View and download the full publication here. This publication appears in “Not So Free to Contract: The Law, Philosophy, and Economics of Unequal Workplace Power,” Journal of Law and Political Economy vol. 3, no. 1, special issue edited by Lawrence Mishel of the Economic Policy Institute.

One common metric of monopsony power is the quit elasticity, measuring how much more likely a worker is to quit a job in response to a wage change. Experimental and quasi-experimental variation in wages across workers within a given job results in quit elasticities in the 2–3 range, implying that a 10% reduction in wages increases the probability of quitting by 20–30%. In a model with monopsonistic employers, a quit elasticity of 2–3 also implies that workers are paid about 80–85% of the value they produce. These results indicate that employer power is pervasive. We present observational evidence that historically disadvantaged groups have systematically lower quit elasticities, indicating they face even greater employer power. Because monopsony power comes from an inability of workers to voluntarily switch jobs, the quit rate and especially the quit elasticity can be a useful metric for judging the health of the labor market. Pervasive employer power alters the analysis of labor market policy in a number of important ways.

I. Introduction

Economists of all political stripes agree that a key restraint on employers in the labor market is the threat of workers quitting for other jobs. In theory, the voluntary flow of workers across firms, and in and out of the labor market, ensures that workers are rewarded according to “the law of one price,” which says that workers with similar characteristics, doing similar jobs, at similar firms, must be paid similar wages. Firms that attempt to underbid other firms should quickly lose all of their employees. Importantly, this logic applies to all attributes of a job, including nonwage benefits and working conditions more generally. When quitting is easy, no abusive bosses can last long, no predatory supervisors can keep their jobs, and other excessively onerous and unpleasant conditions of work will either be compensated for or competed away. Employers will bend over backward to provide amenities and wages to retain labor, and workers will be roughly indifferent across the jobs they can hope to get. As Milton Friedman wrote in Capitalism and Freedom (1962, 19), “the employee is protected from being coerced by his employer by the existence of other employers for whom he can work.”

The easy, and voluntary, flow of workers from job to job is one of the defining features of a competitive labor market. Competition, in turn, implies that firms cannot exercise power over their workers beyond what was agreed to in the employment contract, an assumption that is pervasive in economics, law, philosophy, and political science. The presumption that labor markets are defined by competition shapes both how we think about labor markets and how labor markets are governed both formally and informally. For example, the importance of an employee’s ability to quit is given great prominence in Epstein’s (1984, 947–8) well-known defense of the employee-at-will doctrine, which was clearly articulated in an oft-quoted passage from Payne v. Western & Atlantic Railroad, 81 Tenn. 518–19 (1884):

[M]en must be left, without interference to buy and sell where they please, and to discharge or retain employees at will for good cause or for no cause, or even for bad cause without thereby being guilty of an unlawful act per se. It is a right which an employee may exercise in the same way, to the same extent, for the same cause or want of cause as the employer.

In other words, the argument for at-will employment legislation, and much of existing employment law doctrine, rests on the presumption that both firms and workers can easily terminate an employment relationship. For workers, this presumption means that one can either easily transition to another, equally good job, or that unemployment is not penurious.

In reality, however, finding another job as good as the job one has is not so easy, and quitting a job is a fraught, complicated, and often costly decision. Jobs are complex, idiosyncratic relationships, and individuals value them for reasons that are often only theirs. A friendly coworker or generous boss, an easy commute, proximity to schools, and access to child or elder care all matter in the preferences a worker has for a job. Further, workers do not know about all the offers in the labor market: Most of us are not veteran job hunters, and we depend on a variety of ad hoc mechanisms and heuristics to find out about jobs that are good fits. The result is that job offers are rare, and even in tight labor markets finding a good new job depends a lot on luck in social networks, location, and skills.

The basic fact that work cannot be separated from workers means that limited mobility in response to wage differences across employers is pervasive, and this is true even without colluding or concentrated employers, noncompete agreements, or a variety of policies exacerbating this limitation. The availability of alternative offers, and so the cost of quitting a job, also likely varies by geography, race, and gender, as evidenced by, for example, persistently higher unemployment among Black men (Couch and Fairlie 2010) and the gendered expectations around female mobility.

When quitting a job is costly, employers have power over their employees—a form of power economists call monopsony power. This power allows firms to accept possibly higher turnover in exchange for paying workers lower wages, providing fewer benefits, and allowing worse working conditions, resulting in firms that are too small, pay too little, and have high turnover. This tradeoff is made possible by the fact that, when quitting is costly, workers are less willing to leave their jobs than they should be, and they are thus less sensitive to differences in wages and working conditions across various jobs. They leave jobs slowly, or not at all, thereby reducing the competitive pressure that firms may feel to eliminate discrepancies, and they are vulnerable to the arbitrary whims of their managers and employers.

Importantly, monopsony power does not require anticompetitive artifices such as noncompete agreements, collusion, or concentrated employers to generate employer market power. That is, monopsony power does not come solely from archetypically anticompetitive activities but rather is inherent to how laissez-faire labor markets work and how workers find and value jobs. As Alan Krueger noted in his Jackson Hole keynote address in 2018:

Although economists’ go-to model of the labor market is often one with perfect competition—where bargaining power is irrelevant because supply and demand determine the wage, and there is nothing firms can do about it—in many applications I think it is more appropriate to model the labor market as imperfectly competitive, subject to monopsony-like effects, collusive behavior by firms, search frictions, and surpluses that are bargained over. As a result of these labor market features, firms should be viewed as wage-setters or wage-negotiators, rather than wage-takers.1 (Krueger 2018, 267)

All monopsony requires is that quitting a job is costly, that these costs result in workers being so reluctant to leave their jobs that firms do not have to adjust wages or attributes of a job to keep workers, and that this cost differs from worker to worker in ways that employers may not be able to, or may choose not to, factor into their pay schemes.

An important by-product of monopsony power is that it gives employers latitude to use contractual devices like noncompete clauses to further lock in workers without having to compensate them for forgoing alternative employment (Starr, Prescott, and Bishara 2020). If the labor market were perfectly competitive, noncompete contracts would be perfectly offset by higher compensating wages (or other amenities)—a notion called a compensating wage differential—because workers would never accept the disamenity of forgoing future employment opportunities without being compensated by their current employer. When labor markets are monopsonistic, a noncompete contract will not necessarily be perfectly offset by higher wages, and so it might pay for an employer to use one. One could imagine an inverted-U relationship between noncompete contracts and the degree of monopsony. At one end, where the labor market is a perfect monopsony, there is no need for a noncompete agreement because there is only one employer and therefore nowhere else for workers to go; at the other end, where the labor market is perfectly competitive, there is no scope to introduce a noncompete agreement without losing all ability to recruit unless workers are sufficiently compensated, which would undermine the reason for using the agreement in the first place. But in between is a range of situations where workers may be willing to accept a noncompete agreement without demanding a commensurate increase in pay or benefits.

While noncompete clauses are one particularly important example, monopsony power means that, in general, the specific job and many of the features of a given employment relationship are designed to increase retention, particularly retention of high-productivity workers, potentially at the expense of less-productive workers in the same firm. This means that workplace cultures, tolerance of harassment, and even things like health benefit packages and workplace safety are not provided to satisfy the interest of the firm’s average workers but instead the most productive, or the ones most likely to quit. This potential for disparate treatment creates an important role for institutions of voice, such as collective bargaining, to represent the interests of all workers but most importantly those who are least likely to quit, because it is these workers over whom employers have the most power.

Although monopsony does not rely on the presence of anticompetitive practices like noncompete and antipoaching agreements and simple, naked collusion, anticompetitive practices certainly exacerbate monopsony. There is evidence that these practices are pervasive in today’s labor market (see Block and Harris 2021), and indeed even Adam Smith noted the predilection for employers to collude in setting wages. Yet merger screening policy overwhelmingly focuses on consumer welfare rather than worker welfare, despite the fact that mergers in thin labor markets have a detectable negative effect on wages (Arnold 2019; Prager and Schmitt 2019; Naidu, Posner, and Weyl 2018). New papers are coming out every few months that document negative effects of labor market concentration on wages, with the takeaway that roughly 10% higher concentration (as measured by the Herfindahl index) implies 0.1% to 1% lower wages.

One common method of measuring the degree of monopsony power—the method we use below—is the quit elasticity, a measure of how much more likely a worker is to quit a job in response to a (small) wage change. Monopsony comes, in part, from a reluctance of workers to quit their jobs. This implies a low quit elasticity because being reluctant to quit a job means that the likelihood an individual quits a job does not change very much in response to wage changes (i.e., a small decline in the wage does not increase the probability of quitting, and a small increase in the wage does not decrease the probability of quitting). Conversely, a competitive labor market implies a very high quit elasticity because even small changes in the wage should result in large-scale changes in quits.

Experimental and quasi-experimental variation in wages across workers within a given job results in quit elasticities in just the 2–3 range, implying that a 10% reduction in firm wages increases the probability that the average worker will quit by 20–30%. While this may seem like a lot, high competition theoretically implies the probability of quitting in response to even a small wage decrease should approach 100%. Moreover, trading off lower wages for somewhat higher turnover may be appealing to many firms, especially in situations where the cost of turnover is low. Evidence from survey data suggests quit elasticities that are even smaller, though these estimates are likely biased due to a failure to account for other labor market dynamics changing at the same time that may affect wages and the likelihood of quitting.

What is the cost to workers, and the profit to employers, of quit elasticities in such low ranges? And how might the impact vary between different groups of workers? Less is known on how the degree of monopsony power varies across historically disadvantaged groups, although related evidence on the comparative weakness of labor market outcomes experienced by these groups (Couch and Fairlie 2010) suggests that they should face higher levels of monopsony power as employers choose to leverage the weaker labor markets. Among our calculations (discussed in more detail below) are the following:

• If employers are strictly profit-maximizing, then a quit elasticity of 2–3 implies that workers are paid about 80–85% of the value they produce for the firm (i.e., marginal revenue product), with employers (or other input suppliers) pocketing the difference.

• Monopsony, as measured by lower quit elasticities, is higher in low-wage sectors such as retail and restaurants, for low-wage workers as measured by the bottom quartile of worker fixed-effects, and during business cycle downturns. These workers (and during these periods) may as a result reap an even lower share of their productivity.

• For men, however, the quit elasticity is 0.16 compared to 0.09 for women, implying that men are almost twice as likely to quit a job due to the same-size wage decline as are women.

• For white workers the quit elasticity is 0.12 compared to 0.07 for Black workers, implying that white workers are also almost twice as likely to quit a job in response to the same-size wage decline as are Black workers. Again, that historically disadvantaged groups—particularly Black workers, according to our results—are less sensitive to wage changes implies that firms have more leeway to actively reduce, or fail to increase, their pay because they are less likely to exercise the primary form of power available to them, which is to quit. And, while the elasticity estimates themselves are small compared to quasi-experimental evidence, the comparison of elasticities between groups is still a valid exercise.

Of course, no one would argue that labor markets are truly perfectly competitive, but viewed from this perspective, it is clear that elasticities of the size that have been estimated to date are far from what are expected in a competitive labor market. Even urban labor markets in low-wage, high-turnover sectors like food services or retail have low quit elasticities, despite being precisely the setting where labor markets ought to be competitive because there should be myriad alternative employment opportunities for workers and the workers ought to be easily substituted by firms.

The presence of monopsony suggests a more expanded view of how to judge the health of a labor market. Specifically, because monopsony power comes from an inability of workers to voluntarily switch jobs whenever they want, the voluntary job separation rate (or quit rate) can be a useful metric for judging the health of the economy, in addition to standard measures like the vacancy or unemployment rate. Its usefulness is partly because there is a “job ladder” that is an important source of wage growth during expansions and a major source of wage stagnation during downturns, when the ladder collapses; it is also partly because quits summarize the ability of workers to leave terrible working conditions that are not adequately compensated for. Further, the quit rate is a flow measure of labor market slackness, and so it moves more quickly than stock measures like unemployment, and relying on the former could speed up the timing of policy changes. Combined, these facets of the quit rate imply it should be a macroeconomic indicator that monetary and fiscal authorities pay attention to. If there is pervasive and persistent involuntary unemployment or underemployment, then there is always a “reserve army” of the unemployed available as new recruits, and so employers do not have to worry so much about preventing quits. Markets that do not guarantee full employment cannot guarantee worker freedom of choice or rely on worker exit to constrain employer exploitation.

One caveat to the literal monopsony interpretation of the labor market is that employers do not have to use all the monopsony power they have. For example, some firms may choose to use pay and/or benefits to elicit higher effort and quality or (even) stronger loyalty from their workers, steps that could in certain circumstances increase firm profitability.2 In order to deter shirking on the job, the threat of costly unemployment must be salient, and so monopsonists (particularly those without extensive workplace surveillance technologies) may have to pay wages higher than workers’ outside options. Incorporating effort implies that the output-reducing prediction of simple monopsony may not be borne out empirically, as powerful employers can impose faster work speed as well as lower wages. Similarly, employers can’t perfectly control their managers, and managerial interest in empire building may prevent employers from exercising all their monopsony power. Employers may also choose to set pay in a way that is largely unresponsive to local labor market conditions because, thanks to their power, they can be indifferent to the wage, paying arbitrary wages without worrying too much about losing all their workers. This indifference is revealed by the surprising degree of uniformity of wages paid by the same employer across disparate labor markets.3 Finally, some firms may choose to pay higher wages either in an attempt to increase their applicant pool and hopefully identify the most productive workers, or as an attempt to produce a reputation that consumers (or politicians) might reward.

II. Dynamic monopsony: The microeconomics of quits

Labor economists have long been divided over whether (and which) labor markets are best characterized by perfect competition or by monopsony. The defining feature of a perfectly competitive labor market is that a firm can hire as many workers as it desires at the prevailing wage rate, but also that it cannot hire any workers without paying that rate. Again, this implies that the firm faces an extremely elastic quit rate, or that decreasing wages even slightly below the market rate should result in essentially all of the workers quitting. The quintessential monopsony is a company town, where the only way a firm can hire more workers is to either draw in people already in the town but out of the labor force or draw new people into the town. In a monopsonistically competitive labor market there may be many employers, but it remains the case that, in order to recruit and retain more workers, firms must raise wages in order to overcome the costs workers incur in either entering the labor market or in switching jobs. The monopsony literature—and the term monopsony—dates back to 1933, when Joan Robinson published The Economics of Imperfect Competition (Robinson 1933). Since then, the literature on imperfect competition in labor markets has ballooned, not just in labor economics but also in law and economics, macroeconomics, international trade, and even industrial organization.4

Besides responding to outstanding puzzles in the minimum wage literature, the recent explosion of papers on monopsony reflects both the growing availability of detailed administrative data linking workers and employers and the decreasing cost of experimental interventions in real-world labor markets. Paradoxically, the decline of union coverage, minimum wages, and the willingness of firms to promote from within has made monopsony more visible. Recent empirical work suggests that a firm’s labor supply elasticity, which represents the combined impact of the change in workers quitting and the change in new employees when the wage changes, lies somewhere between 2 and 6, which implies that a 10% decrease in the wage decreases labor supply to the firm by 15–50%.

The sensitivity of firm-specific employment to wages, or the elasticity, is a direct measure of how much wage-setting power employers have. Set wages too low, and an employer will lose all its workers. Set wages too high, and an employer will lose profits. The just-right level of wages will trade off profit per worker with the number of workers staying at the firm. The more sensitive employment is to the wage—i.e., the larger the elasticity—the steeper this trade-off is, so that any attempt to lower worker wages will result in a much larger exodus of workers. In contrast, if this labor supply elasticity is low, then firms can pay workers low wages without having to worry about too many quitting. Variation in the elasticity a firm faces can come from differences in the type of work the firm requires, the location of the firm and the local labor market conditions the firm faces, and the pool of available labor from which the firm can hire.

Direct estimates of monopsony power that are obtained in thick labor markets—where there are both many workers and many employers—offer the most compelling evidence that monopsony is pervasive. Thick labor markets should mean that there are lots of workers looking for a job and lots of employer competition, and that no individual firm is large enough to be able to unilaterally alter conditions in the labor market; thus, labor supply elasticities to any one firm ought to be quite high. Generally, these estimates rely on the fact that a substantial fraction of people who quit employment at one firm become a new employee at another, meaning that there is a tight link between the average quit elasticity and the average labor supply elasticity to that firm. When the firm increases its wages, it both loses few workers and gains new workers, who will be quitting other firms.

Some of the most credible evidence on monopsony power in the United States is provided by the few randomized controlled experiments in which wages are randomized for identical jobs in markets with many wage-setters and little in the way of barriers to entry and movement between firms. Caldwell and Oehlsen (2018) randomized wages for Uber drivers, including those who also drove for Lyft. Examining the rate at which drivers switched between the two, they found a surprisingly low elasticity of between 4 and 5, suggesting that if the implied hourly wage was 10% higher at Uber (Lyft), then 40–50% of workers would switch to Lyft (Uber). Given that workers literally just have to switch apps on their phone, this is a surprisingly low rate of switching in response to a large wage difference. Dube, Jacobs, Naidu, and Suri (2020) experimentally varied wages for an identical task and found substantial monopsony power even on (putatively thick) Amazon Mechanical Turk.5 And, in a nonexperimental setting, Dube, Giuliano, and Leonard (2019), examining a large national retailer that implemented a national wage increase policy pretty much independently of any local labor market conditions, found that quits happened predominantly among those who didn’t get raises while their peers did. This paper illustrates both dimensions of monopsony: the constraints on internal wage differentiation and the extent of labor market competition. Workers quit a lot in response to arbitrary wage differences with their peers, but not so much in response to their own wages alone.6 A meta-analysis of the literature by Sokolova and Sorensen (2021) found an average labor supply elasticity of around 4, implying that a 10% increase in the wage is associated with a 40% increase in employment from a combination of fewer people leaving the firm and new people joining.

Other important evidence on labor supply elasticities comes from firm-specific shocks to labor productivity. In this literature, firms are suddenly hit with a shock that raises the productivity of labor (e.g., a patent is approved, a federal procurement bid is won, or student enrollment at a university goes up). If the employer has to raise or lower firm-specific wages, that suggests they have labor market power, and the resulting change in employment can be used to estimate labor supply elasticities. For example, Lamadon, Mogstad, and Setzler (2022) find a labor supply elasticity between 6 and 6.5 using shocks like this, while Goolsbee and Syverson (2019) find a labor supply elasticity of 1.9 (for full professors) to 7.8 (for assistant professors). One limitation of this literature is that firms can adjust many dimensions of compensation (e.g., hours) in response to productivity shocks (either upward or downward), making isolating the effect of wages difficult.

A. Monopsonistic competition: Evidence from matched worker firm data

In Bassier, Dube, and Naidu (2022), the quit elasticity is estimated by matching workers who are at the same firm and paid the same wage at time t–1 but leave to different firms at time t. They then look at how long workers stay at that next firm as a function of the wage change they got between t−1 and t. Workers lucky enough to land at a high-wage firm will stay longer before leaving again, and workers who land at a relatively low-wage firm will likely leave quickly.

This relationship is summarized in Figure A, reproduced from their paper. It shows a declining relationship, so that the probability of quitting decreases as the wage increases. Faster declines (a steeper slope) mean less monopsony power because workers are more sensitive to the wage change. The magnitude is similar to the other experimental and quasi-experimental estimates described above, where the quit elasticity was about 2 and implied a labor supply elasticity of around 4. So, if the employer decreases wages by 10%, the probability of quitting rises by 20%, it becomes 20% harder to get new recruits, and so overall labor supply to the firm falls by 40%. Quits matter, but they are hardly adequate as the sole device for restraining the capricious or exploitative whims of employers.

Figure A

Further, the connection between aggregate labor market tightness and degree of employer power can be seen in differences in the labor supply elasticity across periods and across groups of workers (Figure B). Recessionary periods with higher unemployment have lower labor supply elasticities, implying more employer power, which makes sense because there are fewer alternative employment opportunities: fewer workers are quitting, and firms are hiring fewer workers. Low-wage workers have lower labor supply elasticities, implying that they face greater employer power. This is perhaps prima facie surprising because low-wage workers have higher turnover in general. But this turnover seems to be not particularly sensitive to the wage, and indeed one can imagine many competing risks, besides those that are wage-related, driving quit behavior for workers facing volatile schedules and other life shocks.

Figure B

B. Heterogeneity: Evidence from the SIPP

The prospect of heterogeneity in monopsony power by gender dates to Joan Robinson’s early work on monopsony, which found women being paid less not because of lower productivity but because of social norms and intra-household bargaining power that dictated that wives move for their husbands and not vice versa. Employers, even nonsexist ones, would lower women’s wages relative to men’s, knowing that women’s outside options were circumscribed. A similar argument would apply to other ascriptive characteristics like race or even education: Employers are looking for “tags” for workers having better or worse outside options, and to the extent the law permits them to condition on those tags, they will. Even a nonracist employer would pay Black workers less, knowing that Black workers have both fewer other job options (as long as some racist employers exist in the same market) as well as lower average wealth and access to credit.7

We can explore worker-level heterogeneity a bit more in depth using the Survey of Income and Program Participation (SIPP).8 Here, we estimate the probability of voluntarily leaving one’s job for either another job or for nonemployment in each month for a given wage level in the previous month, while also adjusting both wages and the probability of quitting for human capital, demographic, and geographic factors.9 One issue to note is that both wages and the probability of quitting are, to some extent, affected by idiosyncratic worker-specific characteristics. That is, they are both endogenous, meaning that these estimates should not be taken as evidence that the causal effect of changes in the wage on the probability of quitting is different across groups, but rather that heterogeneity is present and worthy of future research. The following figures report the (residual) relative probability of quitting for each (residual) wage level, where wages are measured relative to the average (residual) wage in the group and the probability of quitting is measured relative to the probability of quitting for the middle-wage group. Residualizing wages and quits adjusts for differences in human capital and demographic characteristics, ensuring that comparisons are made among people with similar characteristics.10

Figures C–F show how the quit elasticity varies across the wage distribution (adjusted for covariates like education, age, geography) for all workers (Figure C) and by gender (Figure D), race/ethnicity (Figure E), and urban versus rural (Figure F). A comparison of slopes across groups gives an estimate of how monopsony power varies among them. Flatter slopes represent higher levels of monopsony power, because they imply that the probability of quitting is less sensitive to changes in the wage.

Figure C

Figure D

Figure E

Figure F

The results are qualitatively similar to those seen in Figure A, though measured in different units. Specifically, the wage (horizontal) axis shows the percent difference between a given wage level and the average wage (which is defined as zero), and the quit rate (vertical) axis shows the percent difference in the probability of quitting between a given wage level and the average wage. Broadly, the quit elasticity tends to be smaller for low and high wages and larger for the middle of the wage distribution, as evidenced by the fact that the lines are largely horizontal in low (high) wages, indicating that the probability of quitting is similar across a wide range of wages (though much higher for low wages than for high wages). In the middle of the distribution monopsony power is lower, since this is the area where the probability of quitting declines as the wage increases. For example, Figure D shows that at a residual wage of -0.6 (an adjusted wage that is 60% below the average), the adjusted probability of quitting for men is about 20% higher than at the average wage (residual wage equals zero) but only 11% higher for women. Normalizing the adjusted probability of quitting to zero at the average wage for men and women shows that the probability of quitting falls faster for men as the wage increases in the bottom half of the wage distribution than it does for women. That is, for low-wage jobs an increase (decrease) in the wage decreases (increases) the probability of quitting more for men than it does for women. Similar patterns hold for Black versus white workers (Figure E) and urban versus rural workers (Figure F).

Figure G shows the quit elasticities implied by the slopes in Figures C–F, confirming that disadvantaged groups have lower quit elasticities and thus face more monopsony power. Variation across groups is meaningful and shows that employers have more power over those disadvantaged in the labor market (women, Blacks, and Hispanics). Overall, the implied quit elasticities in the SIPP—a 10% wage decrease resulting in a 1.2% increase in the probability of quitting (i.e., a quit elasticity of 0.12) are on the low end of the range seen in the literature using survey data. These are also very small relative to estimates identified using exogenous variation, but in any case nowhere close to the very high numbers (say 5 or higher) that perfect competition would imply. While the estimated quit elasticity is likely biased in this exercise, the estimates are meant to illustrate the differences across groups, as the wage variation here is clearly endogenous to many omitted factors, unlike the quasi-experimental estimates discussed above.

Figure G

Men as a whole have the highest elasticity, with a 10% increase in the wage resulting in a 1.6% decrease in the probability of quitting, while Black individuals face the most monopsony power, with a 10% wage increase associated with a 0.7% decline in the probability of quitting. So, while elasticities are low for all groups, meaning employer power is pervasive, the results show employers have more power over disadvantaged groups in the labor market (women, Blacks, and Hispanics) than advantaged groups (whites and men). Again, the estimated elasticities are certainly biased toward zero due to important omitted variables that are correlated with both the wage and whether an individual quits a job, but we believe that comparisons across groups are still valid, at least for illustrative purposes.

C. Job ladders: The macroeconomics of quits

Recent research affirms the important role of quitting and labor market competition in driving wage growth. Traditionally, we measure the business cycle with unemployment, or the stock of workers out of a job who are actively looking for one. But this measure of the state of the labor market, while important, misses how tight labor markets help all workers by raising the degree of labor market competition. The job ladder (Moscarini and Postel-Vinay 2016), mentioned above, implies that during periods of expansion even employed workers get more and more offers for better jobs, leaving their employers no choice but to either raise wages or recruit new workers. At first employers recruit from the unemployed, and thus may not need to raise wages much, but once the pool of unemployed starts falling, employers begin poaching from each other, driving wage growth. Indeed, Moscarini and Postel-Vinay (2018) show that nominal wage growth responds to the quit rate even controlling for the unemployment rate, which supports the notion that the quit rate is an independent signal of a strong labor market along with the unemployment rate. A 1% decrease in the unemployment rate raises nominal wage growth by about 1 percentage point—the classic “wage curve” relationship between unemployment and wages. However, the authors also find that, even controlling for unemployment, a 1% increase in the employer-to-employer rate (i.e., leaving a job for another job, an admittedly imperfect proxy for quits) increases nominal hourly wage growth by 4 percentage points, and even more for people who actually switch jobs. The fall in the quit rate of roughly 0.5 over the Great Recession is thus associated with about a 2% fall in real wage growth. This pattern suggests that macroeconomic measurement may want to gauge tightness of the labor market based not just on the overall unemployment rate but also on the frequency of quits, in particular quits in response to wage differences across firms.

We have some spotty historical measures of quits, at least at an aggregate level, for manufacturing. Figure H shows the quit rate in manufacturing from two sources, the Historical Statistics of the United States (HSUS) and the Job Openings and Labor Turnover Survey (JOLTS), as well as the overall unemployment rate (from HSUS and the Bureau of Labor Statistics).11 If you want to see what a really tight labor market looks like, consider World Wars I and II, when the quit rate was the highest it has ever been, coincident with historically low unemployment rates. This pattern is consistent with the high pressure of the wartime economy and suggests that the labor market has never been as friendly to worker choice as it was then, even during the recent recovery.

Figure H

Similarly, the lowest level of quits in manufacturing was observed during the historically highest rates of unemployment during the Great Depression. While we don’t have quit data for the late 19th century, the presence of substantial unemployment in the 20 years prior to World War I suggests that, even during the archetypical laissez-faire, freedom-of-contract period of Lochner, worker choice was circumscribed by labor market slack. Figure I, which shows the relationship between quits and the unemployment rate for the whole economy, not just manufacturing, confirms that quits rise substantially when unemployment drops or is low, and it shows that the positive correlation between quits and low unemployment remains true today. Times when most of the labor force has a job are also times when the employed have the most freedom to leave their jobs.

Figure I

However, the very fact of cyclicality suggests that the market does not reliably deliver freedom in the workplace. The business cycle is as capricious as and less accountable than any monarch. If quitting is the primary force that disciplines employer behavior, then the cyclicality of quits suggests that employers are not always tightly constrained by market forces. In fact, employers are rarely constrained by market forces as the economy is rarely in periods of sustained low unemployment. As long as there is structural involuntary unemployment, employers will have the whip hand in setting wages and working conditions. The very fact of booms and busts tracking the quitting behavior of workers suggests that the real freedom experienced by workers, measured by the job options available rather than the current job, is higher when the economy is running hot, and choices are dramatically limited when unemployment is high. This is particularly true for historically disadvantaged groups and now generally true for less-educated workers, because of both the persistently higher rates of unemployment faced by these workers and the higher cyclicality in employment among these groups (particularly for men).

Some evidence on the sources of differential Black unemployment is provided in Couch and Fairlie (2010), who looked at matched Current Population Survey data on transitions from unemployment (and nonparticipation) for Black and white workers. They found that the gap in racial unemployment rates is highest at the lows of the business cycle and that little of this is explained by standard human capital variables like education and age. Thus, tight labor markets likely play an important role in mitigating racial inequality.

One practical way to implement these labor market proxies for economic freedom into macroeconomic policymaking is to use the quit rate for marginalized groups to augment measures of labor market slack, as mentioned above. Full employment becomes measured not just by the number of people looking for work who are unable to get it, but also by the real freedoms experienced by workers with jobs, proxied by their readiness to exercise exit.

One disadvantage of this measure is that the raw quit rate alone doesn’t distinguish between workers who don’t quit because they have a great job and those who don’t quit because they have terrible outside options. For example, Reich and Prins (2020), using data from the National Longitudinal Survey of Youth, show that prior incarceration reduces the responsiveness of quitting to job satisfaction. While people with prior incarceration are more likely to quit in absolute terms as their jobs worsen or options improve, their response has less to do with job conditions than it does for individuals without any experience of incarceration. For reasons such as this, raw quit rates should be supplemented by estimates of the quit elasticity, which manipulate the quality of the job or the outside options and measure the quit response.

But on average, the job ladder/dynamic monopsony model implies that increases in the quit rate are also likely to track the quit elasticity. Just by the fact that quits are most common at low-wage jobs and much rarer at high-wage jobs, the average level of quits is likely, in practice, to be correlated with the elasticity of labor supply facing the firm: when quits are common, the gap in quits between high- and low-wage firms is much higher, so the elasticity is higher. This is consistent with the procyclicality of the quit elasticity measured above. Why? Because if you’re at a firm that pays a very high wage for a given job, you aren’t likely to quit, because there aren’t many offers that are better than the job you have. So, the bulk of quitting is happening for workers at lower wages. Since the number of quits at the top of the labor market is small and pretty constant, the business-cycle quit variation is likely driven by workers leaving low-wage firms, and so the difference in quits between low-wage and high-wage firms, i.e., the sensitivity of quits to wage differences, is highest when the overall quit rate is high.

Some preliminary evidence suggests that this combination of high tightness with high unemployment happened during the post-COVID-19 Great Resignation, in which quits were high, the quit elasticity was high,12 and real wages were growing at the bottom of the distribution, even as unemployment was elevated (if declining).

D. Implications of real-world quit elasticities and employer power

There is a long history of the presumption of competitiveness in labor markets being used to weaken legal protections for workers. Bagenstos (2020) presents evidence that the presumption of no employer power is still pervasive in the law and has the adverse impact of undercutting the constitutional, statutory, and common law basis of workplace protections. Specifically, he examines the Lochner-era freedom-of-contract assumptions being used to justify at-will employment, forced arbitration, and limited enforcement of health and safety regulation. Essentially, almost any employment condition that a worker will accept is presumed to be the outcome of competition in the labor market, where contracts offered by employers can readily be turned down by workers. This presumption was illustrated in the U.S. Supreme Court’s recent decision on forced arbitration in Epic Systems Corp. v. Lewis, 138 S. Ct. 1612 (2018), in which the majority opinion, written by Justice Gorsuch, invokes freedom of contract in the first sentence. Were the forced arbitration agreements genuinely bilateral? Petitioner Epic Systems e-mailed its employees an arbitration agreement requiring resolution of wage and hours claims by individual arbitration. If the employees “continue[d] to work at Epic,” they would be “deemed to have accepted th[e] Agreement” (at 1649). The underlying presumption was that workers could readily quit to another job as good as the one they had at Epic. The dissent by Justice Ginsburg focused on the employer-employee power imbalance: “To explain why the Court’s decision is egregiously wrong, I first refer to the extreme imbalance once prevalent in our Nation’s workplaces, and Congress’ aim in the NLGA [Norris-LaGuardia Act] and the NLRA [National Labor Relations Act] to place employers and employees on a more equal footing” (at 1633).

The opinion in 1905’s Lochner v. New York, 198 U.S. 45 (1905), held that “the freedom of master and employee to contract with each other in relation to their employment, and in defining the same, cannot be prohibited or interfered with, without violating the Federal Constitution” (at 64). In a nutshell, the argument held that, since there was no externality or social problem (in the baking industry), there was no need for the state to use its police power to regulate labor markets. Until weakened by the New Deal, Lochner held considerable sway in preempting labor market regulation.

One could imagine the justification of Lochner being that there was no externality (public health, for example) that would warrant restricting the liberty to contract. But when employers have market power over wages, and worker effort and outside options are difficult to observe, then the wage chosen by the employer is too low, and the chance of a given worker getting hired is also too low, employment is too low, and thus output/revenue is too low. These outcomes are potentially something a sovereign state would have an interest in, allowing use of the police power (particularly when the state is funded out of labor income taxes).

Were Lochner-era labor markets competitive? The clearest evidence of whether this was the case would be quit elasticities estimated on 19th century data, but only indirect evidence on the extent of monopsony in pre-New Deal labor markets is available. In the South, of course, Jim Crow agricultural labor markets were characterized by overt collusion and legal restrictions on competition (Naidu 2010). Looking at the North, Naidu and Yuchtman (2018) use the census of manufacturing establishment samples from the 1850–1880 censuses to estimate the degree of rent-sharing, although the absence of linked worker-firm data makes it difficult to identify cleanly. Nonetheless, the implied labor supply elasticity facing the firm in the 19th century was roughly 2—consistent with a quit elasticity of 1—suggesting that even the unregulated open labor markets of the 19th century did not look frictionless or perfectly competitive. More evidence can be found in Henry Ford’s famous experiment of raising the wage from $2.30 to $5.00; company records indicate that the quit rate fell from 370% to 54% (Raff and Summers 1987). While this is a dramatic fall in quits, it is not so large given the magnitude of the wage change. As a point estimate it would imply a quit elasticity of -0.72, and an arc labor supply elasticity would be roughly 2. These are low and well within the range of contemporary estimates.

The upshot of this calculation of the quit elasticity is that Ford had plenty of labor market power, but, like the Amazon of today, he chose not to use it all. One interpretation (Levy 2021) is that Ford was so obsessed with output and production that he simply chose to not exercise the monopsony power he had in order to implement his Taylorist, high-effort production process. Ford wasn’t just maximizing profits, he was building an empire (and was later sued for this by his shareholders).

Even Lochner-era labor markets were characterized by considerable monopsony power. Employers didn’t have to worry about workers all leaving in the event of a wage cut; indeed, Hanes (1993) argues that it was the institutional memory of violent 19th-century strikes that deterred wage cuts in practice. The dents in this “hired hands” regime really began to appear only with the onset of World War I and the short-lived rash of accompanying unionization, and these early “high road” employers persisted into the 1920s and beyond.

E. Monopsony, exploitation, and freedom at work

When quits aren’t perfectly responsive to wages, employers know that they can pay lower wages, offer worse working conditions, and generally provide a subpar working experience, exactly because they know that not all workers will quit. Sure, some workers will quit, but enough will stay on at the lower wage to make it profitable. This wage/turnover trade-off is one of the core dilemmas facing every employer and human resources manager: How well must you treat your workers in order to keep turnover down? In economic terms, this is dynamic monopsony in action. The upshot is that workers are paid below their marginal product, and outside options (e.g., unemployment insurance, family and friend networks, alternative job offers, and plain old wealth) influence a worker’s wages and working conditions. So the rate of return to skill is depressed, workers invest less in those skills, and employers perpetually complain about the lack of skilled workers.

Measuring monopsony with the quit elasticity gives a simple formula for the (neoclassical) exploitation rate, which is the difference between the wage and the marginal product of labor. If the quit elasticity is Eq, we can approximate the labor supply elasticity by (2 − unemployment rate) × (quit elasticity).13 The fraction of marginal productivity captured by workers can then be approximated by:

where u is the unemployment rate and e < 1 (for effort) is a placeholder for all the other constraints on wage-setting employers are facing, in particular the need to provide incentives, maintain morale, and manage employee behavior. β can be thought of as the degree of bargaining power a worker has and depends on unemployment, the quit elasticity, and countervailing constraints on employer wage-setting as measured by e. As the labor market approaches perfect competition and full employment, (2 − u)Eq gets larger and larger, and β approaches 1. But for the values of the quit elasticity we have, around 2–3, and unemployment around 3–6%, and with no other constraints so that e = 0, in unrestricted monopsony β should be between 0.7 and 0.85, and so workers are receiving only 70–85% of their marginal product. In this clear, empirically operationalizable sense, workers are exploited.

The remaining share of income would show up in some combination of pure profit shares and capital shares. Because book values of investment goods and commercial real estate may capitalize the value of labor market rents, it is difficult to disentangle pure profit from returns on investment. Nonetheless, a recent working paper (Seegmiller 2021), as an example of one of a few recent studies, finds that roughly a third of capital income may be due to monopsony power.

What is a worker’s marginal revenue product? Forget about defining it in some transcendental, economywide way. Empirically it is a simple object: the causal effect of a given worker’s employment on a given firm’s revenue,14 holding all other inputs constant, or the amount of additional revenue a given worker generates for a firm. Neoclassical economists gave the idea of marginal product—the additional output that comes from hiring an additional worker—a bad name because they used it as a measure of the economywide contribution of a worker, independent of particularities of job and industry and social networks and luck. But when market power is pervasive, there is also pervasive misallocation of workers across firms and jobs, and so the notion of a single marginal product ought to be scrapped. At best, it is a thought experiment that lives at the level of a worker-job match, not an economywide concept of value. Both critics and economists alike would be well served by not fetishizing the concept of marginal product beyond that of a simple, context-specific, causal effect. But let’s use it here as an illustration.

Under pretty standard assumptions, we can approximate marginal product with a rescaled average product, so that MPL = θY, where Y is output per worker, i.e., the average productivity of labor, and θ is a scaling factor between 0 and 1 relating average productivity to marginal productivity (say around 0.7 if the production function is Cobb-Douglas).15 If we consider the relationship between productivity and pay and its changing nature over time, we can decompose it into the component due to θ and the component due to β, because the wage w = βθY. The relationship between average productivity and marginal productivity (θ) is determined by things like production technology, organizational change, and structural change in the economy, which are themselves induced by macroeconomic policy, public-good provision, and the effects of international trade. As these have all changed, so has θ. We can include product market power in θ as well. But β, which can be thought of as the fraction of value a worker produces that is paid to the worker, has also changed owing to deunionization, increased employer surveillance, and the collapse of internal labor markets/within-firm equity norms (all of which reduce e and thus reduce β), even if the underlying degree of monopsony, as measured by Eq, has not changed. Further we can imagine that this decomposition is different between college-educated and non-college-educated workers, and so some of the change in wage inequality could be explained by changes in bargaining power as well as changes in productivity.

Stansbury and Summers (2020) argue that the decline of the labor share can be accounted for purely by falling worker rents (which they somewhat confusingly call “worker power,” but rents, defined as payments to workers above their outside options, that can be withheld by employers are in fact employer power [Bowles and Gintis 1992]). Though they discuss unions, the “firm-size premium,” and other sources of rents, technological changes may have made the need to use wages as a labor-disciplining device (efficiency wages) less of a concern than before, so that a component of falling firm rents is improvements in workplace surveillance and easier supervision. (A not-small share of recent patent increases has been in technologies that facilitate the tracking and monitoring of workers.) What the technological elimination of rents has made possible is lower unemployment (i.e., the NAIRU—the non-accelerating inflation rate of unemployment—may have fallen), but also lower wages. Perhaps the increased surveillance and measurement of workers is a contributor to the increased salience of monopsony; the offsetting constraint of labor discipline has become less important. But whether workers have power when wage premia are vulnerable to the discretion of a manager or employer could be quarreled with, and indeed would contradict most discussions of the notion of power. The presence of rents is necessary but not sufficient to diagnose employer power: It depends on who can credibly threaten to deny rents to whom. A union premium is a very different rent from a labor discipline rent because the latter actually represents employers exercising power over workers.

Beyond exploitation, the fact of employer market power gives the capacity for employer domination. When firms have some latitude to choose wages and employment to maximize profits, they have some room for exercising what Pettit (1997) calls “arbitrary whim.” An employer can be generous, and benevolent, maintaining a great work environment, paying high wages, and providing great benefits, but equally an employer can be vindictive and abusive, lowering safety standards, underproviding health care and pensions, and paying low wages. Monopsony implies a capacity for employer domination, where employers can wiggle wages a little bit with little loss of profits but with large effects on worker welfare. Pervasive market power means that the market doesn’t completely constrain the capricious decisions of employers and managers.

The private power of employers has recently become a concern of political theorists and philosophers. Gourevitch (2014) shows that concern with arbitrary employer decisions on wages and shopfloor governance was an important component of the “labor republicanism” of the late 19th-century American labor movement. Along these lines, Anderson (2019) likens one’s ability to quit employers to choosing which communist dictatorship one hopes to work for. Monopsony (and labor discipline) give economic foundations to these writers, supplying empirical and theoretical reasons why employers have scope for exercising political power inside the workplace.

But Cowen (in the same volume) argues that Anderson exaggerates the situation. In his view there is little empirical evidence for the monopsony model of labor markets (i.e., a situation in which a large employer dominates the market) and no evidence therefore of pervasive employer power. That is, Cowen assumes that competitive labor markets, absent labor concentration, restrain employers. Moreover, Cowen argues that even firms with some market power over workers have an economic incentive to provide greater freedoms to workers. Specifically, he claims that firms’ profit motives constrain their ability to exploit workers because firms need to protect their reputations among consumers and retain and recruit workers. Cowen presumes that when employers provide freedoms they can secure a lower wage from employees that reflects the value of those freedoms (based on the assumption that compensation costs are equalized across opportunities, so the greater cost to the employers of any particular employee freedom is offset by the lower wage).

The evidence suggests that Cowen is wrong on the facts. Further, Cowen doesn’t pursue the logic of quality choice under conditions of market power (Spence 1975). Monopsony means that workplace public goods and governance rules (like benefits or, say, harassment and safety policies) are supplied not to maximize the benefit of the average worker at a workplace but rather to retain the “marginal worker,” i.e., the worker most willing to quit or having the widest outside option. So if richer and male and white workers have better outside options, then workplace amenities, including things like culture and training as well as health and other benefits, will be designed to keep those workers from leaving (a situation that could make them either over- or underprovided). But this scenario creates an important role for institutions of voice (like unions) and workplace democracy to represent the interests of average, rather than just marginal, workers in the design and implementation of workplace constitutions.

III. Conclusion

An indicator for the degree of worker power in a laissez-faire labor market is the level of quits, which closely tracks unemployment, but also the sensitivity of quits to changes in wages, the so-called quit elasticity. This measure of competition is a lower bound on the level of freedom workers have in the labor market, and it is consistent with recent work in philosophy theorizing nondomination in markets. Proxying aggregate labor market conditions with quits of historically marginalized populations can go further in measuring full employment than can proxying simply the aggregate stock of unemployment. The fact that quits are not enormously sensitive to changes in wages and job characteristics suggests that competition alone cannot deliver robust protection against arbitrary employer demands and market fluctuations. Rather, a variety of legal rebundling of and restrictions on the rights transacted in labor markets may be required in order to create an environment in which equal power between workers and their employers can be a reality rather than a frictionless fiction.

About the authors

Suresh Naidu (suresh.naidu@gmail.com) is professor of economics and public affairs at Columbia University, external faculty at the Santa Fe Institute, and a research associate with the National Bureau of Economic Research.

Michael Carr (michael.carr@umb.edu) is associate professor of economics at the University of Massachusetts Boston.

Notes

1. Krueger in a footnote (1) went further: “Notice that I don’t call these features ‘imperfections.’ They are the way the labor market works. The assumption of perfect competition is the deviation from the norm of ‘imperfection’ as far as the labor market is concerned.”

2. The classic example of this is the labor discipline model, in which firms voluntarily pay wages above a worker’s next best alternative in order to elicit maximum effort from workers for which effort might be hard to determine (e.g., workers who work in groups, or service providers where quantity is hard to measure).

3. Two pieces of evidence for this: When big national employers change wages, they often don’t announce different wages for different labor markets, but rather blanket policies like a $15 an hour internal minimum wage or a 10 cent increase across all workers contingent on previous wage (Dube, Giulani, and Leonard 2019; Derenoncourt et al. 2021). Secondly, there is a surprising amount of “bunching” in the wage distribution that the availability of paycheck-level data has allowed us to document with much greater certainty than before (Dube, Manning, and Naidu 2019).

4. A recent article reviewing the literature is Manning (2021).

5. Amazon Mechanical Turk (MTurk) describes itself as “…a crowdsourcing marketplace that makes it easier for individuals and businesses to outsource their processes and jobs to a distributed workforce who can perform these tasks virtually.”

6. Emanuel and Harrington (2020) also use nationally uniform wage-setting relative to local differences in pay to recover recruitment and quit elasticities, with implied labor supply elasticities somewhat higher (between 6 and 9), but the variation here partially comes from cross-market variation in outside wages, not from within-firm variation in wages across identical workers.

7. Whatley and Sedo (1998) show that Black workers were less likely than white workers to quit Ford Motor Company despite having worse working conditions (albeit similar pay). Fryer, Pager, and Spenkuch (2013) find offered and current Black wages are 17% lower, even conditional on previous wages, suggesting differences in outside options. Looking directly at survey data, they find that Black workers have reservation wages 7% lower than white workers. In a monopsonistic labor market, the presence of background racial differences in outside options generates racial differences in wages unrelated to productivity at the current job.

8. We use the 2004, 2008, and 2014 panels of the SIPP. The SIPP is a nationally representative survey of the United States. Each panel is longitudinal, lasting three to six years, and each comprises a new representative sample. We limit the sample to individuals between ages 25 and 59, with nonimputed earnings, work hours, and employment status, who were employed in month t−1. We then estimate the probability of quitting between month t and t−1, conditional on the wage in t−1 and additional human capital, demographic, and geographic controls.

9. Specifically, we adjust wages for race, gender, education, residence (or not) in an urban area, and a cubic in age. By voluntary, we mean quitting due to reasons aside from the employer. This includes quitting to take another job, because of working conditions, because of one’s own health or family health issues, because of child care responsibilities, and because of other reasons specific to the person.

10. Residual wages and quits are (log) hourly wages or an indicator for whether an individual quit a job where the effect of education, age, geography, and other observable wage determinants have been removed. Ranking workers using residual wages, instead of the actual wage, ensures that when we compare workers across demographic groups we are making an apples-to-apples comparison. For example, it is well documented that wages increase with age, and it is also likely the case that the probability of quitting one’s job decreases with age. If, for example, Black male workers are on average younger than white male workers, then comparison of Black and white men using observed wages might be biased by the fact that both race and age are different between the two groups. Using residual wages solves this problem.

11. The unemployment concept implemented prior to 1938 is different, but a harmonized series is provided by HSUS. See Carter (2006).

12. See, e.g., Dube 2021.

13. This follows from Chapter 4 of Manning (2003) under the assumption that the share of recruits from nonemployment is equal to the unemployment rate, that there are no voluntary separations to unemployment, and that the elasticity of recruits from unemployment is 0, all of which are plausible simplifying assumptions.

14. We use marginal product as a placeholder for both quantity-based marginal product as well as marginal revenue product, as we’re not discussing market power in product markets here.

15. To be distinguished from the Marxian exploitation rate, which is the difference between the wage and the average product of labor. See Elster 1978 for a discussion.

References

Anderson, Elizabeth. 2019. Private Government: How Employers Rule Our Lives (and Why We Don’t Talk About It). Princeton University Press.

Arnold, David. 2019. Mergers and Acquisitions, Local Labor Market Concentration, and Worker Outcomes. Economics Department, Princeton University.

Bagenstos, Samuel. 2020. Lochner Lives On: Lochner Presumption of Equal Power Lives in Labor Law and Undermines Constitutional, Statutory, and Common Law Workplace Protections. Economic Policy Institute.

Bassier, Ihsaan, Arindrajit Dube, and Suresh Naidu. 2022. “Monopsony in Movers: The Elasticity of Labor Supply to Firm Wage Policies.” Working Paper no. 27755. National Bureau of Economic Research.

Block, Sharon, and Ben Harris. 2021. Inequality and the Labor Market: The Case for Greater Competition. Brookings Institution.

Bowles, Samuel, and Herbert Gintis. 1992. “Power and Wealth in a Competitive Capitalist Economy.” 21 Philosophy & Public Affairs 324.

Caldwell, Sydnee, and Emily Oehlsen. 2018. “Monopsony and the Gender Wage Gap: Experimental Evidence from the Gig Economy.” Working Paper. Massachusetts Institute of Technology. https://app.scholarsite.io/sydnee-caldwell/articles/monopsony-and-the-gender-wage-gap-experimental-evidence-from-the-gig-economy.

Carter, Susan B. 2006. “Labor Force.” In Chapter Ba of Historical Statistics of the United States, Earliest Times to the Present: Millennial Edition, edited by Susan B. Carter, Scott Sigmund Gartner, Michael R. Haines, Alan L. Olmstead, Richard Sutch, and Gavin Wright. Cambridge University Press.

Couch, Kenneth A., and Robert Fairlie. 2010. “Last Hired, First Fired? Black–White Unemployment and the Business Cycle.” 47 Demography 227.

Derenoncourt, Ellora, Clemens Noelke, David Weil, and Bledi Taska. 2021. “Spillover Effects from Voluntary Employer Minimum Wages.” Working Paper no. 29425. National Bureau of Economic Research.

Dube, Arindrajit (@ArinDube). 2021. “This is preliminary, but….” Twitter, October 29, 2021, 6:43 p.m. https://twitter.com/arindube/status/1454217257765126146.

Dube, Arindrajit, Laura Giuliano, and Jonathan Leonard. 2019. “Fairness and Frictions: The Impact of Unequal Raises on Quit Behavior.” 109 American Economic Review 620.

Dube, Arindrajit, Jeff Jacobs, Suresh Naidu, and Siddharth Suri. 2020. “Monopsony in Online Labor Markets.” 2 American Economic Review: Insights 33.

Dube, Arindrajit, Alan Manning, and Suresh Naidu. 2019. “Monopsony and Employer Mis-Optimization Account for Round Number Bunching in the Wage Distribution.” Unpublished manuscript.

Elster, J. 1978. “Exploring Exploitation.” 15 Journal of Peace Research 3.

Emanuel, Natalia, and Emma Harrington. 2020. “The Payoffs of Higher Pay: Elasticities of Productivity and Labor Supply with Respect to Wages.” Working Paper. Harvard University.

Epstein, Richard A. 1984. “In Defense of the Contract at Will.” 51 University of Chicago Law Review 947.

Friedman, Milton. 1962. Capitalism and Freedom. University of Chicago Press.

Fryer, Roland G., Jr., Devah Pager, and Jörg L. Spenkuch. 2013. “Racial Disparities in Job Finding and Offered Wages.” Journal of Law and Economics 56, no. 3: 521–53.

Goolsbee, Austan, and Chad Syverson. 2019. “Monopsony Power in Higher Education: A Tale of Two Tracks.” Working Paper no. 26070. National Bureau of Economic Research.

Gourevitch, Alex. 2014. From Slavery to the Cooperative Commonwealth: Labor and Republican Liberty in the Nineteenth Century. Cambridge University Press.

Hanes, C. 1993. “The Development of Nominal Wage Rigidity in the Late 19th Century.” 83 American Economic Review 732.

Krueger, Alan B. 2018. “Luncheon Address: Reflections on Dwindling Worker Bargaining Power and Monetary Policy.” Proceedings of the Jackson Hole Economic Policy Symposium, “Changing Market Structures and Implications for Monetary Policy,” Jackson Hole, Wyo., August 23–25. https://www.kansascityfed.org/documents/6984/Lunch_JH2018.pdf.

Lamadon, Thibaut, Magne Mogstad, and Bradley Setzler. 2022. “Imperfect Competition, Compensating Differentials, and Rent Sharing in the U.S. Labor Market.” 112 American Economic Review 169.

Levy, J. 2021. Ages of American Capitalism: A History of the United States. Random House.

Manning, Alan. 2003. Monopsony in Motion: Imperfect Competition in Labor Markets. Princeton University Press.

Manning, Alan. 2021. “Monopsony in Labor Markets: A Review.” 74 ILR Review 3.

Moscarini, Giuseppe, and Fabien Postel-Vinay. 2016. “Did the Job Ladder Fail After the Great Recession?” 34 Journal of Labor Economics S55.

Moscarini, Giuseppe, and Fabien Postel-Vinay. 2018. “The Cyclical Job Ladder.” 10 Annual Review of Economics 165.

Naidu, Suresh. 2010. “Recruitment Restrictions and Labor Markets: Evidence from the Postbellum U.S. South.” 28 Journal of Labor Economics 413.

Naidu, Suresh, Eric A. Posner, and Glen Weyl. 2018. “Antitrust Remedies for Labor Market Power.” 132 Harvard Law Review 536.

Naidu, Suresh, and Noam Yuchtman. 2018. “Labor Market Institutions in the Gilded Age of American Economic History.” In Oxford Handbook of American Economic History, Vol. 1, edited by Louis P. Cain, Price V. Fishback, and Paul W. Rhode, 329. Oxford University Press.

Pettit, Philip. 1997. Republicanism: A Theory of Freedom and Government. Clarendon Press.

Prager, Elena, and Matthew Schmitt. 2019. “Employer Consolidation and Wages: Evidence from Hospitals.” Working Paper. Washington Center for Equitable Growth.

Raff, Daniel M.G., and Lawrence H. Summers. 1987. “Did Henry Ford Pay Efficiency Wages?” 5 Journal of Labor Economics S57.

Reich, A.D., and S.J. Prins. 2020. “The Disciplining Effect of Mass Incarceration on Labor Organization.” 125 American Journal of Sociology 1303.

Robinson, Joan. 1933. The Economics of Imperfect Competition. Palgrave Macmillan.

Seegmiller, B. 2021. “Valuing Labor Market Power: The Role of Productivity Advantages.” Working Paper. https://www.bryanseegmiller.com/publication/jmp/.

Sokolova, Anna, and Todd Sorensen. 2021. “Monopsony in Labor Markets: A Meta-Analysis.” 74 Industrial and Labor Relations Review 27.

Spence, A. Michael. 1975. “Monopoly, Quality, and Regulation.” 6 Bell Journal of Economics 417.

Stansbury, Anna, and Lawrence H. Summers. 2020. “Declining Worker Power and American Economic Performance.” 2020 Brookings Papers on Economic Activity 1.

Starr, Evan, James J. Prescott, and Norman Bishara. 2020. “Noncompete Agreements in the U.S. Labor Force.” 64 Journal of Law and Economics 53.

Whatley, Warren C., and Stan Sedo. 1998. “Quit Behavior as a Measure of Worker Opportunity: Black Workers in the Interwar Industrial North.” 88 American Economic Review 363.

Worker mobility—the ability to find and take another job—is at the core of worker power, and, conversely, worker immobility is at the core of employer power. But how easy is it for a worker to leave a job and look for another? 

In this paper, we present evidence of barriers to worker mobility along two dimensions: labor market considerations (can a worker find another job?) and financial considerations (can a worker afford to transition to another job?).

With regard to labor market barriers, we assess each step in the job search and job match process and find that worker mobility is greatly limited by the availability of jobs to which workers can move; the time it takes to search for and secure another job, if it’s available; and the quality of the available jobs and quality of the new job, if secured. Moreover, limitations in labor market mobility are often dependent on the current job: Does it have hours or working conditions that make on-the-job search difficult? Does it have the scheduling or time-off policy that supports interviewing? Does it provide a positive or negative signal about the worker? The existence of these kinds of constraints suggests that worker immobility can be a reinforcing process. Finally, mobility is contingent on the degree of labor market discrimination. We illustrate, in a stylized model, how a Black worker would need to devote nearly four times the effort to receive the same number of offers as a similar white worker.

With regard to financial barriers to mobility, we find that, even in a job-to-job transfer—the best-case scenario for worker mobility because unpaid time off is minimized—workers can experience gaps in benefit coverage or compensation. Even noncompensation aspects of a new job can carry financial costs, such as commuting, and a change in job location can require a new child care arrangement. Workers facing unemployment spells between jobs must usually finance the spells on their own, a major hurdle for the huge share of U.S. households whose savings add up to just several hundred dollars. Workers willing to enhance their labor market prospects by moving must not only pay the costs of the move but also have access to considerable savings to cover advance rent and security deposits or to cover expenses while awaiting the sale of a home. And much like in the labor market, there is persistent and well-documented discrimination in the housing market that raises these costs for people of color. Finally, access to credit can smooth out job transition costs, but it is not universal and reflects clear racial differences. Black and Hispanic Americans also have considerably less wealth to tap into than white Americans.  

Our assessment of these and other labor market and financial considerations illustrates the extent to which barriers to mobility can make moving jobs a luxury, rather than a right. The theoretical context of these findings is dynamic monopsony: the harder it is for a worker to leave, the more power an employer has over that worker’s wages.

I. Introduction

How easy is it for a worker to leave a job and look for another? In the “at-will” employment framework, an employer cannot legally prevent a worker from leaving a job, but what does the concept say about the right, or even the ability, to find a new one?

In this paper, we present evidence of barriers to worker mobility along two dimensions: labor market considerations (can a worker find another job?) and financial considerations (can a worker afford to transition to another job?). With regard to labor market barriers, we assess each step in the job search and job match process and find:

• Worker mobility is greatly limited by the availability of jobs to which workers can move; the time it takes to search for and secure another job, if it’s available; and the quality of the available jobs and quality of the new job, if secured.

• Limitations in labor market mobility are often dependent on the current job: Does it have hours or working conditions that make on-the-job search difficult? Does it have the scheduling or time-off policy that supports interviewing? Does it provide a positive or negative signal about the worker? The existence of these kinds of constraints suggests that worker immobility can be a reinforcing process.

• Mobility is contingent on the degree of labor market discrimination. We illustrate, in a stylized model, how a Black worker would need to devote nearly four times the effort to receive the same number of offers as a similar white worker.

With regard to financial barriers to mobility, we find:

• Even in a job-to-job transfer—the best-case scenario for worker mobility because unpaid time off is minimized—workers can experience gaps in benefit coverage or compensation. Even noncompensation aspects of a new job can carry financial costs, such as commuting, and a change in job location can require a new child care arrangement.

• Workers facing unemployment spells between jobs must usually finance the spells on their own, a major hurdle for the huge share of U.S. households whose savings add up to just several hundred dollars.

• Workers willing to enhance their labor market prospects by moving must not only pay the costs of the move but also have access to considerable savings to cover advance rent and security deposits or to cover expenses while awaiting the sale of a home. And much like in the labor market, there is persistent and well-documented discrimination in the housing market that raises these costs for people of color.

• Access to credit can smooth out job transition costs, but it is not universal and reflects clear racial differences. Black and Hispanic Americans also have considerably less wealth to tap into than white Americans.

Our assessment of these and other labor market and financial considerations illustrates the extent to which barriers to mobility can make moving jobs a luxury, rather than a right. The ability, or inability, to find new employment has both practical and theoretical implications. Workers would likely call a lack of job options a real hardship and a barrier to higher wages, better living standards, and basic economic security. Economists would call an inability to search for or find a new job an example of a search friction and trace its origins and effects in a theoretical framework. Indeed, worker mobility is at the core of worker power, and, conversely, worker immobility is at the core of employer power. Hence, we start with a discussion of monopsony and how researchers define and identify the concentration of employer power.

II. Monopsony in practice

Worker mobility—the ability to find and take another job—exists on a spectrum, from those who can easily move to those who cannot. What explains the variation? One could ascribe the reason to the worker and his or her education, experience, occupation, or industry. But that is just another way of saying that workers with different educations, experience, occupations, or industries face different labor markets, and, again, some labor markets have more mobility than others. Hence, the real question is, what explains the variation in mobility in labor markets?

Perfect monopsony is defined as the presence of a single buyer in a market (the inverse of a perfect monopoly, in which there is a single seller). In labor markets, monopsony, or monopsonistic power, is the term given to describe a market-disrupting concentration of employer power that allows the employer to pay workers less than a competitive wage. A competitive wage is a worker’s marginal productivity; paying less than a competitive wage means paying a worker less than what the worker contributes to the firm. We often conceptualize monopsony in a static framework, in which a single dominant employer—a mining company or a manufacturing plant in a rural area—controls the labor market. But researchers have found diffuse and pervasive evidence of monopsony beyond this static framework; workers are paid below a competitive wage in many geographies, markets, and industries (Ashenfelter, Farber, and Ransom 2010; Dube, Manning, and Naidu 2018; Sokolova and Sorensen 2021; Stansbury and Summers 2020; Webber 2015). Instead, concentration of employer power today is conceptualized as dynamic monopsony (Manning 2003, 2021; Naidu and Carr 2022), in which the difficulty or inability to leave a position that pays less than a competitive wage gives an employer monopsonistic power to lower wages.

There are many potential contributors to dynamic monopsony. Card (2022) discusses the concentration of employers in labor markets, collusive “no poaching” agreements between firms, and noncompete clauses attached to hiring contracts. The latter two can be interpreted as evidence of monopsony in addition to contributors to monopsony. Moreover, all of these contributors have a common theme of restricting worker movement, either through the reduced number of alternative employers, reduced access to those employers, or restricted movement to those employers. Hence, the crux of monopsony is the limitation of a worker’s outside options and the ability to exercise them (Schubert, Stansbury, and Taska 2021). Worker mobility, then, can be seen as a study of search friction—anything that prevents a worker from immediately starting a new job, from the time needed to find another position to the ability to accept it. Because it is hard for workers to leave, and employers know this, they can exercise monopsonistic power and pay lower than competitive wages.

Often, researchers look for evidence of monopsony in observed wages and the distribution of wages across firms, time, industries, or geographies that reflect rising employer power or the decline in worker power. The proof of monopsonistic power is found in the resulting wages. In this paper, we take the reverse approach. Rather than look for evidence of the result, we assess evidence of the cause, namely, search frictions. Are workers mobile? Are they able to move jobs? We take a practical approach to this question and discuss the many barriers and difficulties workers face in leaving a job.

III. Labor market constraints

The primary component of worker mobility is whether the worker can find a job. The job-finding process is conceptually straightforward: workers decide to look for a job, perform the job search, and, if another job is found, decide whether to accept it. However, each of these steps carries its own assumption: that there is a job available, that the worker has the ability to dedicate time to search for it, and that a new job will be of sufficient quality and not result in disruptions to the worker’s income or standard of living. As we will show, these assumptions are not trivial.

A. Available jobs

To start, we consider the first step in the job-finding process and its associated assumption: workers decide to look for a job, which assumes that another job is available. The primary barometer of job availability is the unemployment rate, which measures the share of workers who are not employed but actively looking for a job. There are numerous means of assessing job availability, but most derive from, or incorporate in some way, the number of unemployed workers. Unemployment itself, or a positive rate of unemployment, is not an indication of no available jobs; it can take time to look for and decide on a position. Even in a market with many available jobs, some unemployment would be expected. However, elevated unemployment rates, or prolonged unemployment durations, are stronger indicators of job unavailability.

Figure A shows the job seeker ratio—the number of unemployed workers in a given month as a share of the number of job openings—from 2000 to 2021. A ratio greater than 1 indicates there are more unemployed workers than available jobs. The ratio spikes during recessions (peaking here at 6.5 in July 2009) and falls during recoveries.

Figure A

Job availability is not a fixed feature of the labor market and instead fluctuates with the business cycle. In the 252 months of available data since December 2000, there have been just 33 in which jobs outnumbered searching workers, or just 13% of the time. Indeed, periods of full employment are rare. When unemployment is higher, workers are less likely to quit their jobs, transitions to new jobs are fewer, and wage growth is lower (Mishel 2022). At the same time, employers can fill vacancies with less effort and even raise expected education and experience requirements for new hires (Modestino, Shoag, and Ballance 2020). Varying job availability means worker mobility is dependent on the broader economy, which is far beyond the control or influence of a single worker.

Trends in the broader economy, though relevant, may matter less than a worker’s local labor market. Although the U.S. is an expansive labor market of over 150 million workers and boasts nearly every combination of industry and occupation, workers are mostly confined to their local labor markets, which comprise the jobs they are qualified for within the geographical boundary of where they are willing to work. (Workers could move across labor markets for a job, but, as we will discuss, that is not financially feasible for many). Local labor markets should not be conflated with a city or locality; rather, a local labor market is defined by where a worker can commute on a daily basis. An individual in a rural area might spend an hour commuting 50 miles to a job, across many towns, while an individual in an urban area might commute for an hour by bus within the same city or borough. For reference, 12% of workers in 2019 had a commute time of less than 10 minutes, and 10% had a commute time of over an hour (Burd, Burrows, and McKenzie 2021).

Two examples illustrate some of the constraints of local labor markets and how they can exhibit monopsonistic features, regardless of size. In Janesville, journalist Amy Goldstein wrote about the 2008 closing of the General Motors manufacturing plant in Janesville, Wis., a town of around 60,000 (Goldstein 2018). The plant employed about 1,200 workers when it closed, and it was not the largest employer in Janesville and the workers did not represent a majority or even a large share of the total labor force. Yet, for the GM workers there were no alternative employers in the city who paid a similar wage for workers with their education and experience levels. Few of the laid-off workers returned to manufacturing in Janesville after the plant closed. Some took retirement buyouts; some commuted weekly to another GM plant hundreds of miles away, returning home to Janesville each weekend; and others pursued retraining at the local community college. Those different jobs, however, did not pay wages at the GM level.

In No Shame in My Game, anthropologist Katherine Newman studied the employees of a fast-food restaurant in Harlem over a two-year period (Newman 2000). Though New York City is one of the largest and wealthiest cities in the world, with millions of jobs, the job options for the Harlem workers were not as vast as the city itself. Most walked to work and could not afford daily subway fare, most had no education beyond a high school degree, and most had little prior job experience. Thus, the fast-food restaurants within their neighborhood were one of their only employment options. With low wages and long hours (at least in the time in which Newman was studying), the workers struggled to find the time or money to finance a job search outside their neighborhood.

The narratives of the laid-off auto workers in Janesville and the fast-food workers in Harlem illustrate what research has found to be the case: worker outcomes are conditioned on the local labor market, and one’s labor market does not have the same geographic boundaries that demarcate cities or counties (Enrico 2011; Manning and Petrongolo 2017). Further, the determinants that make a labor market more or less favorable to workers in terms of availability of jobs or level of pay are vast and outside of the worker’s control. For example, a Walmart opening up in a county is associated with a reduction in both retail employment and retail earnings (Neumark, Zhang, and Ciccarella 2008; Wilshire 2022). Comparing across local labor markets, researchers have found differences in employer concentration and market power, and those differences are associated with differences in worker outcomes: markets with more employer power have lower wages and more wage inequality (Azar et al. 2020; Benmelech, Bergman, and Kim 2020; Rinz 2018). An individual worker would have little influence on these and other features of the local labor market, but those features would greatly influence the worker’s own earnings and mobility.

It is possible that the importance of local labor markets will erode in the future as more workers work remotely and are therefore not tethered to the labor market in which they reside. In the beginning of the COVID-19 pandemic in May 2020, efforts to contain community spread via social distancing resulted in 35.4% of workers in the U.S. working from home (Bureau of Labor Statistics 2020). Researchers estimate that 37% of jobs in the U.S. economy could be performed at home (Dingel and Neiman 2020). Before the pandemic, in January 2020, just 2.5% of job postings were for remote work; that share increased to roughly 7% by the end of 2021 (Judes, Adrjan, and Sinclair 2021; Kolko 2021). However, the ability to work remotely is associated with workers who have higher educational attainment and higher earnings (Desilver 2020; Dingel and Neiman 2020; Kolko 2021). While some workers may see fewer constraints from local labor markets as a result of working remotely, that will not be the case for all or even most workers.

B. Ability to search for jobs

Aside from the availability of jobs, a worker must dedicate time to search for a new position. This is not a trivial undertaking, nor is job searching equally accessible or successful for all workers. We can see this immediately if we think of three contexts for job searching: searching on-the-job during work hours, searching on-the-job after work hours, and searching while unemployed. Each context supports a different level of effort and has a different cost for the searching worker.

Workers who want to change jobs without having a gap in income maintain their current employment while searching. Practically speaking, some workers, particularly those with access to the internet and with low levels of active supervision or monitoring, will be able to search during the workday. A paralegal at a law firm, for example, uses a computer all day, and, though he or she may need to be available for calls or emails, it is unlikely a lawyer is constantly or even frequently examining the paralegal’s screen. But workers without computers or phones on hand who are actively engaged in a task would find search during the workday difficult to impossible. A child care worker, for example, has little access to a computer or internet at work, and is occupied with a physical task of minding children.

Employed workers who cannot sufficiently search on the job must search outside of work hours, but they face logistical challenges. Most job postings and applications are completed over the internet, yet according to the American Community Survey 15% of individuals do not have an internet subscription (Martin 2018). But the limiting factor for most workers is time. People with jobs work on average 8.1 hours a day, and people who are job searching search on average 2.8 hours a day (Bureau of Labor Statistics 2019). It would not be appropriate to add those estimates together, as they vary too much across individuals and work statuses or situations, but they illustrate the considerable time commitment job search requires and how difficult that can be to accommodate when one has a full-time job.

Securing an interview request takes considerable effort. According to a 2018 survey, nearly 40% of seekers who submitted 1–10 applications failed to receive any interview requests, compared with 26% who submitted 11–20 applications and 19% who submitted 21–80 (Dalton and Groen 2020). A separate survey of employed workers found that, of workers who received an offer over the prior four months (whether they were searching or not), the majority received only one (Federal Reserve Bank of New York 2022).

Compared with employed workers, unemployed workers have fewer time commitments that might interfere with their job search, but also no earned income to support themselves until a job is found. Workers can become unemployed through layoffs or firm closings or by newly entering or reentering the labor market. Unemployment in a discussion of at-will work arrangements, however, refers to those who voluntarily quit in order to find a new job. The only workers who can exercise this option, then, are those who can afford a period without earned income. 

Each of these contexts of job search imply significant disparities in income. Occupations that support search during the workday are higher paid, on average, than those that do not. A reasonable approximation for the share of occupations that can be performed primarily on a computer with access to the internet is the share of occupations that can be performed remotely, and, according to Dingel and Neiman (2020), discussed above, that share is 37%—and those jobs pay more than jobs that can’t be performed remotely. A separate study estimated that a higher share, around 44% of jobs, could be performed remotely, yet the share was much higher (67–70%) among those with a bachelor’s degree than among those without a high school diploma (11–17%). It is safe to assume, then, that search during the workday is a luxury and more likely an option for the highest earning and the highest educated.

Job searching outside of work hours is more time consuming, a fact that disadvantages mothers who already work a “second shift” of child care and housekeeping (Hochschild and Machung 2012). Even in 2019, full-time employed married mothers spent an average of 75 minutes more per day on household activities, purchasing goods and services, and caring for children than full-time employed married fathers (Bureau of Labor Statistics 2019). In addition, job searching requires a computer and an internet connection, and the 15% of households without broadband skews with income: only 3% of households with annual incomes of more than $150,000 lacked broadband, compared with 25% with income less than $25,000. And taking time off for a job search requires significant savings, which we discuss in the next section.

Further, a significant share of job matches comes through an individual’s network (Granovetter 1973). Researchers have investigated many aspects of networks and how they relate to the probability of being hired, wages, and tenure in the job, and whether the importance of networks is found across the wage distribution (Simon and Warner 1992; Ioannides and Datcher Loury 2004; McDonald 2015; Schmutte 2015, 2016; Brown, Setren, and Topa 2016). A key finding relevant to the investigation of worker mobility is that networks are often a channel of labor market inequality. The more white men there are in someone’s network, for example, the more job leads an individual can be expected to have (McDonald 2011). Jobs found through friends tend to pay more, but the premium is higher if the friend is white rather than Black (Tenev 2020). In terms of worker mobility, the interpretation of these findings is that networks aid in mobility, but the extent of the assistance varies along key dimensions of inequality. It is not clear whether new forms of networking via social platforms such as LinkedIn or Indeed augment or diffuse the inequality aspects of network connections or create new ones. LinkedIn’s own research claims that 85% of jobs are filled via networking (Adler 2016), but the use of social networking sites has various implications for success, ranging from whether the person is an extrovert (Davis et al. 2020) to how old they are and how they look in their profile photo (Krings et al. 2021).

Applications, moreover, are just the beginning. The vast majority of job offers come only after an interview (Dalton and Groen 2020). Again, employed workers are not equally able to accommodate a job interview, given that most interviews occur during the workday and that not all workers have paid time off. In 2021, 23% of private-sector workers did not have paid sick leave, 21% did not have paid vacations, and 54% did not have paid personal leave (Bureau of Labor Statistics 2021). Workers without paid leave would have to take a pay cut if the interview fell during the workday. Yet even workers with paid leave may need permission to miss work. Interviews can also be difficult to schedule if workers do not have control over their shifts. Only about 45% of workers know their work schedule less than a month in advance, and about 20% know their work schedule less than one week in advance (Reeves 2020). Hourly retail and food service workers face even more unpredictability: one-third had less than one week’s notice of their schedules, and their schedules varied in total hours week-to-week (Schneider and Harknett 2019). Even a single interview can be difficult to arrange, but many jobs require multiple interviews.

In conclusion, the tasks, pay, scheduling, and benefits of certain jobs make it difficult to find another one. The exigencies of job search are not equally accommodated by every job, and thus worker mobility is highly varied based on the current job and whether it affords search, how the work schedule is set, and whether it offers time off.

One could argue that as long as the worker has some ability to move to another position, potentially with features that enable easier mobility, limits to worker mobility are short term or job specific. However, workers of color are discriminated against in the hiring process, a barrier that is neither short term nor job specific.

Researchers test for labor market discrimination through audit studies, in which resumes are generated and sent in response to job postings and callback rates are measured. Critically, the resumes feature some intentional but specific variation, and the callback rate for interviews is a test of that feature’s labor market penalty. One prominent audit study, for example, changed the names on resumes to Black-sounding names (such as Lakisha and Jamal) or white-sounding names (such as Emily and Greg); white names received 50% more callbacks (Bertrand and Mullainathan 2004). Audit studies have been tested in numerous settings and iterations, but the findings are remarkably consistent and have not changed in the past 30 years: Black and Hispanic workers are called back less for job interviews than white workers (Quillian et al. 2017). Discrimination applies to both the high-wage and low-wage segments of the labor market. Audit studies have found that Black Harvard graduates have callback rates on par with white public university graduates (Gaddis 2015), and Black and Hispanic workers with clean records have similar callback rates as white workers with felony records (Pager, Western, and Bonikowski 2009). A recent study, which sent out 83,000 applications to the largest 108 employers in the U.S., found that discrimination against Black applicants was present and driven by highly discriminatory firms representing about a quarter of the firm sample (Kline, Rose, and Walters 2021).

The differential difficulty in job finding contributes to the persistent differences in Black and white economic outcomes. The differences in hiring rates can explain half of the unemployment rate differences between Black and white workers (Forsythe and Wu 2021), and these differences, along with wealth disparities and others, push down wages (Stelzner and Bahn 2021).

Figure B uses a stylized, hypothetical model of job search results to illustrate how prolonged a search can be. We assume two success rates—17.5% and 35%—of each application securing an interview and securing an offer, and map out how many applications are necessary to receive five job offers. The climbing lines, scaled on the left axis, show the number of interviews per application. The bars, scaled on the right axis, show the number of job offers per application. For example, at 25 applications at a 35% success rate, the applicant has garnered nine interviews and three offers, compared with just four interviews and one offer at a 17.5% success rate. The relative steepness of the lines shows how quickly applications are converted to interviews, and the differential height of the bars shows how many offers are associated with application rates. At a 35% success rate, an individual would need to apply for 40 jobs in order to get 14 interviews that yielded five job offers. To get those five offers at a 17.5% success rate, an individual would need to apply for 150 jobs in order to get 26 interviews. These differences are not trivial—being half as likely to be called back results in a job search process three times longer. Assuming two hours per application and four hours per interview, this is the difference between 136 hours and 404.

Figure B

Job search is not equally or easily accomplished for workers in different occupations, but discrimination in the labor market adds a pernicious and permanent disparity in the success of that search.

C. Match of job and worker

The final step and the accompanying assumption in worker mobility is that the worker find and can find a job of sufficient quality. That is, conditional on jobs being available and conditional on workers being able to search for those available jobs, the offered job must match the worker.

Match is multidimensional. It encompasses skill, pay, benefits, location, schedule, working conditions, on-the-job growth, and earnings potential, among other factors. It is not the case that there is a consistent ranking across workers of which attribute is most important. Working mothers who manage child care responsibilities, for example, have been found to be very sensitive to commute times in assessing employment opportunities, even turning down higher-paying offers that might disrupt the commute, and by extension child pickup and drop-off (Manning and Petrongolo 2017). The nonwage attributes of a job, whether they are the working conditions or other amenities, are key determinants of job preferences (Maestas et al. 2018) and job satisfaction (Sockin 2021).

Hence, the quality of the match is determined by the worker, workers are not uniform in what they value in a match and how they assess a match, and, as a result, it is difficult for a researcher to determine the quality of matches in assessing barriers to worker mobility. How large an assumption it is that a worker can find a suitable job and, similarly, how limiting match quality is in worker mobility are difficult to measure. We could, for example, look to the rejection rate of job offers to get a sense of match quality and mobility. Glassdoor, a job search website, finds that the rejection rate of job offers varies with broader macroeconomic conditions, but generally is in the range of 15–19% (Sockin and Zhao 2020). This rejection rate would suggest that match quality is an issue but not a large one—the majority of job offers are accepted. However, individuals presumably limit their search to jobs that they would want, so the acceptance rate is conditional on the application rate, and we do not observe both. The details of the job, and the wage and nonwage features, may be withheld until an offer is made. As noted previously, the majority of workers receive only one job offer, making rejection less likely, and individuals who feel that there are few suitable jobs available may be discouraged from searching at all.

Observing workers who move from one job to another—job switchers—does not necessarily yield information on job match either, because workers with limited mobility may move to a worse, low-quality job. Earnings are a fundamental indicator of job quality and likely an important aspect of job match. Models of lifetime earnings have found that switching to higher-paying jobs is a key contributor to total earnings growth over a worker’s career (Topel and Ward 1992). But in fact, only about half of job switchers between 1996 and 2010 went into a higher-paying job (Wiczer 2016). The average wage increase between jobs was 2.6%, and a fair share of switchers lost half of their earnings from the prior job. Part of the reason is that a voluntary job switch is not free of negative aspects. Workers could switch because they were warned that they would be fired, they could have had a negative experience with a colleague, or they could have experienced sexual harassment (Nyström and Zhetibaeva Elvung 2015; Willness, Steel, and Lee 2007).

Indeed, interpreting switching to lower-quality jobs (at least measured by earnings) introduces the question of whether quality is a limiting condition to mobility at all. That is, does switching into a lower-paying job support the conclusion that pay (or other measures of quality) does not always inform a worker’s assessment of a match or that pay (or other measures of quality) does not limit mobility? Workers either do not care about quality or they do care about quality, but do not let it function as a limit to mobility. Both could be, and likely are, true to an extent. A key aspect of this question, though, is how well a searching worker would be able to assess quality before starting the job.

As a theoretical aside, the requirements of match in job search could be interpreted as a search friction and evidence of dynamic monopsony (as we present it here) or, as an alternative, could be interpreted as product differentiation, a separate channel for monopsony. As Naidu, Posner, and Weyl (2018) explain, employers have preferences over workers regarding such factors as their skills, education, and experience but also their personality and work style. At the same time, workers have preferences over jobs, as noted above. The labor market’s two-sided preferences mean that it is “doubly differentiated” and therefore naturally thinner than a single-differentiated market would be. For example, a person shopping for clothing, like a pair of jeans, can have many different preferences: cut, fit, style, material, and even sourcing and sustainability. The store selling the jeans, on the other hand, only cares that any potential buyer pay the posted price, and it does not discern between buyers based on their age, height, hair color, commitment to fashion, or even their intention to actually wear the clothes. If stores had to meet those preferences before selling, it would be much harder to buy clothing. Two-sided preferences make matching more difficult.

Critically, in the labor market’s two-sided preference match, workers know less than employers and thus there is information asymmetry in job search, a factor that tilts labor markets toward monopsony. Learning about a job’s features, especially the culture of the workplace, during the search process can be difficult. Research has found that an employer’s reputation is important to worker’s preferences (Benson, Sojourner, and Umyarov 2020), yet employers actively try to discourage transparency about working conditions via nondisclosure agreements (Sockin, Sojourner, and Starr 2022), and fear of retaliation can be a deterrent to workers commenting at sites like Glassdoor about former employers (Sockin and Sojourner 2020). Even while employed, workers often have little information about their wages relative to those of their coworkers, and learning that they are paid less decreases job satisfaction (Card et al. 2012). Observed job switches into lower-quality jobs, therefore, cannot be assumed to be a worker’s specific intention: the worker may not know that the job is of lower quality.

Some would argue that match is not a barrier to leaving jobs. In theory, so long as any job is available, worker mobility is not limited. The first half of that statement is certainly true: a worker may not prefer to move from being a teacher to, say, a cashier, but the worker can move. The latter half is problematic: moves have consequences, for both future mobility and trajectories, as well as for the power of monopsony. These consequences are illustrated by the low-wage labor market, where there is often an available job.

Analysis of the employment and earnings trajectories of workers in the low-wage labor market suggests that low-wage employment can be a trap. For one thing, low-wage jobs have a much higher likelihood of leading to unemployment than do high-wage jobs, creating a “low-pay/no-pay” cycle of employment (Fok, Scutella, and Wilkins 2015; Mosthaf, Schank, and Schnabel 2014). Second, low-wage jobs don’t pay sufficient wages to support savings, so if workers becomes unemployed they can’t afford a nonworking break and must take the next available job, which is also low wage. Third, many low-wage jobs offer little on-the-job training or investment, so that unemployed low-wage workers are less qualified for different, better jobs (Cuesta and Salverda 2009). Finally, and particularly important to downward mobility, low-wage employment has a scarring effect, similar to unemployment or resume gaps, that reduces future employment prospects (Stewart 2007).

Taken together, low-wage employment greatly increases the chance of future unemployment and decreases the chance of future high-wage employment (Mosthaf, Schank, and Schnabel 2014). Critically, mobility out of low-wage employment into high-wage employment has declined over the last 30 years (Schultz 2019), and research consistently finds that the trap is particularly deep for women (Mosthaf, Schnabel, and Stephani 2011). The ability to move down the job ladder is not proof of worker mobility unless there is demonstrated ability to move back up.

Further, having access to any available job versus a suitable job can augment employer power and decrease wages. Being in an occupation in a labor market with more outside options—defined as the full set of jobs that workers in that occupation move from or to—is associated with higher wages than being in the same occupation in a labor market with fewer outside options (Schubert, Stansbury, and Taska 2021). This finding helps delineate between movement and mobility; even monopsonistic labor markets will feature movement across jobs, but that movement is not the same as competitive mobility. Hence, match quality may limit some movement and will limit mobility.

D. Immobility and labor policy

Our approach so far in demonstrating that worker mobility is limited in practice has been to examine how difficult it is to leave a current situation. A complementary approach to interrogating the feasibility of leaving an employment situation is to assess if workers do not leave when there is reason they should.

Consider a perfectly competitive labor market. Not only are workers competing for jobs, but employers are competing for workers. If a job has features that workers do not like, employers must either raise wages to compensate for that feature or eliminate it. If they do not, workers—perfectly mobile in this perfectly competitive market—will leave. The ability to leave any employment situation freely would thus force employers to improve working conditions and wages or lose their workers. It also follows that in a perfectly competitive labor market there is no need for labor regulation; competition is the regulation that marries working conditions to workers’ standards.

The introduction of labor regulations and their effect is one kind of indication of worker mobility: the greater the lack of mobility, the less power workers have to improve working conditions and the more regulatory intervention is necessary (Stelzner and Paul 2020). The U.S. labor regulatory regime spans multiple pieces of legislation, including the Fair Labor Standards Act, the Occupational Safety and Health Act, the Civil Rights Act, and the Equal Pay Act. In a sense, the existence of these laws is an explicit acknowledgment of lack of worker power in improving working conditions and an implicit acknowledgment of lack of worker mobility.

Not everyone agrees with the need for or use of regulation in labor or other markets, and not everyone would interpret its existence and application as proof of shortcomings in the market. However, the continued use of regulation and enforcement is fresh evidence that leaving for a similar or better job is not always a feasible option. For example, since the passage of the Occupational Safety and Health Act (OSHA) of 1970, the rate of workplace injury has decreased fivefold (Brown 2020). Some of this decrease is attributed to the changing industrial composition of the workforce in the past five decades, but research into the effect of OSHA inspections on safety finds that they reduce the incidence of on-the-job injuries and days away from work (Levine, Toffel, and Johnson 2012; Li and Singleton 2019).

Further, employer power is a key moderator of regulatory effectiveness. Researchers have found that where employer power is stronger, such as in workplaces with high numbers of Hispanic workers in areas where immigration enforcement has increased (Grittner and Johnson 2021), regulation is less effective. Conversely, where employer power is weaker, such as in union-organized workplaces (Sojourner and Yang 2022), regulation is more effective. Worker mobility drives the effectiveness of labor regulations as well. A study of sexual harassment reporting found that reports decreased in number, but increased in severity, based on the unemployment rate at the time and the value of unemployment insurance benefits. The authors concluded that workers with fewer outside options are less likely to report sexual harassment (Dahl and Knepper 2021). Thus, the need for regulations and the struggle to enforce them are indicators of limits to worker mobility.

IV. Financial constraints

The corollary to the labor market constraints that limit worker mobility are the financial constraints. The former concerns the availability of alternative jobs and the difficulty in finding them, the latter concerns the cost of finding and taking them.

Job mobility for established workers (excluding here job search and finding by new entrants or re-entrants to the labor market who have not worked for a considerable period and those who were laid off) occurs under two circumstances. Either they move job-to-job (JTJ) or they quit, have a period of unemployment dedicated to search, and find a new job, i.e., move job-to-unemployment-to-job (JUJ). A worker can take time off between positions in a JTJ move, but that time off is not dedicated to, or motivated by, the search for work because the worker already has a position. The period of time off in a JUJ move is to support job search. We have discussed previously the practical needs of both JTJ and JUJ in terms of job search. Here we discuss the financial costs of job transitions.

A. Job-to-job transitions

Financially the best-case scenario for worker mobility is to transition between jobs without needing to finance time off between paid employment. However, even this type of transition can entail costs for a worker, aside from a potentially negative difference in wages.

As long as a firm is not violating federal nondiscrimination standards, nonwage compensation like retirement contributions, health insurance, and transportation benefits are not required to be offered uniformly to all employees. And a new employee does not have to be offered the same benefits as more tenured employees; for example, a worker might not be able to enroll in the firm’s retirement plan until after 12 months of tenure, or a worker might have to restart accrual of paid time off for sick days or vacation days. So workers can experience gaps in coverage or compensation, and this transition cost is greater if the moving worker also lost benefits from leaving, left during an early step in the vesting schedule, or was not paid unused time off.

Nonwage benefits, or losses associated with transitioning from one job to another, might seem marginal, especially relative to potential wage gains. However, benefits constrain worker mobility. Job lock refers to workers who do not leave their firms because doing so would disrupt or cause a loss in their health insurance coverage. Early estimates of job lock suggested that it reduced voluntary departures by 25% (Madrian 1994) and affected workers who had, or whose spouses or children had, preexisting conditions (Gruber and Madrian 1994; Madrian 1994). The inverse is also true; easier access to health insurance is associated with moves to higher-paying jobs (Farooq and Kugler 2022). Of course, the health insurance landscape has evolved since job lock was first estimated, but even after the preexisting coverage mandate of the Affordable Care Act of 2010, job lock continues to affect certain workers who are concerned with switching out a doctor’s network with a new insurer (Kent et al. 2020).

Finally, even noncompensation aspects of a new job can carry financial costs, such as commuting. Three-fourths of American workers drive alone to work, and the daily cost of a driving commute is $8-13 (U.S. Census Bureau 2018, 2020). Over a full-time work year, that equates to $2,080-3,380; moving could increase that cost. Similarly, longer shifts or longer commutes could incur a need for additional child care. If work extends beyond the hours a licensed home care provider or center-based care is open, a worker might have to find different care, pay for extended care, or hire a part-time caregiver like a nanny or babysitter to fill in the hours before the parent’s work ends. In addition, a change in job location could require a new child care situation if the worker previously had an arrangement at or near the former employer; note that 7% of employers offer onsite child care (Matos, Galinsky, and Bond 2017). Disruption of care is stressful. Surveys of parents have long established that finding affordable child care for their preschool children is difficult, the options are limited, and, even when found, the cost is burdensome (Care.com 2021; Harvard T.H. Chan School of Public Health 2016; Pew Research Center 2015). Up to half of all spending on children goes toward child care (Hubener, Rojas, and Tseng 2018); affording child care is a key stressor among parents of young children and reduces maternal mental health (Lyons-Ruth et al. 2002; Mistry et al. 2007). Hence, any job change that requires a corresponding change in care can be financially costly and mentally difficult.

For high-income workers, job-to-job costs such as health insurance, transportation, or child care could be larger in number but smaller in relative financial burden, greatly diminishing the inhibitions on their mobility in the labor market. Yet, for certain workers—low-income workers, those with preexisting conditions, mothers—those costs can be prohibitive. Health insurance job lock is not only an example of this type of constraint to worker mobility but also an excellent way to conceptualize it: there is a financial component of a job that is not guaranteed or maintained with a new employer, and some components are valued by the worker above their strict monetary value.

B. Job-to-unemployment-to-job transitions

Workers can quit their jobs and search while unemployed, but it’s a risky move without earnings and no definitive timeframe. Studies of unemployed workers find that the more financial cushion one has in unemployment, the better the job upon reemployment (Farooq, Kugler, and Muratori 2020; McCall and Chi 2008).

Though workers who voluntarily quit their jobs are generally not eligible for unemployment insurance benefits, some states make exceptions for quits for medical reasons that are related to the job; quits related to domestic violence; quits to care for an ill family member; quits for a job that did not pan out (because of, say, a rescinded offer); quits by a worker who had to quit to relocate for a spouse’s job; and quits made under “constructive discharge,” i.e., a situation in which a worker enduring harassment or unsafe working conditions cannot continue in the position. Applying for unemployment for any of these reasons requires appealing an initially denied claim, which can be a difficult-to-prove and protracted process. In 2021, an average of 26.4% of claimant appeals were awarded by lower authorities, ranging from 10.7% in Texas to 58.3% in California; at the end of that year, 75% of all pending appeals cases were 41–360 days old, and 9% were more than 360 days old (U.S. Department of Labor 2021).

Without help from unemployment insurance, workers must finance the spell on their own (though it is important to note that unemployment insurance benefits are not generous, and beneficiaries would not likely be able to cover all expenses without additional financial resources). Most workers would likely seek help from their families (Edwards 2020; Edwards and Wenger 2019), and some might move in with a family member during the process (Wiemers 2014). But not every worker has a family relationship that includes financial support, and not every spouse earns enough to support two workers. Family support is a privilege, not a given.

The median duration of unemployment in the U.S. ranges from nine to 20 weeks, depending on the overall unemployment rate (Bureau of Labor Statistics 2021). Even assuming the shortest duration of nine weeks, a worker must cover two months of living expenses: rent or a mortgage, utilities, food, health insurance, and perhaps child care. In 2019, the median family had $5,300 across savings accounts, checking accounts, prepaid cards, and money market funds. However, the median family income was also $58,000, and those with lower incomes would have much less or even nothing saved (Bhutta, Bricker, et al. 2020). Also in 2019, 37% of families did not have enough cash on hand to cover a $400 emergency expense, a proximate indication of having less than $400 in savings (Federal Reserve 2021).

Having two months of living expenses saved can be difficult for cash-strapped households, even with assistance. The JP Morgan Chase Institute analyzed the balances of its account holders in the months after the U.S. government, in response to the COVID-19 pandemic, sent economic impact payments to American households in April 2020 and January 2021. Account balances grew after the payments, with the largest increases occurring in the accounts of the lowest-income households. But within six months, balances were steadily declining, with larger increases and steeper falls for the lowest income (Farrell et al. 2020; Greig, Deadman, and Noel 2021). The advanced child tax credit payments in place from July to December 2021 further buoyed account balances, particularly for low-income households (Greig, Deadman, and Sonthalia 2022), but throughout this period balances for the lowest-income households (with incomes from $12,000 to $26,000) ranged from $1,000 to $2,000.

Of course, the expenses arising from deciding to become unemployed are not unexpected. A worker could save for that goal, but how reasonable is it to assume that a worker is able to do so? Lower-income families are less likely to save and, if saving, save at lower rates, due not to a preference against saving but to practical limitations in doing so (Dynan, Skinner, and Zeldes 2004). For example, rent often rises much faster than incomes, presenting an acute challenge to accruing savings. The price for average rents in U.S. cities increased 73% between 2000 and 2019, while median income rose by just 10%. Just before the pandemic, 38% of renting households were rent burdened, meaning that they spent more than 30% of their income on rent, and 17% were severely rent burdened, spending more than 50%. Of these rent-burdened households, two-thirds did not have $400 in savings and half had functionally nothing in savings (Pew Research Center 2018). Rent is just one part of the household budget, but illustrative of the overall challenges in maintaining savings under the financial pressure of expenses and the often unpredictability of income (Morduch and Schneider 2019).

C. Moving

Many of the labor market constraints a worker faces could be lifted or lessened if the worker were willing to move.

Moving is not uncommon. Between 2015 and 2019, roughly 13% of Americans moved residence each year, though the vast majority of those moves (65%) were within the same county. For our purposes, that means they were likely (though not necessarily, as noted in the example of Harlem labor markets) within the same labor market. Within-state cross-county moves accounted for an additional 17% of moves, and cross-state moves accounted for 14% (Frost 2020). Together, that means only 2% of Americans move out of their county (in-state) and an additional 2% move out of their state in a year. Across all moves, just one in five are job related, and moving within the U.S. for job-related reasons has been declining (Jia et al. 2022; Molloy, Smith, and Wozniak 2011, 2017).

Moving is not costless, and it may not be feasible even if a higher-paying job awaits. Renters must often pay two months of rent upfront—one for the first month’s rent and one for a security deposit—and although a security-deposit refund may be due from their prior residence, states vary in how quickly a landlord is required to pay over the funds (some states allow as many as 60 days). For the 65% of Americans in households who own their primary home, the timing frictions are greater, since they often must sell their current home (which could incur risks, as it is often the largest component of total wealth (Bhutta, Bricker, et al. 2020; Bricker et al. 2020). Homeowners could be paying for housing in both locations for a long period.

Moving also entails material and time costs. According to recent data the cost to hire movers for an in-town move of a studio apartment can be $400 (Perry and Allen 2021), and though renting a truck could be as low as $20 plus mileage (U-Haul International n.d.), not all workers have the physical capability to move themselves. About 20.1 million Americans age 18–64 have disabilities, half of which are ambulatory disabilities (U.S. Census Bureau 2021). Moreover, moving may entail the loss of social or family networks.

And much like in the labor market, there is persistent and well-documented discrimination in the housing market. The Department of Housing and Urban Development, using paired testing in which two prospective renters or buyers are sent to find housing in a locality, has found that Black, Hispanic, and Asian renters are told about and shown fewer units than white renters, and Black and Asian home buyers are shown fewer homes (U.S. Department of Housing and Urban Development 2014; Urban Institute 2015). Discrimination also occurs in the home appraisal process, reducing the equity individuals receive from their homes (Gunderson 2021; Kamin 2021; McMullen 2021; Rothstein 2017). While this loss may not prevent a worker from moving to take a job, it can increase the opportunity cost of selling.

D. Borrowing and wealth

One strategy for dealing with financial constraints of job switching is borrowing. With access to credit, workers should be able to smooth out job transition costs.

Access to credit, however, is not universal: 5% of Americans are “unbanked,” meaning they do not have access to a bank account and must rely on alternative financial services, and 17% do not have access to a credit card. For individuals in households with less than $25,000 in income, 16% are unbanked and 44% do not have a credit card (Federal Reserve 2021). For reference, $25,000 of annual income for a full-time, full-year worker translates to an hourly wage of $12.02, or $4.77 higher than the federal minimum wage of $7.25 and higher than the minimum wage in 37 states as of January 1, 2022 (U.S. Department of Labor 2022).

Access to credit also reflects clear racial differences. Blacks and Hispanics are denied credit at roughly twice the rate of whites, even within the same income groups. As shown in Table 1, which summarizes findings from the Federal Reserve’s annual report on the economic well-being of U.S. households, the share of Black and Hispanic adults denied credit is nearly twice and in some cases three times as high as the denial rate for white adults, regardless of income. Among families with $100,000 or more in income, only 7% of white adults were denied credit compared with nearly one in four—23%—of Black adults.

Table 1

Black and Hispanic Americans also have considerably less wealth than white Americans. In 2019, the median wealth of white families was $188,200, about eight times higher than the median wealth of Black families, at $24,100, and five times higher than Hispanic wealth, at $36,100. Differences in average wealth are wide as well: $983,400 for white families, $142,500 for Black families, and $165,500 for Hispanic families (Bhutta, Chang, et al. 2020). Research has found that being Black is a stronger predictor of wealth than the type of job an individual has (Addo and Darity 2021).

Disparities in banking, credit, and wealth are directly related to a workers’ ability to smooth out financial constraints. For example, in an emergency, families can draw on their cash and their near-liquid equity assets such as stocks, mutual funds, and retirement accounts Among white families, virtually all (98.8%) hold cash, averaging $8,100, and 60.8% have equity holdings, averaging $50,600. Extremely high shares of Black and Hispanic families have cash (96.8% and 95.5%, respectively), but their average balances are much lower, at $1,500 and $2,000. Only 33.5% of Black families have equity holdings (average value $14,400), as do just 24.2% of Hispanic families (average value $14,900) (Bhutta, Chang, et al. 2020).

Access to credit, liquid assets, and equity is a key component of preventing financial hurdles from becoming financial constraints. Not all Americans have credit, cash, or equity, and those disparities are present by income and race. Translated to the financial hurdles of worker mobility, lower-wage earners and workers of color are less mobile that their higher-income or white peers.

V. Conclusion

The evidence presented here shows how worker mobility is limited in the U.S. Searching for and securing another job requires time and endurance, and switching to another job can be costly. The theoretical context of these findings is dynamic monopsony: the harder it is for a worker to leave, the more power an employer has over that worker’s wages.

About the author

Kathryn Anne Edwards (Kathryne@rand.org) is an economist at the RAND Corporation and a professor at the Pardee RAND Graduate School. Her research focuses on the intersection of labor markets and public policy, including unemployment and unemployment insurance, women’s labor supply after children, the challenge facing women in retirement, poverty alleviation, and Social Security.

References

Addo, Fenaba R., and William A. Darity. 2021. “Disparate Recoveries: Wealth, Race, and the Working Class after the Great Recession.” 695 Annals of the American Academy of Political and Social Science 173. doi: 10.1177/00027162211028822.

Adler, Lou. 2016. “New Survey Reveals 85% of All Jobs Are Filled via Networking.” LinkedIn. https://www.linkedin.com/pulse/new-survey-reveals-85-all-jobs-filled-via-networking-lou-adler/.

Ashenfelter, Orley C., Henry S. Farber, and Michael R. Ransom. 2010. “Modern Models of Monopsony in Labor Markets: A Brief Survey.” IZA Discussion Paper No. 4915. doi: 10.2139/ssrn.1599013.

Azar, José, Ioana Marinescu, Marshall Steinbaum, and Bledi Taska. 2020. “Concentration in U.S. Labor Markets: Evidence from Online Vacancy Data.” 66 Labour Economics 101886. doi: 10.1016/j.labeco.2020.101886.

Benmelech, Efraim, Nittai K. Bergman, and Hyunseob Kim. 2020. “Strong Employers and Weak Employees: How Does Employer Concentration Affect Wages?” Journal of Human Resources, December 14. doi: 10.3368/jhr.monopsony.0119-10007R1.

Benson, Alan, Aaron Sojourner, and Akhmed Umyarov. 2020. “Can Reputation Discipline the Gig Economy? Experimental Evidence from an Online Labor Market.” 66 Management Science 1802. doi: 10.1287/mnsc.2019.3303.

Bertrand, Marianne, and Sendhil Mullainathan. 2004. “Are Emily and Greg More Employable than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination.” 99 American Economic Review 991. doi: 10.1257/0002828042002561.

Bhutta, Neil, Jesse Bricker, Andrew Change, Lisa Dettling, Serena Goodman, Joanne Hsu, Kevin Moore, Sarah Reber, Alice Henriques Volz, and Richard Windle. 2020. “Changes in U.S. Family Finances from 2016 to 2019: Evidence from the Survey of Consumer Finances.” 106 Federal Reserve Bulletin 1.

Bhutta, Neil, Andrew C. Chang, Lisa J. Dettling, and Joanne Hsu. 2020. “Disparities in Wealth by Race and Ethnicity in the 2019 Survey of Consumer Finances.” FEDS Notes, September 28.

Bricker, Jesse, Sarena Goodman, Kevin B. Moore, Alice Henriques Volz, and Dalton Ruh. 2020. “Wealth and Income Concentration in the SCF: 1989–2019.” FEDS Notes, September 28.

Brown, Jeff. 2020. “Nearly 50 Years of Occupational Safety and Health Data.” 9 Beyond the Numbers, n.p.

Brown, Meta, Elizabeth Setren, and Giorgio Topa. 2016. “Do Informal Referrals Lead to Better Matches? Evidence from a Firm’s Employee Referral System.” 34 Journal of Labor Economics 161. doi: 10.1086/682338.

Burd, Charlynn, Michael Burrows, and Brian McKenzie. 2021. Travel Time to Work in the United States: 2019. ACS-47. U.S. Census Bureau.

Bureau of Labor Statistics. 2019a. “Table A-1. Time Spent in Detailed Primary Activities and Percent of the Civilian Population Engaging in Each Activity.” https://www.bls.gov/tus/a1-2019.pdf.

Bureau of Labor Statistics. 2019b. “Table A-6A. Time Spent in Primary Activities and the Percent of Married Mothers and Fathers Who Did the Activities on an Average Day by Employment Status, Average for the Combined Years 2015-19.” https://www.bls.gov/tus/tables/a6-1519.htm.

Bureau of Labor Statistics. 2020. “Table 1. Employed Persons Who Teleworked or Worked at Home for Pay at Any Time in the Last 4 Weeks Because of the Coronavirus Pandemic by Selected Characteristics.” https://www.bls.gov/cps/effects-of-the-coronavirus-covid-19-pandemic.htm.

Bureau of Labor Statistics. 2021a. “National Compensation Survey: Employee Benefits in the United States, March 2021.” https://www.bls.gov/ncs/ebs/benefits/2021/home.htm.

Bureau of Labor Statistics. 2021b. “Table A-12. Unemployed Persons by Duration of Unemployment.” https://www.bls.gov/webapps/legacy/cpsatab12.htm/Table-A-12-Unemployed-persons-by-duration-of-unemployment.

Card, David. 2022. “Who Set Your Wage?” Presidential Address to the Annual Meetings of the American Economic Association, January 8. https://www.aeaweb.org/articles?id=10.1257/aer.112.4.1075.

Card, David, Alexandre Mas, Enrico Moretti, and Emmanuel Saez. 2012. “Inequality at Work: The Effect of Peer Salaries on Job Satisfaction.” 102 American Economic Review 2981. doi: 10.1257/aer.102.6.2981.

Care.com. 2021. “This Is How Much Child Care Costs in 2021.” Care.Com Resources. https://www.care.com/c/how-much-does-child-care-cost/.

Cuesta, Maite Blázquez, and Wiemer Salverda. 2009. “Low-Wage Employment and the Role of Education and On-the-Job Training.” 23 Labour 5. doi: 10.1111/j.1467-9914.2008.00431.x.

Dahl, Gordon B., and Matthew M. Knepper. 2021. Why Is Workplace Sexual Harassment Underreported? The Value of Outside Options amid the Threat of Retaliation. Working Paper No. 29248. National Bureau of Economic Research. doi: 10.3386/w29248.

Dalton, Michael, and Jeffrey Groen. 2020. “How Do Jobseekers Search for Jobs? New Data on Applications, Interviews, and Job Offers.” 9 Beyond the Numbers, n.p.

Davis, Joanna, Hans-Georg Wolff, Monica L. Forret, and Sherry E. Sullivan. 2020. “Networking via LinkedIn: An Examination of Usage and Career Benefits.” 118 Journal of Vocational Behavior 1. doi: 10.1016/j.jvb.2020.103396.

Desilver, Drew. 2020. Before the Coronavirus, Telework Was an Optional Benefit, Mostly for the Affluent Few. Pew Research Center. https://www.pewresearch.org/fact-tank/2020/03/20/before-the-coronavirus-telework-was-an-optional-benefit-mostly-for-the-affluent-few/.

Dingel, Jonathan I., and Brent Neiman. 2020. How Many Jobs Can Be Done at Home? Working Paper No. 26948. National Bureau of Economic Research. doi: 10.3386/w26948.

Dube, Arindrajit, Alan Manning, and Suresh Naidu. 2018. Monopsony and Employer Mis-Optimization Explain Why Wages Bunch at Round Numbers. Working Paper No. 24991. National Bureau of Economic Research. doi: 10.3386/w24991.

Dynan, Karen E., Jonathan Skinner, and Steve Zeldes. 2004. “Do the Rich Save More?” 112 Journal of Political Economy 397.

Edwards, Kathryn Anne. 2020. “Who Helps the Unemployed? Workers’ Receipt of Public and Private Transfers.” 9 IZA Journal of Labor Economics 1. doi: 10.2478/izajole-2020-0004.

Edwards, Kathryn Anne, and Jeffrey B. Wenger. 2019. “Parents with an Unemployed Adult Child: Consumption, Income, and Savings Effects.” 8 IZA Journal of Labor Economics 1. doi: 10.2478/izajole-2019-0001.

Enrico, Moretti. 2011. “Local Labor Markets.” In Handbook of Labor Economics, Vol. 4, edited by D. Card and O. Ashenfelter, 1237. Elsevier.

Farooq, Ammar, and Adriana Kugler. 2022. “Impacts of Public Health Insurance on Occupational Upgrading.” 75 ILR Review 225. doi: 10.1177/0019793920928066.

Farooq, Ammar, Adriana D. Kugler, and Umberto Muratori. 2020. Do Unemployment Insurance Benefits Improve Match Quality? Evidence from Recent U.S. Recessions. Working Paper No. 27574. National Bureau of Economic Research. doi: 10.3386/w27574.

Farrell, Diana, Erica Deadman, Fiona Greig, and Pascal Noel. 2020. Household Cash Balances during COVID-19: A Distributional Perspective. JP Morgan Chase Institute.

Federal Reserve. 2021. Report on the Economic Well-Being of U.S. Households in 2020. https://www.federalreserve.gov/publications/2021-economic-well-being-of-us-households-in-2020-overall-financial-well-being-in-2020.htm.

Federal Reserve Bank of New York. 2022. Survey of Consumer Expectations. https://www.newyorkfed.org/microeconomics/sce/labor#/.

Fok, Yin King, Rosanna Scutella, and Roger Wilkins. 2015. “The Low-Pay No-Pay Cycle: Are There Systematic Differences across Demographic Groups?” 77 Oxford Bulletin of Economics and Statistics 872. doi: 10.1111/obes.12084.

Forsythe, Eliza, and Jhih-Chian Wu. 2021. “Explaining Demographic Heterogeneity in Cyclical Unemployment.” 69 Labour Economics 1. doi: 10.1016/j.labeco.2020.101955.

Frost, Riordan. 2020. Who Is Moving and Why? Joint Center for Housing Studies.

Gaddis, S. Michael. 2015. “Discrimination in the Credential Society: An Audit Study of Race and College Selectivity in the Labor Market.” 93 Social Forces 1451. doi: 10.1093/sf/sou111.

Goldstein, Amy. 2018. Janesville. Simon and Schuster.

Granovetter, Mark S. 1973. “The Strength of Weak Ties.” 78 American Journal of Sociology 1360. doi: 10.1086/225469.

Greig, Fiona, Erica Deadman, and Pascal Noel. 2021. Family Cash Balances, Income, and Expenditures Trends through 2021. JP Morgan Chase Institute.

Greig, Fiona, Erica Deadman, and Tonya Sonthalia. 2022. Household Pulse: The State of Cash Balances at Year End. JP Morgan Chase Institute.

Grittner, Amanda, and Matthew S. Johnson. 2021. Deterring Worker Complaints Worsens Workplace Safety: Evidence from Immigration Enforcement. SSRN Scholarly Paper No. 3943441. Social Science Research Network. doi: 10.2139/ssrn.3943441.

Gruber, Jonathan, and Brigitte C. Madrian. 1994. “Health Insurance and Job Mobility: The Effects of Public Policy on Job-Lock.” 48 ILR Review 86. doi: 10.2307/2524628.

Gunderson, Erica. 2021. “Home Appraisal Discrimination Puts a Dent in Black Wealth.” WTTW News, October 16.

Guyot, Katherine, and Richard V. Reeves. 2020. “Unpredictable Work Hours and Volatile Incomes Are Long-Term Risks for American Workers.” Brookings. https://www.brookings.edu/blog/up-front/2020/08/18/unpredictable-work-hours-and-volatile-incomes-are-long-term-risks-for-american-workers/.

Harvard T.H. Chan School of Public Health. 2016. “Child Care and Health in America.” NPR/Robert Wood Johnson Foundation/Harvard School of Public Health. https://www.rwjf.org/en/library/research/2016/10/child-care-and-health-in-america.html.

Harvard T.H. Chan School of Public Health. 2021. “Household Experiences in America During the Delta Variant Outbreak.” NPR/Robert Wood Johnson Foundation/Harvard School of Public Health. https://www.rwjf.org/en/library/research/2016/10/child-care-and-health-in-america.html.

Hochschild, Arlie, and Anne Machung. 2012. The Second Shift: Working Families and the Revolution at Home. Penguin Random House.

Hubener, Evan, Arcenis Rojas, and Neil Tseng. 2018. “Tradeoffs in the Expenditure Patterns of Families with Children.” 7 Beyond the Numbers, n.p.

Ioannides, Yannis M., and Linda Datcher Loury. 2004. “Job Information Networks, Neighborhood Effects, and Inequality.” 42 Journal of Economic Literature 1056. doi: 10.1257/0022051043004595.

Jia, Ning, Raven Molloy, Christopher Smith, and Abigail Wozniak. 2022. “The Economics of Internal Migration: Advances and Policy Questions.” Finance and Economics Discussion Series (FEDS). https://www.aeaweb.org/articles?id=10.1257/jel.20211623&&from=f.

Judes, Alexandre, Pawel Adrjan, and Tara Sinclair. 2021. “Will Remote Work Persist after the Pandemic?” Indeed Hiring Lab. https://www.hiringlab.org/2021/12/16/will-remote-work-persist-after-the-pandemic/.

Kamin, Debra. 2021. “Black Homeowners Face Discrimination in Appraisals.” New York Times, August 25.

Kent, Erin E., Janet S. de Moor, Jingxuan Zhao, Donatus U. Ekwueme, Xuesong Han, and K. Robin Yabroff. 2020. “Staying at One’s Job to Maintain Employer-Based Health Insurance among Cancer Survivors and Their Spouses/Partners.” 6 JAMA Oncology 929. doi: 10.1001/jamaoncol.2020.0742.

Kline, Patrick, Evan K. Rose, and Christopher Walters. 2021. Systemic Discrimination among Large U.S. Employers. SSRN Scholarly Paper No. 3898669. Social Science Research Network. doi: 10.2139/ssrn.3898669.

Kolko, Jed. 2021. “Remote Job Postings Double During COVID-19.” Indeed Hiring Lab. https://www.hiringlab.org/2021/03/16/remote-job-postings-double/.

Krings, Franciska, Irina Gioaba, Michèle Kaufmann, Sabine Sczesny, and Leslie Zebrowitz. 2021. “Older and Younger Job Seekers’ Impression Management on LinkedIn.” 20 Journal of Personnel Psychology 61. doi: 10.1027/1866-5888/a000269.

Levine, David I., Michael W. Toffel, and Matthew S. Johnson. 2012. “Randomized Government Safety Inspections Reduce Worker Injuries with No Detectable Job Loss.” 336 Science 907. doi: 10.1126/science.1215191.

Li, Ling, and Perry Singleton. 2019. “The Effect of Workplace Inspections on Worker Safety.” 72 ILR Review 718. doi: 10.1177/0019793918801575.

Lyons-Ruth, Karlen, Rebecca Wolfe, Amy Lyubchik, and Ronald Steingard. 2002. “Depressive Symptoms in Parents of Children under Age 3: Sociodemographic Predictors, Current Correlates, and Associated Parenting Behaviors.” In Child Rearing in America: Challenges Facing Parents with Young Children, edited by Neal Halfon, Kathryn Taaffe McLearn, and Mark A. Schuster, 217. Cambridge University Press.

Madrian, Brigitte C. 1994. “Employment-Based Health Insurance and Job Mobility: Is There Evidence of Job-Lock?” 109 Quarterly Journal of Economics 27. doi: 10.2307/2118427.

Maestas, Nicole, Kathleen J. Mullen, David Powell, Till von Wachter, and Jeffrey B. Wenger. 2018. The Value of Working Conditions in the United States and Implications for the Structure of Wages. Working Paper No. 25204. National Bureau of Economic Research. doi: 10.3386/w25204.

Manning, Alan. 2003. Monopsony in Motion: Imperfect Competition in Labor Markets. Princeton University Press.

Manning, Alan. 2021. “Monopsony in Labor Markets: A Review.” 74 ILR Review 3. doi: 10.1177/0019793920922499.

Manning, Alan, and Barbara Petrongolo. 2017. “How Local Are Labor Markets? Evidence from a Spatial Job Search Model.” 107 American Economic Review 2877. doi: 10.1257/aer.20131026.

Martin, Michael. 2018. “Computer and Internet Use in the United States, 2018.” American Community Survey Reports. https://www.census.gov/content/dam/Census/library/publications/2021/acs/acs-49.pdf.

Matos, Kenneth, Ellen Galinsky, and James Bond. 2017. National Study of Employers. Families and Work Institute, Society for Human Resource Management.

McCall, Brian, and Wei Chi. 2008. “Unemployment Insurance, Unemployment Durations, and Re-Employment Wages.” 99 Economics Letters 115. doi: 10.1016/j.econlet.2007.06.006.

McDonald, Steve. 2011. “What’s in the ‘Old Boys’ Network? Accessing Social Capital in Gendered and Racialized Networks.” 33 Social Networks 317. doi: 10.1016/j.socnet.2011.10.002.

McDonald, Steve. 2015. “Network Effects across the Earnings Distribution: Payoffs to Visible and Invisible Job Finding Assistance.” 49 Social Science Research 299. doi: 10.1016/j.ssresearch.2014.08.016.

McMullen, Troy. 2021. “For Black Homeowners, a Common Conundrum with Appraisals.” Washington Post, January 21.

Mishel, Larry. 2022. The Missing Clause in ‘Freedom of Contract’: Full Employment. Economic Policy Institute.

Mistry, Ritesh, Gregory D. Stevens, Harvinder Sareen, Roberto De Vogli, and Neal Halfon. 2007. “Parenting-Related Stressors and Self-Reported Mental Health of Mothers with Young Children.” 97 American Journal of Public Health 1261. doi: 10.2105/AJPH.2006.088161.

Modestino, Alicia Sasser, Daniel Shoag, and Joshua Ballance. 2020. “Upskilling: Do Employers Demand Greater Skill When Workers Are Plentiful?” 102 Review of Economics and Statistics 793. doi: 10.1162/rest_a_00835.

Molloy, Raven, Christopher L. Smith, and Abigail Wozniak. 2011. “Internal Migration in the United States.” 25 Journal of Economic Perspectives 173. doi: 10.1257/jep.25.3.173.

Molloy, Raven, Christopher L. Smith, and Abigail Wozniak. 2017. “Job Changing and the Decline in Long-Distance Migration in the United States.” 54 Demography 631. doi: 10.1007/s13524-017-0551-9.

Mosthaf, Alexander, Thorsten Schank, and Claus Schnabel. 2014. “Low-Wage Employment versus Unemployment: Which One Provides Better Prospects for Women?” 3 IZA Journal of European Labor Studies 21. doi: 10.1186/2193-9012-3-21.

Mosthaf, Alexander, Claus Schnabel, and Jens Stephani. 2011. “Low-Wage Careers: Are There Dead-End Firms and Dead-End Jobs?” 43 Zeitschrift für ArbeitsmarktForschung 231. doi: 10.1007/s12651-010-0036-4.

Muhl, Charles J. 2001. “The Employment-at-Will Doctrine: Three Major Exceptions.” Monthly Labor Review (January) 3.

Naidu, Suresh, and Michael Carr. 2022. “If You Don’t Like Your Job, Can You Always Quit?” Economic Policy Institute.

Naidu, Suresh, Eric A. Posner, and Glen Weyl. 2018. “Antitrust Remedies for Labor Market Power.” 132 Harvard Law Review 536.

Neumark, David, Junfu Zhang, and Stephen Ciccarella. 2008. “The Effects of Wal-Mart on Local Labor Markets.” 63 Journal of Urban Economics 405. doi: 10.1016/j.jue.2007.07.004.

Newman, Katherine S. 2000. No Shame in My Game: The Working Poor in the Inner City. 3rd Edition. Vintage.

Nyström, Kristina, and Gulzat Zhetibaeva Elvung. 2015. “New Firms as Employers: The Wage Penalty for Voluntary and Involuntary Job Switchers.” 29 Labour 348. doi: 10.1111/labr.12055.

O’Connell, Ann. 2020. “Deadline for Returning Security Deposits, State-by-State.” Nolo. https://www.nolo.com/legal-encyclopedia/chart-deadline-returning-security-deposits-29018.html.

Pager, Devah, Bruce Western, and Bart Bonikowski. 2009. “Discrimination in a Low-Wage Labor Market: A Field Experiment.” 75 American Sociological Review 777. doi: 10.1177/000312240907400505.

Perry, Christin, and Samantha Allen. 2021. “How Much Do Movers Cost?” Forbes Advisor.

Pew Research Center. 2015. Parenting in America. https://www.pewresearch.org/social-trends/2015/12/17/parenting-in-america/.

Pew Research Center. 2018. American Families Face a Growing Rent Burden. https://www.pewtrusts.org/en/research-and-analysis/reports/2018/04/american-families-face-a-growing-rent-burden.

Quillian, Lincoln, Devah Pager, Ole Hexel, and Arnfinn H. Midtbøen. 2017. “Meta-Analysis of Field Experiments Shows No Change in Racial Discrimination in Hiring over Time.” 114 Proceedings of the National Academy of Sciences 10870. doi: 10.1073/pnas.1706255114.

Rinz, Kevin. 2018. “Labor Market Concentration, Earnings Inequality, and Earnings Mobility.” Working Paper No. 2018-10. U.S. Census Bureau.

Rothstein, Richard. 2017. The Color of Law. Liverlight.

Schmutte, Ian M. 2015. “Job Referral Networks and the Determination of Earnings in Local Labor Markets.” 33 Journal of Labor Economics 1. doi: 10.1086/677389.

Schmutte, Ian M. 2016. “Labor Markets with Endogenous Job Referral Networks: Theory and Empirical Evidence.” 42 Labour Economics 30. doi: 10.1016/j.labeco.2016.06.005.

Schneider, Daniel, and Kristen Harknett. 2019. “Consequences of Routine Work-Schedule Instability for Worker Health and Well-Being.” 84 American Sociological Review 82. doi: 10.1177/0003122418823184.

Schubert, Gregor, Anna Stansbury, and Bledi Taska. 2021. Employer Concentration and Outside Options. SSRN Scholarly Paper No. 3599454. Social Science Research Network. doi: 10.2139/ssrn.3599454.

Schultz, Michael A. 2019. “The Wage Mobility of Low-Wage Workers in a Changing Economy, 1968 to 2014.” 5 RSF: The Russell Sage Foundation Journal of the Social Sciences 159. doi: 10.7758/rsf.2019.5.4.06.

Simon, Curtis J., and John T. Warner. 1992. “Matchmaker, Matchmaker: The Effect of Old Boy Networks on Job Match Quality, Earnings, and Tenure.” 10 Journal of Labor Economics 306.

Sockin, Jason. 2021. Show Me the Amenity: Are Higher-Paying Firms Better All Around? SSRN Scholarly Paper No. 3957002. Social Science Research Network. doi: 10.2139/ssrn.3957002.

Sockin, Jason, and Aaron Sojourner. 2020. What’s the Inside Scoop? Challenges in the Supply and Demand for Information on Employers. SSRN Scholarly Paper No. 3596666. Social Science Research Network. doi: 10.2139/ssrn.3596666.

Sockin, Jason, Aaron Sojourner, and Evan Starr. 2022. Non-Disclosure Agreements and Externalities from Silence. SSRN Scholarly Paper No. 3900285. Social Science Research Network. doi: 10.2139/ssrn.3900285.

Sockin, Jason, and Daniel Zhao. 2020. Why Do Candidates Reject Job Offers? Glassdoor. https://www.glassdoor.com/research/app/uploads/sites/2/2020/02/20-02-27-Glassdoor-Econ-JobRejection-Report.pdf.

Sojourner, Aaron, and Jooyoung Yang. 2022. “Effects of Union Certification on Workplace-Safety Enforcement: Regression-Discontinuity Evidence.” 75 ILR Review 373. doi: 10.1177/0019793920953089.

Sokolova, Anna, and Todd Sorensen. 2021. “Monopsony in Labor Markets: A Meta-Analysis.” 74 ILR Review 27. doi: 10.1177/0019793920965562.

Stansbury, Anna, and Lawrence Summers. 2020. “The Declining Worker Power Hypothesis: An Explanation for the Recent Evolution of the American Economy.” Brookings Papers on Economic Activity (Spring).

Stelzner, Mark, and Kate Bahn. 2021. “Discrimination and Monopsony Power.” Review of Black Political Economy, June 18, n.p. doi: 10.1177/00346446211025646.

Stelzner, Mark, and Mark Paul. 2020. “Monopsony and Collective Action in an Institutional Context.” Review of Social Economy, October 7, n.p. doi: 10.1080/00346764.2020.1829017.

Stewart, Mark B. 2007. “The Interrelated Dynamics of Unemployment and Low-Wage Employment.” 22 Journal of Applied Econometrics 511. doi: 10.1002/jae.922.

Tenev, Nicholas H. 2020. “Social Connections and Racial Wage Inequality.” SocArXiv, May 13. https://osf.io/preprints/socarxiv/vm82w/.

Topel, Robert H., and Michael P. Ward. 1992. “Job Mobility and the Careers of Young Men.” 107 Quarterly Journal of Economics 439. doi: 10.2307/2118478.

U-Haul International. n.d. “U-Haul Truck Rentals: Moving Trucks for Local and One Way Moves.” https://www.uhaul.com/Truck-Rentals/.

Urban Institute. 2015. “Exposing Housing Discrimination.” https://www.urban.org/features/exposing-housing-discrimination.

U.S. Census Bureau. 2018. “Commuting Costs.” https://www2.census.gov/programs-surveys/ahs/2017/infographs/2017%20Housing%20Profile%20Commuting%20Costs.pdf.

U.S. Census Bureau. 2020. “Mode of Transportation to Work.” Table S0801. https://www2.census.gov/programs-surveys/commuting/guidance/acs-1yr/DY2019-Percent-drove-alone.pdf.

U.S. Census Bureau. 2021. “National Disability Employment Awareness Month: October 2021.” Stats for Stories. https://www.census.gov/newsroom/stories/disability-employment-awareness-month.html.

U.S. Department of Housing and Urban Development. 2014. “Paired Testing and the Housing Discrimination Studies.” Evidence Matters. https://www.huduser.gov/portal/periodicals/em/spring14/highlight2.html.

U.S. Department of Labor. 2021. “Unemployment Insurance Appeals Time Lapse and Average Age of Appeals.” https://oui.doleta.gov/unemploy/ui_insurance_appeal.asp.

U.S. Department of Labor. 2022. “Consolidated Minimum Wage Table.” https://www.dol.gov/agencies/whd/mw-consolidated.

Webber, Douglas A. 2015. “Firm Market Power and the Earnings Distribution.” 35 Labour Economics 123. doi: 10.1016/j.labeco.2015.05.003.

Wiczer, David. 2016. Nearly Half of Job Switchers Earn Less in Their New Roles. Federal Reserve Bank of St. Louis.

Wiemers, Emily E. 2014. “The Effect of Unemployment on Household Composition and Doubling Up.” 51 Demography 2155. doi: 10.1007/s13524-014-0347-0.

Willness, Chelsea R., Piers Steel, and Kibeom Lee. 2007. “A Meta-Analysis of the Antecedents and Consequences of Workplace Sexual Harassment.” 60 Personnel Psychology 127. doi: 10.1111/j.1744-6570.2007.00067.x.

Wilshire, Justin. 2022. Walmart Supercenters and Monopsony Power: How a Large, Low-Wage Employer Impacts Local Labor Markets. Washington Center for Equitable Growth.

The assumption of a perfectly competitive labor market is central to some of the most widely accepted theories in the field of labor economics. But the persistent threat of unemployment, in combination with prohibitive conditions imposed by employer practices, public policy, incomplete information about job opportunities, and geographic immobility, means that workers often cannot change jobs or employers easily and without cost.

This imbalance of power disproportionately disadvantages black workers: One of the most durable and defining features of the U.S. labor market is the 2-to-1 disparity in unemployment that exists between black and white workers. Attempts to explain the gap often cite observed average differences in human capital—particularly, education or skills—between black and white workers as a primary cause. But African Americans have made considerable gains in high school and college completion over the last four-and-a-half decades—both in absolute terms as well as relative to whites—and those gains have had virtually no effect on equalizing employment outcomes. Indeed, the significant racial disparities in unemployment that are observed at each level of education, across age cohorts, and among both men and women are the strongest evidence against the notion that education or skills differentials are responsible for the black-white unemployment gap.

Another defining feature of racial inequality in the labor market is the significant pay disparities between black and white workers. In 2019, the typical (median) black worker earned 24.4% less per hour than the typical white worker. This is an even larger wage gap than in 1979, when it was 16.4%. Controlling for racial differences in education, experience, and the fact that black workers are more likely to live in lower-wage Southern states leaves an unexplained gap of 14.9% in 2019 (out of a total average gap of 26.5%). This is up from an unexplained gap of 8.6% in 1979 (out of a total average gap of 17.3%). Any simple or rational explanation for this disparity is further complicated by the fact that racial wage gaps among men are significantly larger than among women.

Racial wage gaps also have widened amid the broader trend of growing wage inequality, as black workers have reaped even fewer gains from increased aggregate productivity than white workers. While net productivity per hour worked increased 69.6% (1.8% per year) between 1979 and 2019, median wages grew by only 14% (0.4% per year). Over this same time, the median wage of black workers grew at a meager 5.2% (0.1% per year) and the median wage of white workers grew 20.0% (0.5% per year).

The economic theories most often invoked to explain observed racial differentials in labor market outcomes are human capital theory, taste-based models of discrimination, and statistical models of discrimination. But each of these models falls short in its attempt to explain long-standing racial disparities in unemployment and pay while blatantly denying the persistence of discrimination. Despite compelling empirical evidence and a solid historical record that points to discrimination as a significant factor in the persistence of racial disparities in the labor market, the interpretation of those disparities is an ongoing debate in the field of economics.

When economists get a statistically significant coefficient on race after estimating a wage equation that controls for standard measures of individual productive capacity (e.g., education, experience) and macroeconomic conditions (e.g., state or regional fixed effects), as well as race and gender, what does that mean? Do we interpret that coefficient as evidence of racial discrimination, or does it reflect some unobserved or omitted variable? Devotees of the conventional economic theories described above tend to dismiss discrimination as a valid or significant explanation of the gaps in favor of the latter interpretation. But if there is some unobserved variable that would explain the statistically significant coefficient on race, it would also have to be strongly correlated with race. In its most basic form, race is nothing more than a socially constructed identifier, defined in the United States primarily by skin color—an arbitrary and superficial physical characteristic that has no relationship to one’s productive capacity. How then should we interpret that correlation?

Stratification economics was developed in response to the inadequacy of conventional economic theory to explain intergroup inequality in general and the persistence of racial disparities in particular. According to stratification economics, while discrimination is unjust, it also serves the functional role of preserving hierarchy. Therefore, persistent racial inequality arises when a dominant group seeks to maintain the hierarchy that affords it some degree of social or economic privilege. Under this framework, identity can be structured so that investing in, or associating with, a group identity can lead to economic returns and benefits. This treatment of identity as endogenous represents a major departure from more conventional economic models but is consistent with a set of alternative theories for explaining stubborn racial gaps in economic outcomes.

When we look at race and labor market discrimination in the context of workers’ bargaining power, it is important we recognize there are at least two complementary goals. With respect to wages, we want to shift the balance of power in a way that puts upward pressure on wages—particularly for wage earners at or below the median—and at the same time close racial wage gaps. We cannot rely on competitive markets alone to do this. Rather, interventions are required to address these inequalities. Appropriate design of those interventions requires that we expand the frameworks we use for understanding power, race, gender, and inequality so that we restructure systems and institutions to prevent discriminatory outcomes rather than enable them. 

Introduction 

The assumption of a perfectly competitive labor market is central to some of the most widely accepted theories in the field of labor economics. On the demand side of this market structure are many firms seeking to fill identical jobs. On the supply side are many workers who possess the same set of requisite skills for a given job opening, all of whom have perfect information about wages and job conditions and are able to move their labor freely, or without cost. The equilibrium price and quantity of labor—the market wage and level of employment, respectively—are those at which the amount of labor supplied by workers is equal to the amount of labor demanded by firms. Workers are paid the marginal product of their labor, and, in the long run, such a perfectly competitive labor market is theoretically at “full employment,” since all who are willing to work at the market wage can find a job that pays that wage.

Under these market conditions, employers are assumed to be wage takers: They are unable to hire or retain workers for less than the going market wage because no workers would willingly accept a job for less when they could easily transfer their labor to a competing firm that pays the market wage. In such markets, any differences in wages must be due to differences in productivity-related human capital.

In reality, however, labor market structures are far from perfectly competitive, and employers are rarely so powerless as to have no discretion in setting wages below marginal productivity or in paying different wages to equally productive and qualified workers. Since there is no absolute empirical standard of full employment—a condition implied by perfect competition—economists often disagree over what the “full employment” unemployment rate is or should be. Nevertheless, the nation’s actual unemployment rate has been above even the Congressional Budget Office’s far-too-conservative estimates of the “natural rate of unemployment”—i.e., the NAIRU, the nonaccelerating inflation rate of unemployment—more often than not during the last 40 years (Bivens and Zipperer 2018), a period characterized by rising inequality. Indeed, since 1979 the monthly unemployment rate has been below 6% only approximately half the time; for blacks the rate has fallen below 6% for only a brief five months preceding the Covid-19 pandemic, and it has remained nearly double the national rate as racial wage gaps have widened.

The threat of unemployment, in combination with prohibitive conditions imposed by employer practices, public policy, incomplete information about job opportunities, and geographic immobility, means that workers often cannot change jobs or employers easily and without cost. This imbalance of power between employers and employees disproportionately disadvantages black workers when racial identity is used to assign privilege or disadvantage in the labor market context.

This paper presents persistent racial inequality in unemployment and wages as outcomes that have been ignored or dismissed and remain unsatisfactorily explained by conventional economic theory. For instance:

• The significant racial disparities in unemployment that are observed at each level of education, across age cohorts, and among both men and women are the strongest evidence against the notion that education or skills differentials are responsible for the black-white unemployment gap.
• Less than half of the observed black-white difference in average hourly wages is explained by differences in education, experience, or region—the main factors presumed to determine pay. However, changes in the racial wage gap track closely with changes in policy, such as civil rights enforcement, and with structural trends contributing to greater wage inequality. 
• Black workers in the public sector face smaller unexplained wage gaps than their counterparts in the private sector. Historically, the appeal of better job opportunities and greater pay equity in the public than in the private sector has contributed to black workers’ disproportionate employment share in the public sector as well as higher average rates of union membership.
• The economic theories most often invoked to explain the observed racial differentials in labor market outcomes—human capital theory, taste-based models of discrimination, and statistical models of discrimination—are historically and empirically inconsistent with the persistence of black-white wage disparities. 

• In contrast to conventional economic models, stratification economics treats group identity (race) as a construct and acknowledges the persistence of racial inequality resulting from discrimination’s functional role in preserving a hierarchy that benefits the dominant group.

The paper proceeds as follows. Section I summarizes the data describing racial disparities in unemployment, and Section II looks at wages. In Section III we highlight examples of how disparities in bargaining power play out for black workers. In Section IV we review the prevailing economic theories used to interpret racial inequalities in labor market outcomes and present challenges to and shortcomings of those theories. In Section V we present stratification economics as a more appropriate framework for understanding the imbalance of power inherent in the social structures that perpetuate racial inequality in labor market outcomes. 

I. Racial disparities in unemployment

One of the most durable and defining features of the U.S. labor market is the large and persistent disparity in unemployment that exists between black and white workers. This disparity is well-documented in decades of publicly reported official estimates from the Bureau of Labor Statistics (BLS) dating back to 1954, when the agency first began reporting rates of unemployment by race (i.e., white and nonwhite1). In 1972, BLS began disaggregating the nonwhite unemployment rate and reporting an unemployment rate for blacks alone. According to this measure, black job seekers are about half as likely to secure employment during a consecutive four-week search period as are white job seekers. Figure A illustrates this pattern, showing that the ratio between the black and white unemployment rates has consistently been about 2-to-1 since 1972. The pattern has persisted across multiple periods of economic growth and contraction, including in 2019 when, after 10 years of job growth, the black unemployment rate fell to a historic low of 6.1% but was still twice as high as the white unemployment rate of 3.0%.

Figure A

Attempts to explain the black-white unemployment rate gap often cite observed average differences in human capital—particularly, education or skills—between black and white workers as a primary reason for the disparity. While conventional human capital theory does not explain the presence of unemployment apart from wage or price rigidities, patterns of unemployment by educational attainment and age are clearly documented in national statistics. According to these data, those with higher levels of education and more potential work experience, as indicated by their age, tend to have lower rates of unemployment than those with lower levels of education and less work experience.

However, observed racial differences in education fail to account for the 2-to-1 black-white unemployment rate disparity, and an exposition of trends in educational attainment and unemployment by race helps to clarify why this explanation falls short. African Americans have made considerable gains in high school and college completion over the last four-and-a-half decades—both in absolute terms as well as relative to whites—but those gains have had virtually no effect on equalizing employment outcomes between black and white workers.

Figure B shows that in 1972 fewer than four out of 10 (36.6%) African American adults age 25 or older had a high school diploma. By 2019 that share had grown to almost nine out of 10 (87.9%); among 25- to 29-year-old African Americans the share had grown to more than nine out of 10 (91.0%), indicating a continuation of the longer-term upward trend. This large increase in high school completion rates among black students helped to narrow the black-white difference in high school completion. In 1972, African Americans trailed whites by 23.8 percentage points (60.4% of whites, compared with 36.6% of blacks). In the most recent data, the difference is only 6.7 percentage points (94.6% for whites versus 87.9% for African Americans).2 

Figure B

College graduation rates have also increased for African Americans. Figure C shows that among those age 25 or older just 5.1% had a four-year college degree in 1972, but by 2019 that share had grown to 26.1%, a fivefold increase. Over the same period, college completion also expanded for whites, but the increase was just over threefold, from 12.6% in 1972 to 40.1% in 2019. As a result, the relative situation of African Americans also improved over this time: In 1972 blacks were just 40.5% as likely as whites to have a four-year college degree (12.6% for whites and 5.1% for blacks), compared to 71.9% today (40.1% for whites and 26.1% for blacks). 

Figure C

However, focusing only on the share of people with a four-year college degree obscures the broader shift to a more highly educated black workforce. Currently, a majority of black high school graduates (55.3%) go on to pursue some level of postsecondary education. More than one-fourth (29.2%) of African Americans age 25 or older have some college education, even if they are not bachelor’s degree holders. This includes 10.6% who earned a two-year associate degree.

Further, there is evidence that, compared to white youth from families with similar income levels, black students are actually more likely to seek higher levels of education, in part because they have fewer less-formal opportunities for economic advancement, such as social networks, family relationships, and institutional mechanisms (Mason 1997; Mangino 2010, 2012). But while these investments in higher education improve employment prospects for black college graduates relative to black noncollege graduates, on average they do not yield outcomes equivalent to those of similarly educated whites.

The significant racial disparities in unemployment that are observed at each level of education (Figure D) are the strongest evidence against the notion that education or skills differentials are responsible for the black-white unemployment gap. In terms of education, the black-white unemployment rate ratio has hovered around 2-to-1 at every level for most of the last 41 years. In that time, only black workers with advanced degrees have approached anything near parity with their white counterparts, as measured by the unemployment rate. In practical terms, this means that black workers are not just twice as likely to be unemployed as similarly educated white workers, but they are often more likely to be unemployed than less-educated whites.

Figure D

The 2-to-1 ratio is also remarkably consistent across age cohorts (Figure E) and among both men and women (Figure F). Older black workers have lower rates of unemployment than younger black workers, but in every age cohort black workers remain roughly twice as likely to be unemployed as white workers.

Figure E

Figure F

These empirical data are consistent with field experiments revealing that black job applicants with equivalent, and sometimes superior, credentials to white applicants are less likely to receive job callbacks (Turner, Fix, and Struyk 1991; Fix, Galster, and Struyk 1993; Bendick, Jackson, and Reinoso 1994). Among the starkest findings in this regard is the audit study of Pager (2003), demonstrating that employers treated whites with criminal records more favorably than blacks without criminal records. Agan and Starr (2018) find that failure to distinguish between applicants with criminal records and those without, as is done through Ban the Box policies, actually reduces outcomes for black applicants without a criminal record. Though research on the impact of Ban the Box policies yields mixed results, discrimination against black workers remains the central unresolved issue. 

While audit and correspondence studies have been criticized for not adequately capturing unobserved characteristics that might influence hiring decisions, Neumark (2012) has shown how robust these findings can be to such considerations. Quillian et al. (2017) show that subsequent field experiments reveal a pattern of discrimination experienced by blacks in particular that has remained constant over time.

Together, these patterns strongly suggest that racial discrimination—and not inadequate education or lack of skills on the part of black workers—is the most plausible explanation for persistent racial disparities in unemployment. Moreover, currently observed racial differences in employment and education have been shaped by the United States’ long history of racial oppression that outright denied or severely limited black American access to the same formal educational and employment opportunities available to whites.

II. Racial disparities in pay 

Another defining feature of racial inequality in the labor market is the significant pay disparities between black and white workers. Most empirical research on black-white wage inequality has taken one of two approaches to estimating and explaining observed differences in pay. Trend analysis has focused on understanding the causes behind the black-white wage gap’s four distinct periods of change: the gap’s dramatic narrowing during the latter part of the 1960s through the 1970s, the reversal of that pattern during the 1980s, the brief period of improvement during the late 1990s, and the post-2000 expansion of the gap.

The other common approach has been to examine how much of the observed pay gap between black and white workers can be attributed to differences in so-called “cognitive skills.” In this section, we focus primarily on presenting patterns and trends in black-white wage disparities between 1979 and 2019. We provide our review and critique of the cognitive skills and human capital literature in subsequent sections.

In 2019, the typical (median) black worker earned 24.4% less per hour than the typical white worker. This is an even larger wage gap than in 1979, when it was 16.4%. Black workers face these significant and growing pay penalties relative to white workers even after controlling for characteristics assumed to be related to productive capacity, like education and experience.

As shown in Figure G, controlling for racial differences in education, experience, and the fact that black workers are more likely to live in lower-wage Southern states leaves an unexplained gap of 14.9% in 2019 (out of a total average gap of 26.5%). This is up from an unexplained gap of 8.6% in 1979 (out of a total average gap of 17.3%).

Figure G

Any simple or rational explanation for this disparity is further complicated by the fact that racial wage gaps among men are significantly larger than among women. Over this same period, the unexplained black-white wage gap increased 7.3 percentage points among men (from 8.6% in 1979 to 14.9% in 2019) and 6.8 percentage points among women. Notably, the unexplained portion of the racial wage gap among women was minimal (1.4%) in 1979 but had expanded to 8.6% by 2019.

These patterns run counter to the notion that productive capacity, as measured by education specifically, is the prevailing determinant of wages. Less than half of the observed black-white difference in average hourly wages is explained by differences in education, experience, or region—some of the main factors presumed to determine pay. While black-white pay differentials are smaller among women than among men, the intersection of race and gender imposes much larger wage penalties for black women relative to white men. As shown in Figure H, in 2019 black women were paid 33.7% less than their white male counterparts, which was a much larger gap than that faced by either white women (25.7%) or black men (22.2%).

Figure H

The trend analysis research finds that changes in the racial wage gap track closely with changes in policy. The narrowing of the gap from the late 1960s through the 1970s can be attributed to the passage of important civil rights legislation (Bound and Freeman 1989; Card and Krueger 1992; Donohue and Heckman 1991), combined with the 1960s economic boom, active enforcement of anti-discrimination and affirmative action policy (Betsey 1994; Fosu 1992; Heckman and Payner 1989; Leonard 1990), and the narrowing of the educational attainment gap between blacks and whites (Carlson and Swartz 1988; Cunningham and Zalokar 1992; Zalokar 1990). On the other hand, the widening of the gap during the 1980s was the result of retrenchment on anti-discrimination policy (Leonard 1990), growing general wage inequality (Blau and Beller 1992), deterioration in the manufacturing sector, and a decline in union representation (Bound and Freeman 1989; Wilson and Rodgers 2016).

The expansion of the black-white wage gap since the 1980s, and certainly during the post-2000 period, is also consistent with the structural trends contributing to greater wage inequality. These include: (1) limited wage growth among middle- and low-wage workers (Gould 2020); (2) above-average growth among the highest-wage workers, particularly chief executive officers (CEOs) and the top 1% (Mishel and Kandra 2021); and (3) racial inequality in hiring, pay, and opportunities for promotion that results in overrepresentation of black workers among low- to middle-wage occupations and underrepresentation among high-wage occupations (Hamilton, Austin, and Darity 2011; Abayomi and Hawkins 2009).

The divergence of productivity growth and median hourly wage growth also points to growing wage inequality in a way that challenges assumptions about competitive labor markets. In competitive labor markets, where it is assumed that employers have no power to set wages below the market wage and workers are paid a wage equal to their marginal productivity, productivity and wages should move together. While this was generally true in the three decades following World War II, beginning around 1973 inflation-adjusted hourly compensation (including employer-provided benefits and wages) grew at a markedly slower rate than economywide productivity. This pattern is documented by Bivens and Mishel (2015), who estimate that over two-thirds of this productivity-pay gap can be explained by rising inequality, which is characterized by greater inequality in compensation and the fact that a smaller share of national income has been going to workers relative to capital owners.

The emergence of the productivity-pay gap calls into question the assumption of wage-taking behavior on the part of employers. An emerging literature on monopsony offers a broader interpretation of employers’ power that goes beyond the simple definition of labor market concentration (i.e., the proverbial one-company town). Rather, monopsony encompasses any power employers have that allows them to cut wages without fear of losing a large share of their workers. While there are several studies estimating employees’ likelihood to exit jobs in response to wage changes (Webber 2015, 2020; Dube, Giuliano, and Leonard 2019; Dube et al. 2020; Bassier, Dube, and Naidu 2020; Azar, Marinescu, and Steinbaum 2019; Langella and Manning 2020; Card et al. 2018; Sokolova and Sorensen 2020), Webber (2020, 18) succinctly concludes: 

The majority of firms compete for workers in labor markets where the typical employee is highly unlikely to move in response to small or even modest changes in their wage. This gives these firms considerable latitude to pay lower wages without worrying about a mass exodus of employees. 

Further, Bassier, Dube, and Naidu (2020) report that employers can “mark down” wages by anywhere from 20% to 50%. While evidence of rising monopsony power in the years since the late 1990s is mixed, studies consistently find that employers can exert more power over low-wage workers, affirming the link between employer power and wage inequalities. Among low-wage hourly workers, indirect wage cuts can also take place in the form of unstable work hours that are too low to qualify the worker for employer-provided benefits, such as health insurance or retirement savings. 

Racial wage gaps also have widened amid the broader trend of growing wage inequality, as black workers have reaped even fewer gains from increased aggregate productivity than white workers. Figure I shows that between 1979 and 2018 median hourly real wage growth fell far short of productivity growth. While net productivity per hour worked increased 69.6% (1.8% per year) during this period, median wages grew by only 14% (0.4% per year). Over this same time, the median wage of black workers grew at a meager 5.2% (0.1% per year) and the median wage of white workers grew 20.0% (0.5% per year).

Figure I

It is also clear from Figure I that the strongest period of wage growth during this time occurred between the mid-1990s and early 2000s. The period of low unemployment and strong wage growth between 1995 and 2000 has been cited as a key contributor to some brief narrowing of the black-white wage gap during this time (Wilson 2015), while others assert that the rise in mass incarceration during the 1990s is responsible for artificially increasing the average wage of black men by removing a disproportionate share of those who were “less skilled” or lower-wage earners from the labor force (Neal and Rick 2014).

Since 2000, the black-white wage gap has continued to widen (Wilson and Rodgers 2016; Gould 2020). One of the most troubling trends of the post-2000 period has been the fact that the black-white wage gap has grown most among workers with a bachelor’s degree, and discriminatory differentials are also higher among the more highly educated (Tomaskovic-Devy, Thomas, and Johnson 2005; Wilson and Rodgers 2016). In fact, while the tight labor market of the late 1990s delivered faster wage growth to black college graduates than white college graduates in the five-year period from 1996 to 2000, the wages of black college graduates fell between 2015 and 2019—corresponding to the last five years of the recovery from the Great Recession, when unemployment rates were closest to those during 1996–2000 (Figure J). By contrast, the wages of white college graduates increased between 2015 and 2019 (Gould and Wilson 2019).

Figure J

III. Examples of bargaining power in action 

The imbalance of power between employers and employees is both a cause and consequence of the racial disparities in labor market outcomes that we have detailed above. One of the things that gives an employee or potential employee greater leverage at the bargaining table is the existence of equally or more attractive employment alternatives—a condition that is facilitated by tighter labor markets. However, over the last four decades there has been insufficient vigilance in fighting unemployment.

As shown in Figure K, between 1979 and 2019 the actual unemployment rate exceeded estimates of the NAIRU by an average of roughly 0.8 percentage points each year. Failure to meet even this arguably too-conservative employment target has weakened the bargaining power of the vast majority of workers, as evidenced by growing wage inequality over this period. Black workers have suffered some of the greatest harm from policy decisions that allowed excessive unemployment in pursuit of misguided inflation targets intended to limit wage growth.

Figure K

The widening of wage gaps between black and white workers over these same years further suggests that the perpetual 2-to-1 unemployment disparity further eroded black workers’ bargaining power relative to white workers. This diminished leverage can affect a worker’s willingness to challenge unfair and unsafe working conditions given concerns about how long it may take to find another job if the worker were to leave or be terminated from the current job as an act of employer retaliation. Long-standing racial disparities in income and wealth also raise the stakes associated with leaving or losing a job more for black workers than for white workers.

Race, unequal power, and the Covid-19 crisis 

The Covid-19 pandemic and recession offer the most recent example of the resilience of racial inequality and stratification in the labor market and how they generate disparate outcomes and unequal bargaining power. When businesses, schools, and other public places responded to the public health crisis in mid-March 2020 by simultaneously closing their doors, three distinct groups of workers quickly emerged. The first group included tens of millions of workers who lost jobs. Workers in the second and third groups both retained their jobs but under very different conditions.

“Essential” frontline workers were required to continue physically reporting to work, while those not in that category were able to work remotely from the safety of their homes. Black workers were least likely to be among those able to retain employment under remote working arrangements: Less than one-fifth had the option to work from home, compared to almost one-third of white workers. Therefore, black (along with Hispanic and Native American) workers were more likely than whites to suffer job loss or be compelled to put their health at risk in exchange for a measure of job security (Gould and Wilson 2020).

One of the structures contributing to such racially disparate impacts is occupational segregation, characterized by overrepresentation of black workers, and especially black women workers, in low-wage occupations and underrepresentation in higher-wage occupations. The Covid-19 crisis popularized the term essential worker, drawing attention to the fact that black workers occupy a disproportionate share of lower-wage jobs in major frontline industries, often with inconsistent work hours (thus, inconsistent pay) and without paid leave or employer-provided health coverage. Rho, Brown, and Fremstad (2020) report that black workers make up a disproportionate share of frontline workers across six sectors of the economy that are considered essential.

While black workers represent 11.9% of all workers, they make up about one in six (17%) of frontline-industry workers. This category includes employment in public transit (26.0%); child care and social services (19.3%); trucking, warehouse, and postal service (18.2%); health care (17.5%); and grocery, convenience, and drug stores (14.2%).

Except for those in health care, none of these workers had any prior professional obligation that would require them to put their health at risk. Absent policy intervention, union representation, or a sympathetic employer, few had any assurances that they would be compensated for the increased risk.

Public-sector vs. private-sector racial wage gaps and the role of unions 

Given the amount of power an employer holds over any individual worker, it becomes necessary to establish a countervailing force that builds sufficient power among workers through a stronger collective voice with which to advocate for higher pay, better benefits, training and promotional opportunities, and protections against discrimination and harassment. In a unionized workforce, for example, collective bargaining results in labor contracts that help to create greater transparency and consistency through clearly defined policies and pay structures. These contracts play a critical role in reducing the potential for pay discrimination by limiting an employer’s discretion in paying different wages to comparably qualified individuals doing the same job and providing workers with critical protections and recourse against other forms of exploitation or mistreatment.

These conditions are more likely to exist in the public sector than in the private sector because in the former a larger share of the workforce is covered by a union contract: 39% of public-sector employees are in a union or covered by a union contract, compared to only 7% of private-sector workers. Historically, the appeal of better job opportunities and greater pay equity in the public than in the private sector has contributed to black workers’ disproportionate employment share in the public sector as well as higher average rates of union membership. Based on empirical analysis of wages among public- and private-sector employees, black workers in the public sector face smaller unexplained wage gaps than their counterparts in the private sector—6.9% versus 16.9%, respectively.3 Within the public sector, postal service jobs have been particularly valuable to black workers because of the uniform wage and benefit structure (all postal employees who have the same job title and job tenure are paid the same nationwide) and higher pay relative to comparable private-sector employment. However, since the 1980s the postal service has been under sustained assault by those who believe the compensation provided by these jobs is too generous.

Newly developed historical data from the early postwar period affirm that collective bargaining has been an effective tool for reducing wage inequality. Based on data compiled for men in several U.S. cities in 1951, Callaway and Collins (2017) found “the [union] wage premium was larger at the bottom of the income distribution than at the middle or higher, larger for African Americans than for whites, and larger for those with low levels of education,” findings that are “consistent with the view that unions substantially narrowed urban wage inequality at mid-century.”

Using data on union households from Gallup surveys dating back to 1936, Farber et al. (2021) similarly found that unions raised wages “between ten and twenty log points, with the less-educated receiving an especially large premium.” While this union effect has been relatively consistent over the last 80 years, patterns of union membership have not been. Unions’ disproportionate representation of “disadvantaged” workers (i.e., not white and not college-educated) began in the mid-1940s and peaked during the 1960s. While black workers continued to have higher rates of union membership than whites in the decades since, as overall union density has declined rates of union membership for black and white workers have converged.

Despite unions being a powerful force for increasing wages among the working class, racism within the labor movement has at times served to perpetuate rather than reduce racial inequality. While racially integrated unions have been instrumental in building support for policies that benefit black workers (Day 2020), tragically there are also examples of white workers using their unions to defend rather than defeat white supremacy.

IV. Challenging the prevailing economic theories used to explain racial disparities in labor market outcomes 

The economic theories most often invoked to explain the observed racial differentials in labor market outcomes described above are human capital theory, taste-based models of discrimination, and statistical models of discrimination. In this section, we briefly review these models and their core assumptions. We then make a case for why each of these models is historically and empirically inconsistent with the facts. 

Challenges to the conventional wisdom of human capital theory 

Adam Smith first introduced the concept of human capital as having an economic value analogous to physical capital in Wealth of Nations. Building upon this concept, Mincer (1958), Schultz (1961), and Becker (1964) popularized what we now know as human capital theory, formalizing a relationship between education, productivity, and earnings. Productivity is the assumed link between education and earnings in the Mincerian earnings function that operationalizes modern human capital theory.

Human capital theory posits that a worker’s earnings are related, directly and solely, to the worker’s productive capacity, represented by an individual’s particular set of skills, knowledge, and abilities, or human capital. Workers can increase their earnings by investing in human-capital-enhancing activities that, presumably, make them more productive. While human capital investments can take multiple forms, including formal education and on-the-job training, we will focus on formal education, the most frequently studied and most often deployed explanation for observed labor market differentials between black and white workers.

For a majority of economists, education is treated as a means of skill accumulation, which translates into greater productivity and higher compensation. However, Darity and Underwood (2021) argue there are three major issues with the human capital theory of wages: (1) it is difficult to precisely calculate productivity, and evidence on the link between more education and increased productivity is far from definitive; (2) the huge wage differential between CEOs, or even the top 1% of all wage earners (Bivens and Mishel 2013), and the typical “line” worker creates an important conundrum; and (3) of primary importance to the discussion outlined in this paper, the presence of labor market discrimination directly contradicts human capital theory.

Regarding the first point, the fundamental question raised by Darity and Underwood (2021) is, what is the function of higher education? Does it generally impart skill- or productivity-enhancing knowledge, or do credentials function as artificial entry barriers to certain occupations? Of primary concern here is the difficulty in precisely calculating one’s productivity apart from the circular move of using wages as a proxy.

We might conceive of one set of occupations where the skills-productivity nexus is consistent with the association between higher education and higher earnings, another set of occupations where the exclusion associated with gaining the credential produces scarcity that raises earnings, and a third set of professions where both factors are at play in determining earnings. The problem is there is no existing research of which we are aware that approaches the connection between higher education and higher earnings, empirically. Thus, we have no way of quantifying which or for how many jobs higher education—or education at any level—leads to a greater capacity to perform job functions. In jobs where employers rate their employees, there is considerable evidence that job performance ratings do not rise uniformly with higher levels of educational attainment for employees. Rather, in some instances there is an inverse relationship between employer ratings and employee educational attainment (Berg 1970).

A second challenge to human capital theory lies with the mere existence of corporate “super salaries.” In 2019, the average CEO at one of the top 350 largest U.S. firms (by sales) earned almost 320 times what the typical employee in those firms’ same industry earned ($21.3 million versus $66,800), constituting an increase of $2.6 million for CEOs and $1,100 for employees from 2018 (Mishel and Kandra 2021). If one were to accept human capital theory at face value, this differential would imply that the average CEO is 320 times as productive as the typical worker, a claim that has the ring of absurdity about it, even if there was a reasonable way to measure the difference.

But one need not consider the extreme cases of corporate executives’ huge compensation packages to recognize the difficulty of our lack of criteria for measuring productivity apart from comparative earnings. The Georgetown University Center on Education and the Workforce’s standard for a good job is one that pays at least $35,000 for workers 25–44 and at least $45,000 for workers 45–64. Inclusive of workers with a bachelor’s degree or higher, this corresponds to median earnings of $65,000 (Carnevale et. al 2018). That is consistent with the average annual compensation for tractor/trailer drivers. Is an investment banker who earns $250,000 in the same year necessarily close to four times as productive? Furthermore, if we did have a reliable independent standard for measuring productivity and found that the investment banker truly is about four times as productive, would that higher level of productivity be attributable to the banker having had a university education? 

The third challenge to human capital theory is the phenomenon of labor market discrimination as evidenced by the persistence of racial disparities in wages and employment that cannot be accounted for by differences in skills. According to human capital theory, black-white differences in unemployment and earnings can largely be explained by black-white differences in skills or education. While, empirically, higher levels of educational attainment are associated with lower average rates of unemployment and higher average wages, there is also strong evidence that the returns to educational attainment are unequal for black and white workers.

By definition, racial discrimination results in an unfair devaluation of black labor, since equivalently productive workers—or in the context of higher education, workers with similar advanced degrees—receive different pay. Thus, discrimination drives a wedge between any ostensible consistent relationship between wages, skills, and productivity. 

Nevertheless, deniers of labor market discrimination as an explanation for the persistence of the black-white wage gap raise questions about whether there are other unobserved characteristics that explain these differences. The literature seeking to explain the racial wage gap as a racial gap in cognitive skills has been controversial, to say the least. This body of research stemmed from cross-sectional analysis of the National Longitudinal Survey of Youth (NLSY), which included scores from the Armed Forces Qualification Test (AFQT). These AFQT scores were used as a proxy for cognitive skills in the estimation of relative wage differentials among black and white men in their 20s and 30s. The finding that inclusion of AFQT scores in a standard human capital model substantially reduced the black-white wage gap (O’Neil 1990; Maxwell 1994; Neal and Johnson 1996) led some to conclude that racial discrimination in the labor market was not a significant factor in the persistence of the wage gap because racial disparities in wages could be explained by differences in cognitive skills.

The validity of that conclusion was challenged on multiple grounds. First, because there was little overlap in AFQT scores of blacks and whites, high correlation between race and test scores presented a problem in estimating the effect on the wage gap (Ferguson 1995; Rodgers and Spriggs 1996). Using AFQT scores to explain black-white pay gaps was also challenged on the basis of the lack of robustness to different model specifications (Mason 1998; Goldsmith, Veum, and Darity 1997). For example, Mason (1998) found that a different measure of intelligence reported in the Panel Study of Income Dynamics (PSID) failed to significantly explain the racial wage gap. Goldsmith, Veum, and Darity (1997) used data from the NLSY to show that the AFQT no longer explains the racial wage gap once psychological measures of “self-esteem” and “locus of control” are included in the wage equation.

Finally, the argument that there are unobserved variables that account for relatively lower skills among black workers is inconsistent with the fact that the racial wage gap is different for men and women (Darity, Guilkey, and Winfrey 1996; Wilson and Rodgers 2016). In fact, given that black women had a slight wage advantage over white women as recently as the early 1980s—something that has never been observed among men—it seems unlikely that some unobserved or unexplained pre-market force at once disadvantages the skill attainment of black men relative to white men while providing an advantage for black women over white women.

Challenges to the conventional wisdom of taste-based model of discrimination 

Becker’s (1957) taste-based model of discrimination is perhaps the best-known neoclassical competitive model used to explain labor market discrimination. Becker’s original model has three central assumptions: (1) labor markets are competitive and employers are motivated by profit maximization; (2) black and white workers are equally productive; and (3) whites have an “externally” acquired “taste for discrimination,” functioning as a preference for white workers. This model posits that discrimination is in fact intentional, if not rational, and allows tastes for discrimination to function through three kinds of agents: employers, employees, and customers.

The biggest challenge to Becker’s taste-based model of discrimination lies in the conclusion that discriminatory wage outcomes are only temporary. In the context of black-white wage differentials, the preference for white workers, or, equivalently, the distaste for black workers, requires black workers to compensate discriminating employers by accepting lower wages. While the discriminatory tastes of employers, employees, or customers may create incentives for workplace segregation, in the long run racial wage differentials are eliminated through competition.

This conclusion is directly refuted by the historical record outlined in Section II, which demonstrates that anti-discrimination policy intervention, not market competition, was responsible for the most significant narrowing of racial wage gaps in the decade following the passage of the Civil Rights Act of 1964, and that lax enforcement of those laws in the decades since, along with other policies that weakened the bargaining power of workers, have contributed to further widening of those gaps.

A second weakness of this model is that it assumes full employment, since all job seekers are presumed hired. The set of possible outcomes then are differentiated by the racial composition of a firm’s workforce, but unemployment—the possibility that someone actively seeking employment will not be hired at all—is not a consideration. 

Challenges to the conventional wisdom of statistical models of discrimination 

Statistical models of discrimination, pioneered by Phelps (1972) and Arrow (1973), allow for less overt and unintentional means by which disparate or discriminatory labor market outcomes manifest. These models are based on the idea that employers have incomplete information about the actual productivity of individual job applicants. However, they may have beliefs about the average productivity of a given group of workers (e.g., black workers, women, or formerly incarcerated individuals) that they assign to individuals belonging to that group. Based on those beliefs, workers belonging to the group assumed to be less productive will have a lower probability of being hired (i.e., a higher unemployment rate) or, when hired, will be offered a lower wage, resulting in a wage gap.

According to statistical models of discrimination, profit-maximizing employers in an environment of imperfect information believe they can distinguish visually between candidates from groups A and B that are drawn from different frequency distributions for ability to perform. Based on the assumption that the mean of group B’s ability distribution is markedly higher than the mean of A’s ability distribution, and there is little difference, if any, in the variance, these employers display a preference for members of group B. 

If, in fact, these employers are wrong and the two distributions are identical, then that fact should be learned over time. As a result, any profit-maximizing employer should become indifferent between members of groups A and B. On the other hand, if those employers are correct, then subsequent inequalities in intergroup outcomes are due to average differences in ability and there is no need for a theory of discrimination at all.

Confounding matters more, the informational assumption that leads employers to rely on knowledge, whether accurate or not, about group affiliation in making individual hiring decisions seems tenuous. It suggests that employers are incapable of making a sound approximation about a candidate’s future potential to perform without relying on the additional signal of group affiliation. This is implausible given the vast resources corporations devote to hiring decisions and the design of screening mechanisms. Over time, an appropriate set of questions or tests should emerge that will facilitate selection, regardless of group affiliation. 

In his open letter to economists, AFL-CIO Chief Economist William Spriggs calls out the inherent racism in the statistical discrimination framework by asking:

How does a model assume that an entire set of actors, observing the infinite diversity of human beings, all settle on race as a meaningful marker independent of history, laws, and social norms? And, miraculously, those same ‘rational’ actors use ‘statistical’ methods to find only negative attributes highly correlated with race. (Spriggs n.d.)

In other words, the only logical reason for taking account of group affiliation or race is to discriminate on the basis of race, and not to improve the accuracy of predictions of an individual’s performance. 

V. Stratification economics as a better framework for understanding persistent racial disparities in the labor market 

Each of the aforementioned economic theories falls short in its attempt to explain long-standing racial disparities in unemployment and pay while blatantly denying the persistence of discrimination. Despite compelling empirical evidence and a solid historical record that points to discrimination as a significant factor in the persistence of racial disparities in the labor market, the interpretation of those disparities is an ongoing debate in the field of economics.

When economists get a statistically significant coefficient on race after estimating a wage equation that controls for standard measures of individual productive capacity (e.g., education, experience) and macroeconomic conditions (e.g., state or regional fixed effects), as well as race and gender, what does that mean? Do we interpret that coefficient as evidence of racial discrimination, or does it reflect some unobserved or omitted variable? Devotees of the conventional economic theories described above tend to dismiss discrimination as a valid or significant explanation of the gaps in favor of the latter interpretation. But if there is some unobserved variable that would explain the statistically significant coefficient on race, it would also have to be strongly correlated with race. In its most basic form, race is nothing more than a socially constructed identifier, defined in the United States primarily by skin color—an arbitrary and superficial physical characteristic that has no relationship to one’s productive capacity. How then should we interpret that correlation?

Getting an answer to that last question requires that we go beyond our standard individual-centered models and consider structural and institutional factors, as well as what we can learn from social, psychological, and historical analyses.

In the absence of that, we continue to lean on human capital models and models of discrimination to our peril. Those models have been inadequate to explain well-documented and persistent patterns of racial inequality, leading to an overemphasis on the shortcomings of individuals and little or no emphasis on fixing biased or discriminatory systems that uphold an economic hierarchy predicated on race. This is the hole that stratification economics was developed to fill.

Stratification economics was developed in response to the inadequacy of conventional economic theory to explain intergroup inequality in general and the persistence of racial disparities in particular. In contrast to Becker’s taste-based discrimination model, or models of statistical discrimination, stratification economics includes a set of theories or models that identify the structural processes that enable the persistence of discrimination and inequality over the long term. This framework emphasizes the effect of group formation, group identity, and group action on an individual’s life outcomes, as opposed to a more conventional framework in which individuals act solely as autonomous optimizers. Stratification economics employs an interdisciplinary approach that incorporates economics, sociology, and social psychology while proposing that one’s relative position matters. Individuals discern relative position by making both intragroup and intergroup comparisons, but a key feature of intergroup comparison is the identification of an outsider group, defined by race, ethnicity, gender, class, religion, or some other demographic characteristic (Darity et al. 2017).

According to stratification economics, while discrimination is unjust, it also serves the functional role of preserving hierarchy. Therefore, persistent racial inequality arises when a dominant group seeks to maintain the hierarchy that affords it some degree of social or economic privilege. Under this framework, identity can be structured so that investing in, or associating with, a group identity can lead to economic returns and benefits. This treatment of identity as endogenous represents a major departure from more conventional economic models but is consistent with a set of alternative theories for explaining stubborn racial gaps in economic outcomes, and these theories help to operationalize stratification economics.4 Those most relevant to the labor market context are Lewis’s (1979) noncompeting groups hypothesis and Swinton’s (1978) labor force competition model of racial discrimination. In each of these, racial identity is associated with aspects of power and social control that are directly incorporated into the analysis.

Lewis (1979) and Swinton (1978) present models of a hierarchical wage or occupational structure and the existence of white worker “coalitions” that allow those who share this group identity to maintain a higher position in that hierarchy by limiting other (i.e., black) workers’ access to higher-status and higher-paying occupations and funneling them into lower-status and lower-wage jobs. The coalitions’ ability to exercise such an influence is based on their position as the majority group, which is a numerical and historical advantage. It is important to note that while applying racial preferences to an existing hierarchical occupational structure (i.e., occupational segregation) is discriminatory, it can be achieved without explicitly invoking or referencing race. Rather, a white worker “coalition” can essentially render excluded workers “noncompeting” by using its majority position in the firm or industry to influence the required credentials for a position, manipulate opportunities to obtain the credentials, or otherwise act as a gatekeeper over entry and promotion to preferred positions, as is often observed in professional environments, including corporate leadership, academia, law, or medicine.

Thus, Lewis concludes that more direct forms of in-market discrimination only become necessary as pre-market efforts to preserve the established racial hierarchy in the occupational structure become less effective. An interesting implication of this conclusion is that investments in human capital that make members of the excluded group more qualified for preferred positions can increase the likelihood that they will experience labor market discrimination. Darity, Dietrich, and Guilkey (1997) find that while black males were making dramatic strides in acquiring literacy between 1880 and 1910 in the United States, simultaneously they were suffering increasing proportionate losses in occupational status due to disadvantageous treatment of their measured characteristics. 

Krueger’s (1963) extension of the trade-based version of the Becker model also has relevance to the discussion of racially disparate labor market outcomes. In that model, white capitalists must value racial group solidarity sufficiently to accept a lower return on their capital as the price they pay for a generally higher level of income for all whites (and higher wages for white workers). In principle, if white capitalists lose from their inability to hire less-expensive black workers, a sufficiently high relative gain in income for white workers can compensate white capitalists for their losses. This prospect advanced by Anne Krueger nearly 60 years ago fits like a glove into stratification economics’ frame of understanding discrimination as an act that yields group benefits and losses.

Further, there is an additional perverse possibility derived from stratification economics. A racial hierarchy of workers can be exploited by owners of capital to subvert worker solidarity and capture a larger share of a worker’s productivity as economic profit. In short, employers can potentially get away with paying black and white workers a wage below their marginal productivity if, on average, the weight white workers place on being relatively better off than black workers is sufficiently high and white workers are paid a wage that is above that of black workers. Thus, both white labor and white capital jointly can benefit from discrimination against black workers. 

Conclusion 

When we look at race and labor market discrimination in the context of workers’ bargaining power, it is important we recognize there are at least two complementary goals. With respect to wages, we want to shift the balance of power in a way that puts upward pressure on wages—particularly for wage earners at or below the median—and at the same time close racial wage gaps. We cannot rely on competitive markets alone to do this. Rather, interventions are required to address these inequalities. Appropriate design of those interventions requires that we expand the frameworks we use for understanding power, race, gender, and inequality so that we restructure systems and institutions to prevent discriminatory outcomes rather than enable them. 

Policies geared toward maximizing employment and limiting the depth and duration of recessions are essential to establishing a new balance of power that makes workers less vulnerable to limited job prospects and low wages. For many of the reasons we have already discussed, these policies are particularly important to improve outcomes of black workers. During the last four decades, the Federal Reserve’s monetary policy decisions have been too contractionary, and those decisions have limited wage growth for the bottom 80% of workers and had an adverse effect on closing the black-white wage gap. While recent revisions to the Federal Reserve’s long-run goals and monetary policy strategy reflect some acknowledgement of the role the central bank plays in reducing or exacerbating racial economic inequality, unless there is an ongoing commitment to avoid prematurely enacting policies that needlessly limit job growth and disproportionately harm black workers, the balance of power is unchanged. Similarly, Congress must avoid excessive and unnecessary fiscal austerity and utilize its power to target funding for job creation in ways that promote racial equity, including a federal job guarantee. A federal job guarantee would eliminate the need for economists to squabble over the full employment unemployment rate, and essentially end the tradeoff between unemployment and inflation by making the Phillips curve vertical at a zero unemployment rate.

Labor unions play an important role in giving workers a stronger collective voice to advocate for higher pay, better benefits, training and promotional opportunities, and protections against discrimination and harassment. The Protecting the Right to Organize (PRO) Act is an important step toward streamlining the process when workers form a union, ensuring that they are successful in negotiating a first agreement, and holding employers accountable for violations of labor law (McNicholas, Poydock, and Rhinehart 2021). Historically, when given an opportunity to join a union, black workers have had the highest rates of union membership and have benefited from better pay and working conditions relative to workers who are not covered by a union contract. Still, the labor movement, like any other U.S. institution, is not immune to racism, and unions must continue to grow as more diverse, inclusive, and dynamic organizations as they serve the vital role of leveling the playing field for all workers. 

Finally, the fact that labor market discrimination has persisted well beyond the passage of Title VII of the Civil Rights Act of 1964 and the establishment of the Equal Employment Opportunity Commission, the federal agency tasked with enforcement of federal anti-discrimination laws, should not be overlooked, or taken lightly. As outlined in Yang and Liu (2021), meaningful accountability for discrimination requires solutions that confront the power and information asymmetries that weaken our enforcement system. Specifically, the authors recommend changes in at least four areas: (1) policies that encourage employer transparency, such as requiring employers to report employment and pay data by race, ethnicity, gender, and occupation, are necessary to fight discrimination and encourage accountability; regular reporting draws attention to discriminatory patterns, but also empowers workers with the information they need to pursue recourse against workplace discrimination; (2) increased funding of federal, state, and local enforcement agencies is necessary to provide the staffing and resources required to investigate the tens of thousands of discrimination charges filed each year and level the playing field for workers seeking justice; (3) revising legal doctrines to better align with the language and purpose of Title VII and other anti-discrimination laws will help to relieve the exceptionally onerous burden workers face in proving cases of discrimination; and (4) legal protections against anti-discrimination should be expanded to cover all workers and protect against practices that coerce employees to waive any rights to legally challenge unfair or unequal treatment.

The pursuit of economic and racial justice requires a serious interrogation of long-accepted assumptions about how the labor market functions and how much power any individual worker has to choose a better alternative. For black workers, there is also a long history of racially motivated exclusion, exploitation, and oppression that contributes to the assumed inferiority of black labor and the normalization of racial inequality. Therefore, meaningful policy changes that will serve to empower all workers and eliminate persistent racial disparities in the labor market also require a serious reckoning with the pervasiveness of racism in the collective thought, actions, institutions, and polices of the United States.

Notes

1. The nonwhite racial category included black workers along with others who did not identify as white, but about 95% of those in the category were black (or “negro”).

2. The 2019 high school and college completion data are based on non-Hispanic white population, while 1972 estimates for whites are not distinguishable by ethnicity.

3. Based on estimates of a wage regression with controls for education, age, state of residence, and union contract coverage status in addition to race and gender. The data used for this analysis included a combined 10 years of data from the CPS ORG (2009–2018).

4. Darity and Mason (1998) provide a more extensive review of economic models with relevance to the theory of stratification economics.

References 

Abayomi, Kobi, and Lee Hawkins. 2009. “Copulas for Tunable Markov Processes: Heuristics for Extreme-Valued Labor Market Outcomes, for African American Males: Fortune 500 CEOs vs. Professional Athletics.” Proceedings of the 2010 Joint Mathematics Meetings, Spring. 

Agan, Amanda, and Sonja Starr. 2018. “Ban the Box, Criminal Records, and Racial Discrimination.” Quarterly Journal of Economics 133, no. 1: 191–235. 

Arrow, Kenneth J. 1973. “The Theory of Discrimination.” In O. Ashenfelter and A. Rees, eds., Discrimination in Labor Markets. Princeton University Press. 

Azar, José, Ioana Marinescu, and Marshall Steinbaum. 2019. “Measuring Labor Market Power Two Ways.” AEA Papers and Proceedings 109.

Bassier, Ihsaan, Arindrajit Dube, and Suresh Naidu. 2020. “Monopsony in Movers: The Elasticity of Labor Supply to Firm Wage Policies.” National Bureau of Economic Research Working Paper no. 27755. https://doi.org/10.3386/w27755.

Becker, G. 1957. The Economics of Discrimination. University of Chicago Press. 

Becker, G.S. 1964. Human Capital: A Theoretical and Empirical Analysis with Special Reference to Education. 3rd Edition. University of Chicago Press. 

Bendick, Marc Jr., Charles W. Jackson, and Victor Reinoso. 1994. “Measuring Employment Discrimination Through Controlled Experiments.” In James B. Stewart, ed., African-Americans and Post-Industrial Labor Markets. Transaction Publishers. 

Berg, Ivar. 1970. Education and Jobs: The Great Training Robbery. Praeger. 

Betsey, Charles. 1994. “Litigation of Employment Discrimination Under Title VII: The Case of African American Women.” American Economic Review 84, no. 2: 98–102. 

Bivens, Josh, and Lawrence Mishel. 2013. “The Pay of Corporate Executives and Financial Professionals as Evidence of Rents in Top 1 Percent Incomes.” Journal of Economic Perspectives 27, no. 3: 57–78. 

Bivens, Josh, and Lawrence Mishel. 2015. Understanding the Historic Divergence Between Productivity and a Typical Worker’s Pay. Economic Policy Institute.

Bivens, Josh, and Ben Zipperer. 2018. The Importance of Locking in Full Employment for the Long Haul. Economic Policy Institute. 

Blau, Francine D., and Andrea Beller. 1992. “Black-White Earnings over the 1970s and 1980s: Gender Differences in Trends.” Review of Economics and Statistics 74, no. 2: 276–86. 

Bound, John, and Richard Freeman. 1989. “Black Economic Progress: Erosion of the Post-1965 Gains in the 1980s?” In Steven Shulman and William Darity, Jr., eds., Question of Discrimination: Racial Inequality in the U.S. Labor Market. Wesleyan University Press. 

Callaway, Brantly, and William J. Collins. 2017. “Unions, Workers, and Wages at the Peak of the American Labor Movement.” National Bureau of Economic Research Working Paper no. 23516.

Card, D., A. Cardoso, J. Heining, and P. Kline. 2018. “Firms and Labor Market Inequality: Evidence and Some Theory.” Journal of Labor Economics 36, no. S1: 13–70.

Card, David, and Alan Krueger. 1992. “School Quality and Black-White Relative Earnings: A Direct Assessment.” Quarterly Journal of Economics 107 (February): 151–200. 

Carlson, Leonard, and Caroline Swartz. 1988. “The Earnings of Women and Ethnic Minorities, 1959–1979.” Industrial & Labor Relations Review 41, no. 4: 530–46. 

Carnevale, Anthony P., Jeff Strohl, Neil Ridley, and Artem Gulish. 2018. Three Educational Pathways to Good Jobs: High School, Middle Skills and Bachelor’s Degree. Georgetown University Center on Education and the Workforce.

Cunningham, James S., and Nadja Zalokar. 1992. “The Economic Progress of Black Women, 1940–1980: Occupational Distribution and Relative Wages.” Industrial & Labor Relations Review 45, no. 3: 540–55. 

Darity, William Jr., Jason Dietrich, and David Guilkey. 1997. ‘‘Racial and Ethnic Inequality in the United States: A Secular Perspective.’’ American Economic Review (May): 301–5. 

Darity, William A. Jr., David Guilkey, and William Winfrey. 1996. “Explaining Differences in Economic Performance Among Racial and Ethnic Groups in the USA: The Data Examined.” American Journal of Economics and Sociology 55 (October): 411–26.

Darity, William A. Jr., Darrick Hamilton, Patrick L. Mason, Gregory N. Price, Alberto Dávila, Marie T. Mora, and Sue K. Stockly. 2017. “Stratification Economics: A General Theory of Intergroup Inequality.” In Andrea Flynn, Susan R. Holmberg, Dorian T. Warren, and Felicia J. Wong, eds., The Hidden Rules of Race: Barriers to an Inclusive Economy. Cambridge University Press. 

Darity, William A. Jr., and Patrick L. Mason. 1998. “Evidence on Discrimination in Employment: Codes of Color, Codes of Gender.” Journal of Economic Perspectives 12: 63–90. 

Darity, William. A. Jr., and Miles Underwood. 2021. Reconsidering the Relationship Between Higher Education, Earnings, and Productivity. Postsecondary Value Commission. 

Day, Meagan. 2020. “Unions Are Essential for Eliminating Racism.” Jacobin, July 7, 2020.

Donohue, John, and James Heckman. 1991. “Continuous vs. Episodic Change: The Impact of Civil Rights Policy on the Economic Status of Blacks.” Journal of Economic Literature 29 (December): 1603–43. 

Dube, Arindrajit, et al. 2020. “Monopsony in Online Labor Markets.” American Economic Review: Insights 2, no. 1: 33–46.

Dube, Arindrajit, Laura Giuliano, and Jonathan Leonard. 2019. “Fairness and Frictions: Impact of Unequal Raises on Quit Behavior.” American Economic Review 109, no. 2: 620–63.

Farber, Henry S., Daniel Herbst, Ilyana Kuziemko, and Suresh Naidu. 2021. “Unions and Inequality over the Twentieth Century: New Evidence from Survey Data.” National Bureau of Economic Research Working Paper no. 24587. 

Ferguson, Ronald. 1995. “Shifting Challenges: Fifty Years of Economic Change Toward Black-White Earnings Equality.” Daedalus 124: 37–76. 

Fix, Michael, George C. Galster, and Raymond J. Struyk. 1993. “An Overview of Auditing for Discrimination.” In Michael Fix and Raymond Struyk, eds., Clear and Convincing Evidence: Measurement of Discrimination in America. Urban Institute Press.

Fosu, Augustin K. 1992. “Occupational Mobility of Black Women, 1958–1981: The Impact of Post-1964 Antidiscrimination Measures.” Industrial & Labor Relations Review 45, no. 2: 281–94. 

Goldsmith, Arthur, Jonathan Veum, and William Darity Jr. 1997. “The Impact of Psychological and Human Capital on Wages.” Economic Inquiry 35 (October): 815–29. 

Gould, Elise. 2020. State of Working America Wages 2019: A Story of Slow, Uneven, and Unequal Wage Growth over the Last 40 Years. Economic Policy Institute. 

Gould, Elise, and Valerie Wilson. 2019. Wage Growth Is Weak for a Tight Labor Market—and the Pace of Wage Growth Is Uneven Across Race and Gender. Economic Policy Institute. 

Gould, Elise, and Valerie Wilson. 2020. Black Workers Face Two of the Most Lethal Preexisting Conditions for Coronavirus—Racism and Economic Inequality. Economic Policy Institute.

Hamilton, Darrick, Algernon Austin, and William Darity Jr. 2011. Whiter Jobs, Higher Wages: Occupational Segregation and the Lower Wages of Black Men. Economic Policy Institute.

Heckman, James J., and Brook Payner. 1989. “Determining the Impact of Federal Antidiscrimination Policy on the Economic Status of Blacks: A Study of South Carolina.” National Bureau of Economic Research Working Paper no. 2854. 

Krueger, Anne. 1963. “The Economics of Discrimination.” Journal of Political Economy 75 (October): 481–86.

Langella, Monica, and Alan Manning. 2020. “The Measure of Monopsony.” Marshall Lecture, Annual Congress of the European Economic Association.

Leonard, Jonathan S. 1990. “The Impact of Affirmative Action Regulation and Equal Employment Law on Black Employment.” Journal of Economic Perspectives 4, no. 4: 47–63. 

Lewis, W.A. 1979. “The Dual Economy Revisited.” The Manchester School 47: 211–29.

Mangino, William. 2010. “Race to College: The Reverse Gap.” Race and Social Problems 2: 164–78. 

Mangino, William. 2012. “Why Do Whites and the Rich Have Less Need for Education?” American Journal of Economics and Sociology 71, no. 3 (July): 562–602. 

Mason, Patrick L. 1997. “Race, Culture, and Skill: Interracial Wage Differences Among African Americans, Latinos, and Whites.” Review of Black Political Economy 25, no. 3: 5–39.

Mason, Patrick L. 1998. “Race, Cognitive Ability, and Wage Inequality.” Challenge 41. 

Maxwell, Nan. 1994. “The Effects on Black-White Wage Differences of Differences in the Quantity and Quality of Education.” Industrial & Labor Relations Review 47: 249–64. 

McNicholas, Celine, Margaret Poydock, and Lynn Rhinehart, 2021. How the PRO Act Restores Workers’ Rights to Unionize: A Chart of the Ways the PRO Act Fixes Major Problems in Current Labor Law. Economic Policy Institute. 

Mincer, J. 1958. “Investments in Human Capital and Personal Income Distribution.” Journal of Political Economy 66: 281–302. 

Mishel, Lawrence, and Jori Kandra. 2021. “Preliminary Data Show CEO Pay Jumped Nearly 16% in 2020, While Average Worker Compensation Rose 1.8%.” Working Economics Blog (Economic Policy Institute), May 27, 2021.

Neal, Derek A., and William R. Johnson. 1996. “The Role of Premarket Factors in Black-White Wage Differences.” Journal of Political Economy 104, no. 5: 869–95. 

Neal, Derek, and Armin Rick. 2014. “The Prison Boom and the Lack of Black Progress After Smith and Welch.” National Bureau of Economic Research Working Paper no. 20283. 

Neumark, David. 2012. “Detecting Discrimination in Audit and Correspondence Studies.” Journal of Human Resources 47 (Fall): 1128–57. 

O’Neill, June. 1990. “The Role of Human Capital in Earnings Differences Between Black and White Men.” Journal of Economic Perspectives 4: 25–45. 

Pager, Devah. 2003. “The Mark of a Criminal Record.” American Journal of Sociology 108 (March): 937–75. 

Phelps, Edmund. 1972. “The Statistical Theory of Racism and Sexism.” American Economic Review 62, no. 4: 659–61. 

Quillian, Lincoln, Devah Pager, Ole Hexel, and Arnfinn H. Midtbøen. 2017. “Meta-Analysis of Field Experiments Shows No Change in Racial Discrimination in Hiring over Time.” PNAS Early Edition.

Rho, Hye Jin, Haley Brown, and Shawn Fremstad. 2020. “A Basic Demographic Profile of Workers in Frontline Industries.” Center for Economic and Policy Research. 

Rodgers, William M. III, and William E. Spriggs. 1996. “What Does the AFQT Really Measure? Race, Wages, Schooling, and the AFQT Score.” Review of Black Political Economy 24 (Spring): 13–47. 

Rodgers, William M. III, and William E. Spriggs. 2002. “Accounting for the Racial Gap in AFQT Scores: Comment on Nan L. Maxwell, ‘The Effect on Black-White Wage Differences of Differences in the Quantity and Quality of Education.’” Industrial & Labor Relations Review 55 (April): 533–41. 

Schultz, T.W. 1961. “Investment in Human Capital.” American Economic Review 51: 1–17. 

Sokolova, Anna, and Todd Sorensen. 2020. “Monopsony in Labor Markets: A Meta-Analysis.” Washington Center for Equitable Growth. 

Spriggs, William. n.d. “Is Now a Teachable Moment for Economists? An Open Letter to Economists from Bill Spriggs.”

Swinton, David H. 1978. “A Labor Force Competition Model of Racial Discrimination in the Labor Market.” Review of Black Political Economy 9: 5–42.

Tomaskovic-Devy, Donald, Melvin Thomas, and Kecia Johnson. 2005. “Race and the Accumulation of Human Capital Across the Career: A Theoretical Model and Fixed-Effects Application.” American Journal of Sociology 111, no. 1: 58–89. 

Turner, Margery, Michael Fix, and Raymond Struyk. 1991. Opportunities Denied, Opportunities Diminished: Racial Discrimination in Hiring. Urban Institute Press. 

Webber, Douglas A. 2015. “Firm Market Power and the Earnings Distribution.” Labour Economics 35: 123. 

Webber, Douglas A. 2018. “Employment Adjustment over the Business Cycle: The Impact of Competition in the Labor Market.” IZA Discussion Paper no. 11887, Institute of Labor Economics (IZA).

Webber, Douglas A. 2020. “Labor Market Competition and Employment Adjustment over the Business Cycle.” Journal of Human Resources 56, no. 1. 

Wilson, Valerie R. 2015. The Impact of Full Employment on African American Employment and Wages. Economic Policy Institute.

Wilson, Valerie, and William M. Rodgers III. 2016. Black-White Wage Gaps Expand with Rising Wage Inequality. Economic Policy Institute. 

Yang, Jenny R., and Jane Liu. 2021. Strengthening Accountability for Discrimination: Confronting Fundamental Power Imbalances in the Employment Relationship. Economic Policy Institute. 

Zalokar, Nadka. 1990. The Economic Status of Black Women: An Exploratory Investigation. U.S. Commission on Civil Rights.

How does codetermination—entitling workers to participate in firm governance, either through membership on company boards or the formation of works councils—affect worker welfare and corporate decision-making?

In 2018, the Reward Work Act and the Accountable Capitalism Act, proposed by Democratic senators, included provisions that would require large companies to allocate 33–40% of the seats on their boards to worker-elected representatives. These proposals emulate the German model of “board-level codetermination,” which originated in the aftermath of World War II and has since spread to many European countries, including Austria, Denmark, Finland, Norway, and Sweden. In addition, the German model of “shop-floor codetermination” through elected works councils has received widespread attention in the past several years, in part due to the widely covered 2014 and 2019 unionization drives at Volkswagen’s Chattanooga, Tennessee, plant.

American corporate law has historically been hostile to such arrangements, which impinge on owners’ or managers’ exclusive discretion. And the academic literature has claimed that involving workers in firm governance impedes efficient decision-making, distorts incentives, and deters capital formation by allowing workers to capture the fruits of investment, ultimately stunting economic growth and leaving both employers and workers worse off. But alternative perspectives in the literature emphasize the potential benefits of codetermination for firms and workers through enhanced trust and information flows. And recent arguments stress that shared governance requirements can mitigate imbalances of power between employers and workers and thereby prevent exploitation.

This paper critically assesses these competing perspectives. We describe the background of existing codetermination laws and ask whether there are successful precedents for proposals to rectify workplace power imbalances through codetermination reforms. We then ask how contemporary codetermination institutions operate in practice. In which areas of decision-making does codetermination boost workers’ influence, and to what extent? How do worker representatives use their newfound authority? Are shared governance arrangements characterized by adversarial struggles between worker representatives and employers, or by cooperative relationships in which worker representatives and employers work together toward mutually agreeable goals? We draw on surveys, interviews, and case studies to answer these questions, and briefly survey the existing quantitative evidence on the economic impacts of codetermination.

We conclude that, historically, codetermination reforms have not been a key stand-alone vehicle for increasing worker power and have instead been intended to supplement core frameworks of union representation and centralized collective bargaining. Contemporary codetermination arrangements mostly function as amicable venues for workers and employers to share information and perspectives and for workers to shape decisions about immediate working conditions. For example, board-level codetermination creates two-way knowledge flows, giving employers a more intimate understanding of company operations and the desires of workers, and giving workers financial and strategic information that may inform collective bargaining strategies. However, the presence of worker representatives on company boards does not substantially shift high-level decision-making; workers usually occupy a minority of seats and therefore lack the ability to outvote shareholders, and often worker representatives defer to shareholder representatives in recognition of the fact that workers benefit when the company performs well. Shop-floor codetermination gives workers some control over decisions about hours and amenities, but (apart from, e.g., German works councils) little control over wage-setting or layoff decisions. One notable exception is that worker representatives’ influence may grow during economic downturns, when qualitative evidence suggests the representatives sometimes play an important role in negotiating wage or hour cuts that prevent layoffs.

Probably reflecting the limited authority conveyed by most existing codetermination arrangements, the quantitative evidence suggests that both board-level and shop-floor codetermination have mostly zero or slight positive impacts on worker and firm outcomes.

On the worker side, minority board-level representation does not affect wages, but it may slightly increase job security and subjective job satisfaction; on the firm side, it has zero or small positive effects on productivity, capital intensity, and profitability. Relatively weak forms of shop-floor codetermination have similarly slight effects on both worker and firm outcomes, while stronger shop-floor codetermination arrangements (which allocate broader and more substantive powers to worker representatives) may slightly boost wages, reduce within-firm earnings inequalities, and raise job security (possibly at the expense of nonincumbent workers). Strong forms of shop-floor codetermination do not appear to worsen firm performance, and may even increase productivity, but there is still a dearth of credible quasi-experimental evidence on the effects of these arrangements, so we are hesitant to make confident pronouncements.

Our overall conclusion is that most existing codetermination arrangements are relatively weak and have, at most, incremental positive effects. This conclusion leaves us unable to decisively confirm or reject the important claim, implicit in American corporate law, that employers must retain exclusive discretion over firm governance or else economic performance will suffer. On the one hand, the existing evidence shows it is possible to involve workers in workplace decision-making in ways that, if anything, weakly improve firm performance while also plausibly benefiting workers. However, the representation arrangements for which we possess the most credible evidence do not involve very substantial restrictions on employer discretion. Causal evidence on the economic performance effects of shared governance arrangements that more substantively limit employer discretion—such as powerful German works councils or parity codetermination in German iron, coal, and steel sector firms—remains scarce. In sum, codetermination laws may perform valuable functions even if they do not substantially affect the balance of power in workplaces.

I. Introduction

In the United States, shareholders and owners exercise exclusive discretion over the governance of private firms. This model of corporate governance aligns with an influential strain of thought, dating back to Friedman (1970), which asserts that shareholder control produces the right incentives for economic growth while not endangering the welfare of workers, who are protected by the forces of labor market competition. Alternative strains of thought claim that pervasive employer labor market power necessitates countervailing “worker power institutions,” and argue that recent ailments suffered by American workers can be traced back to the steady decline of pro-worker institutions over the past five decades (Stansbury and Summers 2020).

This latter perspective has motivated recent proposals to boost worker power by giving workers formal rights to participate in workplace governance. In 2018, the Reward Work Act and the Accountable Capitalism Act, proposed by Democratic senators, included provisions that would require large companies to allocate 33–40% of the seats on their boards to worker-elected representatives. These proposals emulate the German model of “board-level codetermination,” which originated in the aftermath of World War II and has since spread to many European countries, including Austria, Denmark, Finland, Norway, and Sweden. In addition, the German model of “shop-floor codetermination” through elected works councils has received widespread attention in the past several years, in part due to the widely covered 2014 and 2019 unionization drives at Volkswagen’s Chattanooga, Tennessee, plant (Liebman 2017; Silvia 2018, 2020).

Other papers in this volume examine the effects on labor market outcomes of specific restrictions on firm decision-making, e.g., in the areas of wage-setting (minimum wages) or the determination of other job characteristics (safety or flexibility regulations). This paper examines the impact of codetermination laws—broader interventions that restructure firms’ internal authority structures by integrating workers into decision-making.

American corporate law has historically been hostile to such arrangements, which impinge on owners’ or managers’ exclusive discretion. In 1981, a landmark U.S. Supreme Court ruling narrowed the scope of unions’ bargaining rights, citing “an employer’s need for unencumbered decision-making” (Harlin 1982). The Chamber of Commerce’s amicus curiae brief in the same case asserted that decisions about aspects of workplace governance apart from wages, hours, and working conditions “are uniquely the central burdens and prerogatives of management…. They are matters over which the collective bargaining process is unlikely to be useful, but likely to be obstructive or destructive.”1

These statements echo influential arguments in the academic literature, which claim that involving workers in firm governance impedes efficient decision-making, distorts incentives, and deters capital formation by allowing workers to capture the fruits of investment, ultimately stunting economic growth and leaving both employers and workers worse off (Jensen and Meckling 1979; Hansmann and Kraakman 2000). In contrast, alternative perspectives in the literature emphasize the potential benefits of codetermination for firms and workers through enhanced trust and information flows (Freeman and Lazear 1995). In addition, recent arguments stress that shared governance requirements can mitigate imbalances of power between employers and workers and thereby prevent exploitation of workers (Anderson 2017; Strine, Kovvali, and Williams 2021).

In this paper, we critically assess these competing perspectives on codetermination. We begin, in Section II, with a historical discussion: We describe the background of existing codetermination laws and ask whether there are successful precedents for proposals to rectify workplace power imbalances through codetermination reforms. Then, in Section III, we ask how contemporary codetermination institutions operate in practice. The aforementioned perspectives assert, respectively, that shared governance beneficially “boosts worker power” or that it harmfully “constrains employer discretion.” Both of these statements are vague and in need of substantial clarification. In which areas of decision-making does codetermination boost workers’ influence, and to what extent? How do worker representatives use their newfound authority? Are shared governance arrangements characterized by adversarial struggles between worker representatives and employers, or by cooperative relationships in which worker representatives and employers work together toward mutually agreeable goals? We draw on surveys, interviews, and case studies to answer these questions. Finally, in Section IV, we briefly survey the existing quantitative evidence on the economic impacts of codetermination, drawing heavily on a recent survey article by Jäger, Noy, and Schoefer (2021).

We conclude the following:

• Historically, codetermination reforms have not been a key stand-alone vehicle for increasing worker power and have instead been intended to supplement core frameworks of union representation and centralized collective bargaining.

• Contemporary codetermination arrangements mostly function as amicable venues for workers and employers to share information and perspectives and for workers to shape decisions about immediate working conditions.

• Board-level codetermination, for example, creates two-way knowledge flows, giving employers a more intimate understanding of company operations and the desires of workers, and giving workers financial and strategic information that may inform collective bargaining strategies. However, the presence of worker representatives on company boards does not substantially shift high-level decision-making; workers usually occupy a minority of seats and therefore lack the ability to outvote shareholders, and often worker representatives defer to shareholder representatives in recognition of the fact that workers benefit when the company performs well.

• Shop-floor codetermination gives workers some control over decisions about hours and amenities, but (apart from, e.g., German works councils) little control over wage-setting or layoff decisions. One notable exception is that worker representatives’ influence may grow during economic downturns, when qualitative evidence suggests the representatives sometimes play an important role in negotiating wage or hour cuts that prevent layoffs.

• Probably reflecting the limited authority conveyed by most existing codetermination arrangements, the quantitative evidence suggests that both board-level and shop-floor codetermination have mostly zero or slight positive impacts on worker and firm outcomes (Blandhol et al. 2020; Jäger, Schoefer, and Heining 2021; Harju, Jäger, and Schoefer 2021). On the worker side, minority board-level representation does not affect wages, but it may slightly increase job security and subjective job satisfaction; on the firm side, it has zero or small positive effects on productivity, capital intensity, and profitability.

• Relatively weak forms of shop-floor codetermination have similarly slight effects on both worker and firm outcomes, while stronger shop-floor codetermination arrangements (which allocate broader and more substantive powers to worker representatives) may slightly boost wages, reduce within-firm earnings inequalities, and raise job security (possibly at the expense of nonincumbent workers). Strong forms of shop-floor codetermination do not appear to worsen firm performance, and may even increase productivity, but there is still a dearth of credible quasi-experimental evidence on the effects of these arrangements, so we are hesitant to make confident pronouncements.

Our overall conclusion is that most existing codetermination arrangements are relatively weak and have, at most, incremental positive effects. This conclusion leaves us unable to decisively confirm or reject the important claim, implicit in American corporate law, that employers must retain exclusive discretion over firm governance or else economic performance will suffer. On the one hand, the existing evidence shows it is possible to involve workers in workplace decision-making in ways that, if anything, weakly improve firm performance while also plausibly benefiting workers. However, the representation arrangements for which we possess the most credible evidence do not involve very substantial restrictions on employer discretion. Causal evidence on the economic performance effects of shared governance arrangements that more substantively limit employer discretion—such as powerful German works councils or parity codetermination in German iron, coal, and steel sector firms—remains scarce. We close by noting that codetermination laws may perform valuable functions even if they do not substantially affect the balance of power in workplaces.

II. A brief history of codetermination

We begin by sketching the historical origins of modern codetermination laws, focusing on the countries with the strongest contemporary codetermination systems: Germany, Austria, the Netherlands, and the Nordic countries. We use the history of German codetermination as a case study, and then note parallels to the historical trajectories of codetermination in the others.

The purpose of this historical discussion is threefold. First, we illustrate that codetermination laws or agreements tended to arise because powerful national labor movements mobilized and overcame employer resistance to shared governance—they did not constitute a sudden empowerment of workers by government fiat. Second, we emphasize that codetermination laws and agreements were one specific byproduct of a wider campaign by unions and labor groups to shift toward an egalitarian relationship of social partnership between labor and capital. Other products of this movement include widespread union representation in workplaces and strong, centralized collective bargaining frameworks; codetermination reforms have often been intended to supplement or extend core frameworks of union representation. Third, we show that labor movements often fell short of securing codetermination arrangements that they believed would result in significant workplace power-sharing; they were instead forced to settle for arrangements that they considered weak or insufficiently radical.

A. Germany

In Germany, the world’s first national codetermination law was introduced in the aftermath of World War I. As McGaughey (2016) notes, German labor movements had been advocating for shared governance since the popular revolutions of 1848–1849, but before World War I these efforts had been successfully suppressed by the aristocracy and by major business owners. The political and economic devastation wrought by the war shattered existing power structures and dramatically worsened the bargaining position of major industrialists, putting labor movements on a stronger footing. In addition, as “workers councils” seized control of several cities in the months following the end of the war, the looming threat of widespread proletarian revolution put immense pressure on employers to placate workers (Beal 1955; Thelen 1991). The result was a series of collective agreements negotiated between employer associations and labor unions, beginning in November 1918 with the Stinnes-Legien Agreement (Winkler 1993). It consisted of a package of reforms, including the introduction of an eight-hour working day, official recognition of labor unions by employers, and the establishment of industry-level collective bargaining frameworks through which unions and employer associations would jointly negotiate standards for wages, hours, and working conditions (Beal 1955; Silvia 2013). In addition, the agreement permitted the creation of “works councils” (shop-floor codetermination institutions under union supervision) in firms with 50 or more employees. The rising German union movement viewed works councils as a promising avenue through which to extend and entrench its influence in workplaces (Addison 2009).

In 1919 and 1920, the political position of the German labor movement worsened as a successful revolution failed to materialize and moderate parties won a parliamentary majority in the Weimar Republic’s first elections (Beal 1955). Under pressure from labor activists and striking workers, the newly elected parliament passed a national codetermination law: the Works Council Act of 1920, which introduced mandatory establishment-level worker representation in firms with 20 or more employees. However, labor activists believed the law allocated far too little power to worker representatives, and 100,000 workers gathered in front of the Reichstag to protest the law’s introduction (Weipert 2012).

Over the next decade, German judges and employers further weakened the works councils established by the act (Thelen 1991; McGaughey 2016), thus eliminating any semblance of substantive codetermination. With the ascent of the Nazis in the early 1930s, the Works Council Act was dealt its final blow as labor groups were banned, union leaders were imprisoned or murdered, and major industrialists regained their former power.

Codetermination was reintroduced after World War II partially through the grassroots efforts of German workers and unions and partially via an external imposition by the British occupiers (Silvia 2013; Zahn 2015; McGaughey 2016). In the immediate aftermath of the war, German workers moved quickly to reestablish labor unions and works councils, taking advantage of the temporarily weak position of employers. Meanwhile, the British imposed geopolitically motivated labor reforms. Business leaders in the German heavy industries had played a crucial role in bankrolling the Nazis and supplying the machinery of both World Wars, and the Allies were determined to prevent a reoccurrence of the same dynamic. They therefore took steps to democratize the heavy industries and decentralize power away from major industrialists. In 1948, an Allied statute imposed “parity codetermination” on large firms in the iron, coal, and steel industries; under parity codetermination, workers elect representatives to 50% of the seats on a company’s board. The statute also formalized the role and rights of works councils, declaring them to be local support bodies for the industry-level labor unions and giving them a set of formal codetermination rights (that, although valued by union leaders, were considered quite weak; Silvia 2013). Finally, the Allies helped reintroduce the short-lived industry-level collective bargaining frameworks set up by the Stinnes-Legien Agreement.

German labor groups, which had initially taken advantage of the decimated postwar position of German employers to push for widespread nationalization, were impressed by parity codetermination. They dropped their demands for nationalization and began instead to push for an economywide adoption of parity codetermination (Scherrer 1983; Silvia 2013). They were only partially successful; in 1952, the German legislature introduced a law requiring only one-third board-level representation in large firms in industries other than iron, coal, and steel, due to strong resistance from resurgent employer associations to the idea of extending parity codetermination requirements. Failure to secure full parity codetermination was seen as a dispiriting defeat for labor groups, who did not view one-third representation as an authentic form of shared governance (Silvia 2013).

Labor groups were further dispirited by the passage, in the same year, of a new Works Council Act, which significantly weakened the works councils that had been established via the Allied statute, ad hoc arrangements, and state-level legislation (Thelen 1991; Silvia 2013). The act narrowed the mandate of works councils and formally separated them from labor unions, in an attempt to curtail the influence of unions. The rights of German works councils were later strengthened by reforms in 1972 and 2001 (Addison et al. 2004).

In the two decades following the 1952 board-level and shop-floor codetermination laws, the primary aim of West German labor movements was to secure the extension of parity codetermination to all large German firms (Silvia 2013; McGaughey 2016). With the decline of the center-right Christian Democrats and ascent of the left-wing Social Democratic Party in the late 1960s and early 1970s, German labor movements came close to achieving their goal. In 1976, a major codetermination reform initiated by the governing Social Democrats extended 50% board-level representation to all German firms with 2,000 or more employees. However, the reform included a crucial concession to the Social Democrats’ coalition partner, the classically liberal, business-friendly Free Democratic Party: Shareholders would be given a tie-breaking vote on company boards, meaning that workers could always be outvoted by unanimous shareholders and hence would enjoy only “quasi”-parity representation. Once again, labor movements were disappointed with this concession.

Since the 1970s, there have been several amendments to German codetermination laws (Page 2018). In 1994, a push to simplify corporate regulations culminated in the abolition of board-level codetermination requirements in small newly formed stock corporations (which had, uniquely, previously been subject to codetermination requirements regardless of their size; this reform is studied by Jäger, Schoefer, and Heining 2021). The reform was very narrow in scope, in part because the ambitions of the Christian Democrat/Free Democrat government were limited to begin with (they aimed only to harmonize regulations between stock corporations and limited liability companies and did not pursue a wider rollback of codetermination requirements), and in part because the government was forced to compromise with the Social Democratic upper house (Bundesrat). A few years later, in 2001, a Social Democratic government passed legislation broadening works councils’ coverage and slightly strengthening their codetermination rights, in response to a gradual decline in works council coverage and a report released by a codetermination review commission (Addison 2009).

Overall, since the reforms of the 1970s, codetermination laws have not been a major source of political conflict in Germany. Employer associations officially oppose codetermination requirements, but a comfortable majority of individual businesses and managers are supportive of them (Paster 2012). Paster suggests that these facts can be reconciled by noting that employer associations strategically overemphasize the views of vocal opponents of shared governance mandates, since opponents have much to gain from abolition but proponents have little to lose (as they would be able to voluntarily retain their codetermination arrangements).

To sum up: From the 1910s to the 1970s, German codetermination reforms were only one element of a wider competition between employers and labor groups, with the economic devastation of the World Wars and the external intervention of the British providing the substantial boost of worker power that enabled labor groups to secure major reforms. These reforms primarily involved the strengthening of unions and the establishment of industry-level collective bargaining frameworks; shop-floor codetermination was intended as a mechanism to supplement the operations of industry-level unions, while parity board-level codetermination was viewed as a stand-alone method of boosting worker power but was never extended beyond the iron, coal, and steel sectors. More broadly, many of the imposed codetermination arrangements, including minority board-level representation as well as the works councils originally established by both Works Council Acts, were perceived by labor groups to be weak and inauthentic forms of shared governance. Only parity board-level representation was considered a really substantive example of codetermination; interestingly, the top-down imposition of parity codetermination by the Allies constitutes perhaps the only historical example of a dramatic equalization of power through the fiat-based imposition of codetermination arrangements. But this occurred in a very unique historical context, and parity codetermination has not since been introduced in any other context.

B. Other countries

The histories of codetermination in Austria, the Netherlands, and the Nordic countries share the three key features we highlighted in the German context.

First, in each of these countries, the introduction of codetermination was enabled by preexisting factors that helped boost workers’ influence. In Austria, codetermination originated largely in parallel with Germany, with worker mobilization following World War I leading to the Austrian Works Councils Act of 1919, and post–World War II reforms reestablishing and extending the codetermination arrangements that arose in the interwar years (Kummer 1960).

Meanwhile, in the Netherlands and the Nordic countries, this boost to worker power came through the establishment of national frameworks for negotiation and collective agreement between powerful union associations and employer associations (e.g., via the Danish Constitution of the Labor Market negotiated in 1899, the first Dutch national agreement in 1914, the Norwegian Basic Agreement of 1935, and the Swedish Saltsjöbaden Agreement of 1938; see Wheeler 2002; Haug 2004a, 2004b; Trampusch 2006; Bergene and Hansen 2016). Under these frameworks, unions and employer associations met regularly to jointly determine national or industry-level standards for wages and working conditions. Labor movements secured the creation of these frameworks through massive and extended strikes and through the legislative efforts of social democratic parties, which were for a long time deeply intertwined with Nordic labor movements (Alestalo and Kuhnle 1986).

Firm-level codetermination arrangements in the Nordic countries were introduced in the decades following the creation of these frameworks, initially through collective negotiations and then through legislation (Bjorheim 1974; Knudsen 2006; Votinius 2012). Codetermination reforms consisted of the allocation of new co-decision-making rights to establishment-level union representatives, who were already present in most workplaces. National unions pursued codetermination rights for their representatives out of a desire to have a say on issues of workplace organization broader than the narrow set of decisions (about wages, benefits, etc.) covered by collective bargaining agreements (Wheeler 2002). Nordic codetermination representatives inherited much of their power from the broader social power of the national unions (Votinius 2012).

Second, in all of these countries, codetermination rights were secured as part of broader packages of reforms aimed at empowering workers. As we have mentioned, codetermination reforms in the Nordic countries simply extended the role of union representatives, whose near-universal presence in workplaces was a result of the organizing and legislative efforts of national unions. In addition, codetermination arrangements were introduced alongside shorter working weeks, systems of unemployment or sickness insurance, and other labor reforms (Van Leeuwen 1997; Haug 2004a, 2004b).

Third, political compromises meant that many of the codetermination reforms in these countries introduced weaker shared governance arrangements that left labor groups unsatisfied. For example, in Norway, the codetermination arrangements established in the 1960s and 1970s were later criticized by labor activists for conveying too little power to workers and focusing too narrowly on firm performance (Bergene and Hansen 2016). In Finland, political compromises in the drafting of a 1990 board-level codetermination law meant that the law applied to far fewer companies than preferred by the Social Democrats (Harju, Jäger, and Schoefer 2021). The 1950 Works Council Act in the Netherlands introduced mandatory works councils in firms with 50 or more employees; however, these councils had to include managers, they had only information and consultation rights without substantive codetermination powers, and their mandate was to improve firm performance rather than to advocate for workers. Thus, these councils were essentially toothless until a 1979 reform substantially strengthened them (Van het Kaar 1997). Similarly, a 1971 reform in the Netherlands gave works councils the right to nominate representatives to company boards, but these nominations could be rejected by the incumbent board. This was changed by a 2004 reform, but it remains the case that works councils cannot nominate candidates who are either an employee of the firm or a representative of a union engaged in a collective agreement with the firm (Van het Kaar 2007); consequently, Dutch worker-nominated board members are only “worker representatives” in a thinner sense.

C. Conclusion

To understand modern European codetermination, we must place it in its historical and institutional context. Firm-level codetermination requirements are only one (often relatively weak) component of a wider institutional environment shaped by European labor movements over the 20th century to advocate for workers. Our description of contemporary codetermination arrangements, in Section III, will frequently refer to interactions between codetermination and this wider institutional structure.

III. Does codetermination shift power in the workplace?

We now draw on detailed qualitative evidence to answer the following questions: In which domains, and to what extent, do existing codetermination arrangements shift power in the workplace? How do worker representatives deploy their powers, and how does shared governance play out in practice? What are the interactions between codetermination and other pro-labor institutions, including unions and collective bargaining frameworks? In Section IIIA we cover board-level codetermination, and in Section IIIB we discuss shop-floor codetermination.

A. Board-level codetermination

1. Existing board-level codetermination laws

Under board-level codetermination, workers elect representatives who fill a share of the seats on their company’s board. As in the United States, the boards of European companies are charged with making major strategic decisions and with appointing and supervising senior executives; board-level codetermination therefore gives workers the right to participate in a limited set of high-level decisions (Conchon 2011; Jäger, Schoefer, and Heining 2021).

As of 2022, a large number of European countries have board-level codetermination laws (Jäger, Noy, and Schoefer, 2021). Virtually all such laws give workers a minority of seats on their company’s board—usually 20% or 33%, ranging up to 50% minus the casting vote in the case of quasi-parity representation in Germany, as described above (ETUI 2020; Jäger, Schoefer, and Heining 2021). Under minority and quasi-parity representation, workers can always be overruled by shareholders voting unanimously, and consequently these codetermination arrangements give workers very little direct decision-making authority. The sole exceptions to this rule are firms in the German iron, coal, and steel sectors. Our discussion in this section focuses on minority and quasi-parity board-level codetermination arrangements, under which shareholders hold majority voting rights.

2. How does board-level codetermination operate in practice?

Board-level worker representatives are upfront about the fact that their minority status leaves them without formal decision-making power. For example, a French worker representative interviewed by Gold, Kluge, and Conchon (2010, 62) noted:

Our action as board-level employee representatives is very limited by the fact that our voting right is not powerful enough. So I know full well that I couldn’t…well, if you like, I’ve never managed to overturn a vote since I was first elected in 1999.

When asked to assess the impacts of board-level codetermination, one Finnish worker representative responded simply:

The employer always has a majority. No direct effect. (Harju, Jäger, and Schoefer 2021, 28)

Probably as a consequence of the fact that worker representatives have little hope of out-voting shareholders, formal voting does not figure prominently in the day-to-day operations of codetermined boards. Instead, board meetings are focused on cooperative dialogue and the mutual sharing of information and perspectives. Boards aim for consensus decisions, and split votes are unusual. (Of course, these collaborative, consensus-oriented discussions occur with the majority status of the shareholder representatives looming quietly in the background, which likely affects worker representatives’ behavior.) For example, worker representatives interviewed by Gold, Kluge, and Conchon (2010) said that:

Our whole modus vivendi on the supervisory board is oriented towards consensus. To that extent, the outcome of formal voting does not carry so much weight. In 11 years on the supervisory board I have never encountered the kind of fundamental conflict with shareholders or managers the question refers to. (A representative from the Czech Republic, p. 28)

…[W]e sit around the same table and we have the same powers and responsibilities, but of course I know where the power lies. Of course, if we come to a vote, then we lose—but the [shareholder representatives] always seek consensus….Very frequently, they ask us, they challenge us, and so they want our opinion. That must have some impact as well, or there’s not much point in asking. (A Finnish representative, pp. 35, 40)

I don’t feel in a minority or any kind of inferiority. Both sides try to achieve unanimity. (A German representative, p. 103)

Boards are able to arrive at consensus decisions in part because of the compliant attitudes of worker representatives. Either in recognition of their inability to overrule shareholders or due to a belief that the interests of workers are mostly aligned with the interests of the company, worker representatives often defer to directors and shareholder representatives, especially when boards make major strategic decisions with important profit implications. For example, Levinson (2000) reports that Swedish worker representatives are almost totally inactive during board-level discussions of company strategy; meanwhile, fewer than 5% of Finnish worker representatives report wielding influence over strategic decisions or decisions about production, outsourcing, or investment (Harju, Jäger, and Schoefer 2021). An Austrian worker representative argued that:

It’s my task to be there for the workforce….It’s the task of the [management] to run the company. I don’t interfere with that….Everyone is concerned with the long-term survival of the company. (Gold, Kluge, and Conchon 2010, 17)

Worker representatives’ reluctance to participate in strategic discussions may also be attributable to a perception that strategic decisions are made out of their view and discussed during board meetings only as a formality. A Finnish representative interviewed by Harju, Jäger, and Schoefer (2021) reported that:

…management has, in fact, already decided the course of action at the stage when I become aware of it. At that point, it is virtually impossible to influence the big lines anymore; maybe you can say your words and negotiate some details.2

If worker representatives are stranded in a perpetual minority and often defer to shareholder representatives, what is their purpose on the board? The qualitative evidence suggests that they serve three main functions. First, they share information with managers and communicate the perspectives and ideas of workers. As two Finnish board-level representatives said:

It often feels that the members of the management group want to talk to me because they feel that they are separated from the employees and want to hear my opinions.

….I can bring the personnel’s thoughts and ideas to the management team very freely. And bring different types of thinking from employees. (Harju, Jäger, and Schoefer 2021, 30–31)

A French board-level representative noted that:

[Shareholder representatives] do appreciate us because they live in their bubble, they’re in their stratosphere. Quite visibly, when I explain things to them that [workers] might find very basic, they’re often completely taken aback….While they regard everything as a cost item, I for my part try to show them that it doesn’t represent a cost when it enables the company to operate better, live better, and even to sustain. That can even serve the shareholders’ interests…. They do listen when I talk like that. (Gold, Kluge, and Conchon 2010, 62)

Possibly due to the benefits of increased access to information (paired with little relinquishment of formal influence), European directors and shareholders hold mostly positive views of board-level codetermination. Levinson (2000) observes that 61% of Swedish directors believe board-level codetermination has net positive effects on companies, citing increases in information-sharing and the legitimacy of decisions; meanwhile, 30% believe the institution has a neutral effect and only 9% believe it has a negative effect. In addition, 80% of Swedish directors report that the degree of cooperation between worker and shareholder representatives is “good” or “very good.” According to Paster (2012), 71% of German executives and 63% of German private investors oppose the repeal of board-level codetermination laws.

The second function of board-level worker representatives is to directly influence decisions about working conditions—an area of decision-making in which shareholders appear (somewhat) willing to allow workers to shape outcomes. Levinson (2000) reports that worker representatives in Sweden are highly active during board-level discussions of personnel or working conditions, and over 90% of Swedish directors claim that worker representatives have a “large” impact on decisions about working conditions. Anecdotally, European worker representatives describe using their platform to secure a variety of goods for their fellow workers, including subsidized commuter tickets, a budget for leisure activities, or expansions of pension eligibility (Gold, Kluge, and Conchon 2010; Harju, Jäger, and Schoefer 2021). Worker representatives also occasionally mention influencing decisions about layoffs, mergers, or wages, though these examples are the exception rather than the rule and there is widespread frustration at the difficulty of affecting important decisions and the unwillingness of employers to listen to worker representatives on these topics (Gold, Kluge, and Conchon 2010). For example, Finnish representatives interviewed by Harju, Jäger, and Schoefer (2021, n.p.) protested that:

We don’t get the opportunity to influence and provide help [in cases of personnel transfers and redundancies]. We can’t influence these matters.

Yes, I can freely participate in the discussion [about layoffs], but usually these issues are not discussed in the board meetings. The agenda is usually decided in advance, and then the board of directors simply goes through the agenda by stating facts rather than having discussions.…It is always the employer who makes the decision on [wage-setting].

Of course, even if worker representatives rarely exert direct influence over layoff decisions, worker representation may indirectly deter layoffs by raising the costs (consultation, negotiation, etc.) associated with firing workers (Keskinen 2017). However, this seems more likely to be true for shop-floor than board-level codetermination, as shop-floor representatives are usually given specific powers over decisions about layoffs or personnel transfers (ETUI 2020).

Third, board-level representatives sometimes use the information they acquire through board meetings to support the activities of shop-floor representatives or union representatives (notably, sometimes the same individual will be both a board-level representative and a shop-floor or union representative). For example:

I feel that I am well-informed about the economic background with regard to [my company]….Needless to say, that is a great help to me in our wage negotiations, in which I am the chief trade union negotiator [under an industry-level collective agreement]. (An Austrian representative in Gold, Kluge, and Conchon 2010, 20)

The benefit to the union of having one or more board-level employee representatives is to get information upfront and to display its stances at a high level, meaning that the union can anticipate events. (A French representative in Gold, Kluge, and Conchon 2010, 54)

My dual role as a [board-level and shop-floor] representative helps me to get more information, which is helpful when dealing with salary negotiations. (A Finnish representative interviewed by Harju, Jäger, and Schoefer 2021, n.p.)

One particularly notable and high-stakes example of cooperation between board-level, shop-floor, and union representatives is the negotiation of “employment pacts” that protect workers from layoffs during recessions in exchange for reductions in compensation. We describe this example in depth in Section IIIB.

However, we should take care not to overrate the importance of institutional interactions; many of the representatives surveyed by Gold, Kluge, and Conchon (2010) report having little to no contact with shop-floor or union representatives and having no input on collective bargaining strategies. One Norwegian representative even notes that all of the financial information that could usefully inform collective bargaining strategies is publicly available. Interactions between board-level representatives and other worker representation institutions are therefore far from a universal phenomenon.

3. Conclusion

As a consequence of workers’ minority vote share under existing laws, board-level codetermination does not allow workers to directly wield decision-making authority. Rather, existing board-level codetermination arrangements enhance information flows between managers and workers, allow workers to secure marginal improvements in working conditions, and may complement other worker representation institutions, including trade unions and shop-floor codetermination.

The available quantitative evidence suggests that these three mechanisms add up to produce neutral or slight positive impacts of board-level codetermination on worker and firm outcomes, as we describe in Section IV. Meanwhile, Finnish worker representatives surveyed by Harju, Jäger, and Schoefer (2021) do not perceive board-level codetermination as particularly impactful. Many believe the institution has no effects at all—citing the powerlessness inherent in a minority vote share or attempts by employers to bypass worker representatives by making decisions unofficially and out-of-view and treating board meetings as a formality. The Finnish representatives who do believe the institution has an impact mostly point to increases in “trust,” “transparency,” or “communication,” or “the staff feeling better taken into account.” They do not claim the institution affects wages, layoffs, or other economic outcomes.

B. Shop-floor codetermination

1. Existing shop-floor codetermination laws

Under shop-floor codetermination, workers elect shop-floor representatives or committees (e.g., “shop stewards” or “works councils”) who participate in day-to-day decisions about working conditions and dismissals. Most countries in Europe, and many countries outside of Europe, have laws that give workers rights to shop-floor codetermination (Jäger, Noy, and Schoefer 2021). The strength and breadth of authority conveyed to shop-floor representatives by these laws varies from country to country.

In the majority of countries with shop-floor codetermination laws, shop-floor representatives are merely given information and consultation rights, meaning that employers must inform shop-floor representatives in advance about planned layoffs or changes to working conditions and must consult them about the changes (ETUI 2020; Visser 2021). However, employers have no general obligation to take the perspective of shop-floor representatives into account, meaning that these laws convey no formal decision-making authority to workers.

In Austria, Germany, and the Nordic countries, shop-floor representatives are given more substantive formal authority, with the breadth of this authority varying across countries. In Austria, shop-floor representatives have co-decision-making rights in several areas, including disciplinary procedures, the allocation of working hours, workplace monitoring technologies, and performance pay systems (Aumayr et al. 2011; ETUI 2020). Austrian shop-floor representatives also have the right to demand external arbitration when employers make decisions with which they disagree in a broader set of categories (ETUI 2020). In Germany, shop-floor representatives have co-decision-making rights over a similar set of areas, can veto dismissals and force the employer to take the issue to a labor court, and (where industry-level collective bargaining agreements permit) can engage in local wage bargaining on behalf of workers. In Sweden and Norway, most changes to working conditions must be negotiated with establishment-level union representatives (ETUI 2020). In the Netherlands, major changes to workplace regulations must be approved by shop-floor representatives (ETUI 2020). Notably, the right to participate in decisions about working conditions or dismissals gives shop-floor representatives some indirect control over production decisions; proposals to alter production techniques by restructuring workflows or introducing automation technologies must pass through worker representatives charged with evaluating the impacts of these changes on workers. We return to this point in Section IV, when we discuss the relationship between codetermination and automation.

Overall, the majority of existing shop-floor representation laws convey very little formal authority to workers, but Austria, Germany, and the Nordic countries give shop-floor works councils or union representatives substantive powers over a variety of decisions relating to immediate working conditions and dismissals or transfers of staff. Shop-floor representatives, by contrast to board-level representatives, are directly granted decision-making authority and are highly involved in day-to-day firm governance; however, they have no direct mandate to deal with higher-level strategic decisions.

2. How does shop-floor codetermination operate in practice?

It is difficult to draw sweeping conclusions about how shop-floor codetermination operates in practice because of the considerable heterogeneity across countries in the responsibilities and rights assigned to shop-floor representatives. Since the available qualitative evidence largely consists of case studies or surveys of Nordic or German shop-floor codetermination, we focus on the activities of shop-floor representatives in countries that allocate them substantive decision-making powers. A few broad conclusions are evident.

First, shop-floor representatives are highly engaged in day-to-day discussions about working conditions, they manage to exert moderate influence over the outcomes of these discussions, and their contributions to these discussions are valued by employers. For example, Swedish shop-floor representatives interviewed by Wheeler (2002) describe influencing decisions about working hours and health and safety, helping set up education and training programs for workers, and helping resolve conflicts among workers or between workers and managers. Managers interviewed in the same study appreciatively cite the influence of the shop-floor representatives, saying that their input improves decision-making and increases worker satisfaction.

The interview evidence from Wheeler (2002) is consistent with broader survey evidence on the impacts of shop-floor representatives and their relationships with managers. In the 2019 European Company Survey, about 50% of managers across Europe claim that worker representatives have a “moderate” or “great” amount of influence on decisions about working conditions (Jäger, Noy, and Schoefer 2021).3 Levinson (2000) cites surveys showing that 80–90% of Swedish managing directors agree that shop-floor representatives exert “large” or “very large” influence over decisions about the workplace environment or working hours.

Meanwhile, managers have mostly positive views of the impacts of shop-floor codetermination on day-to-day decision-making. In the 2013 European Company Survey, about 80% of managers agree that worker representatives behave in a constructive and trustworthy way, that worker representation increases employee buy-in to decisions, and that worker representation “grants a competitive edge.” That said, the majority of managers indicate that they prefer to consult workers informally (authors’ own calculations), a point explored in-depth by Jäger, Noy, and Schoefer (2021). Levinson (2000) shows that 80–90% of Swedish managers approve of shop-floor representation, believing that the institution causes decisions to be “better rooted among employees.” Swedish managers also reject the claim that shop-floor representation impedes timely or effective decision-making or is a drain on resources—repudiating the U.S. Supreme Court’s assertion, quoted in the introduction, that involving workers in decisions that are “the central burdens and prerogatives of management” necessarily drags out or worsens decision-making.

The second general observation we can make is that, while shop-floor representatives are also highly engaged in discussions about layoffs, outsourcing, or personnel transfers, they wield less influence in this area than over decisions about working conditions. Representatives interviewed by Wheeler (2002) describe instances where they delayed layoffs or negotiated more generous severance packages, but report a general inability to prevent layoffs from happening. In the European Company Survey, only 25% of managers claim that employee representatives wield a “moderate” or “great” amount of influence over decisions about dismissals (Jäger, Noy, and Schoefer 2021).4

While shop-floor representatives appear unable to routinely influence dismissals, they may be more able to affect layoff decisions during economic crises. German works councils, for example, have a long history of negotiating “employment pacts” during recessions that ward off layoffs in exchange for cuts to wages or hours—effectively permitting firms to adjust employment on the intensive, rather than extensive, margin (Rehder 2003; Burda and Hunt 2011). This practice appears to be enabled by other kinds of worker representation as well. For example, Gregoric and Rapp (2019) show that Scandinavian firms with board-level codetermination were less likely to lay off workers during the Great Recession and more likely to cut wages or hours instead; Burdín and Dean (2009) show that Uruguayan worker-managed firms behave similarly.

Why might worker representation, specifically, enable this behavior? Traditional firms are reluctant to adjust wages or hours downward (Bewley 2002), perhaps because workers learn to reflexively resist proposed cuts to compensation out of fear that such cuts can be used to opportunistically exploit them. Worker involvement in decision-making might give workers access to the information they need to verify that cuts are genuinely necessary, or it may increase trust and enhance the legitimacy of decision-making enough to permit firms to propose wage or hour cuts. By thus enabling intensive-margin employment cuts, worker representation can benefit both workers and firms by insulating workers from unemployment and preserving productive worker-firm matches.

That said, the ability of European firms to avoid layoffs by cutting wages and hours during crises is also driven by other European labor market institutions, including short-time work policies, working time accounts, and clauses in industry-level collective bargaining agreements (Burda and Hunt 2011; Rinne and Zimmermann 2012; Herzog-Stein, Lindner, and Sturn 2018). We should not attribute this phenomenon entirely (or, perhaps, even predominantly) to codetermination.

The third general observation we can make is that shop-floor representatives, at least outside of Germany, do not exert much influence over wage-setting in their role as codetermination representatives. Here, it is crucial to draw clear distinctions between different European worker representation institutions, which often blend together. In countries with “single-channel” shop-floor representation, such as the Nordic countries, establishment-level union representatives function both as codetermination representatives (who have co-decision-making rights) and as union representatives (who have rights to engage in local wage negotiations and collective bargaining). Often, the distinction between “co-decision-making” and “negotiation” breaks down in practice, and shop-floor representatives simply engage in general advocacy on behalf of workers (Sippola 2012). Crucially, however, any authority that these shop-floor representatives have to influence wage-setting comes through their role as union representatives, not through their role as codetermination representatives. In cases where the two roles do come apart—for example, Finnish law allows for the election of codetermination representatives who are not union representatives—the codetermination representatives report wielding very little influence over wage-setting, and point to union representatives as the parties responsible for securing better wages (Harju, Jäger, and Schoefer 2021).

Meanwhile, countries like Germany have “dual-channel” shop-floor representation, meaning that shop-floor codetermination arrangements are clearly separate from shop-floor union representation (ETUI 2020). In Germany, shop-floor codetermination representatives (“works councils”) do sometimes engage in wage negotiations (as permitted by the works council law), and there is indeed some evidence that German works councils boost wages and create mild within-firm wage compression (Hirsch and Mueller 2020). That said, sectoral bargaining conducted by industry-level unions remains by far the most dominant form of collective negotiation in Germany.

Overall, under both single-channel and dual-channel regimes, it is collective bargaining and union-based negotiation that ultimately influence wage-setting; while shop-floor codetermination can affect decisions about, e.g., the adoption of performance pay schemes, the institution is not set up to influence overall wage levels.

3. Conclusion

Shop-floor representatives in the Nordic countries and Germany wield moderate authority in day-to-day firm governance, which they use to shape nonpecuniary aspects of working conditions. They are largely unable to influence routine decisions about layoffs or wage-setting, but they may have a greater capacity to affect these decisions during economic crises. Relationships between shop-floor representatives and employers are generally amicable, with both parties viewing shop-floor shared governance as mutually beneficial.

This qualitative evidence is once again consistent with quantitative evidence on the impacts of shop-floor representation, which suggests the institution has zero impacts on wages, may slightly reduce separations, and may improve subjective job quality (Addison 2009; Keskinen 2017; Harju, Jäger, and Schoefer 2021). We now turn to surveying the quantitative evidence on the impacts of codetermination.

IV. What are the economic impacts of codetermination?

Section III paints the following picture: Board-level and shop-floor codetermination arrangements affect some decisions about working conditions, result in increased information flows and increased worker trust in company management, and have little impact on major decisions—including decisions about wage-setting, layoffs, investment, and company strategy.

If we had to extrapolate from this qualitative characterization of codetermination to a prediction about the economic impacts of the institution, we would probably conjecture that codetermination has few impacts on observable economic outcomes and mildly improves nonpecuniary aspects of job quality. In particular, on the worker side, we would predict that codetermination does not affect wages and that it reduces turnover—either by directly insulating workers from layoffs during crises or by increasing job quality and hence reducing voluntary separations. On the firm side, we would predict null or small positive effects of codetermination on firm performance—not negative effects—for two reasons. First, the qualitative evidence is inconsistent with all of the channels through which negative effects of codetermination on firm performance are hypothesized to materialize. Codetermination does not give workers influence over wage-setting or decisions about investment or expansions, meaning that the “hold-up” and “worker rent-seeking” mechanisms postulated by Jensen and Meckling (1979) cannot get off the ground. Additionally, surveys of managers suggest that codetermination does not significantly slow down or obstruct decision-making. Second, the survey evidence described in Section III suggests that European managers, directors, and even investors have mostly positive views of codetermination, which would be hard to reconcile with the institution having substantial negative impacts on firm performance. Managers and directors even cite positive impacts of codetermination on decision-making, information flows, and trust, which might lead us to expect small positive impacts on firm performance.

We would also predict that shop-floor codetermination has greater impacts than board-level codetermination, at least in countries like Germany or Sweden that grant shop-floor representatives meaningful powers. Our qualitative discussion in Section IIIA showed that board-level representatives almost completely lack substantive decision-making authority, while in Section IIIB we noted that shop-floor representatives in Germany, Austria, and the Nordic countries exercise at least some authority over decisions about working conditions, may also influence layoff decisions, and (in Germany) can also be involved in wage negotiations.

Happily, these predictions are largely consistent with the available quantitative evidence on the economic impacts of codetermination, which we now briefly summarize.

A. Sources of evidence

The available evidence comes from two sources. First, a large set of studies, surveyed by Jäger, Noy, and Schoefer (2021), estimates the partial-equilibrium effects of codetermination on individual firms and their workers by comparing codetermined to noncodetermined firms. The central challenge faced by these studies is that firms with and without codetermination (e.g., firms above versus below the size thresholds for codetermination mandates, or firms whose workforces do versus do not take up their rights to codetermination) may differ unobservably in ways that make a simple comparison of their outcomes misleading. Several studies therefore use “quasi-experimental” techniques that attempt to isolate variation in codetermination status that is plausibly unrelated to these underlying unobservable factors.

For example, Jäger, Schoefer, and Heining (2021) and Harju, Jäger, and Schoefer (2021) analyze reforms in Germany and Finland that abolished or introduced codetermination requirements for specific subsets of firms delineated by size or legal form. They use micro datasets on individual firms and their workers to compare the outcomes of similar firms, affected versus unaffected by the reforms, before and after the reforms. By zooming in on these specific comparisons, these studies uncover the causal effects of the codetermination reforms under the reasonable assumption that affected and unaffected firms would have experienced identical trends in outcomes in the absence of the reforms.

Studies comparing the outcomes of codetermined and noncodetermined firms vary in the plausibility of the assumptions required to make their estimates have a causal interpretation. Consequently, Jäger, Noy, and Schoefer (2021) do not simply aggregate the findings of all existing studies, but instead weight existing results by the plausibility of the underlying “identifying assumptions.” Their survey, and our brief summary below, focuses mainly on the studies with the most persuasive quasi-experimental strategies.

The second source of evidence consists of cross-country event study estimates by Jäger, Noy, and Schoefer (2021), which try to uncover the general-equilibrium impacts of codetermination reforms on aggregate economic outcomes and the quality of industrial relations, using a similar difference-in-differences (comparing affected versus unaffected countries, before and after a reform) strategy at the country level. This strategy is less reliable at the level of countries than at the level of firms, since sample sizes are smaller and it is harder to find similar comparison countries that plausibly are on an otherwise similar trend, but these event study estimates nevertheless constitute the best available evidence on the general-equilibrium impacts of codetermination laws.

We now summarize the conclusions from these two sources of evidence.

B. Worker outcomes

First, both board-level and shop-floor codetermination have few (if any) impacts on observable worker outcomes, with shop-floor codetermination having slightly stronger effects. Board-level representation has zero or very small positive impacts on wage levels, with recent studies finding point estimates on the order of 1–2% and confidence intervals that include zero (Blandhol et al. 2020; Jäger, Schoefer, and Heining 2021; Harju, Jäger, and Schoefer 2021). Meanwhile, some evidence indicates that shop-floor representation in Germany moderately boosts wages and narrows within-firm earnings gaps, perhaps thanks to the special wage negotiation rights held by German works councils (Jirjahn and Smith 2018; Hirsch and Mueller 2020; Schnabel 2020).

Board-level representation also does not appear to reduce voluntary turnover, which constitutes revealed-preference evidence that this form of codetermination does not substantially improve job quality (Jäger, Schoefer, and Heining 2021; Harju, Jäger, and Schoefer 2021). However, strong forms of shop-floor codetermination are associated with lower voluntary turnover (Addison, Schnabel, and Wagner 2001). Both forms of codetermination do seem to reduce involuntary separations (i.e., layoffs), and may commensurately be accompanied by slight reductions in hiring in codetermined firms (Addison, Schnabel, and Wagner 2001; Keskinen 2017; Harju, Jäger, and Schoefer 2021). Finally, there is suggestive evidence that both kinds of codetermination improve subjective job quality (Harju, Jäger, and Schoefer 2021). Cross-country event studies confirm that codetermination reforms do not appear to affect wage levels, the labor share, or income inequality (Jäger, Noy, and Schoefer 2021).

C. Firm performance

Second, both types of codetermination have neutral or small positive impacts on firm performance, including productivity, capital intensity, revenue, and profitability (Jäger, Schoefer, and Heining 2021; Harju, Jäger, and Schoefer 2021). There is even some evidence that German works councils raise productivity (Mueller and Stegmaier 2017). This is consistent with the results of cross-country event studies, which find no effects of codetermination on productivity growth, capital formation, or growth in gross domestic product (Jäger, Noy, and Schoefer 2021).

D. Industrial relations

Finally, codetermination laws do not appear to improve the quality or cooperativeness of a country’s industrial relations, though the evidence here is murkier. Some scholars argue that codetermination institutions have been responsible for shaping cultures of cooperative industrial relations, e.g., in Germany (Thelen 1991). However, Jäger, Noy, and Schoefer (2021) find no evidence that codetermination reforms affect a country’s subsequent strike intensity and find no cross-sectional correlation between the “cooperativeness” of a country’s industrial relations and whether the country has codetermination laws; that said, they do find suggestive evidence for increases in union density as a result of codetermination reforms. In addition, they argue that the qualitative historical evidence suggests that codetermination arose in countries with preexisting cultures of social partnership and worker-management dialogue, rather than causing the development of such cultures.

Overall, the empirical evidence is not yet conclusive on this front.

E. Institutional complementarities

Jäger, Noy, and Schoefer (2021) also discuss whether evidence on the economic impacts of codetermination in Europe can be translated to the United States, given major differences in institutional context between the U.S. and Europe (for example, unions are much weaker and collective bargaining coverage is much lower in the U.S.). The qualitative evidence we have surveyed contributes to this discussion by highlighting historical and contemporary complementarities between codetermination and other worker representation institutions—including the fact that codetermination has historically played a subsidiary role to industry- or national-level trade unions and collective bargaining frameworks, and that many contemporary worker representatives describe their ability to supply unions with information as one of their main functions. These important complementarities mean we should be cautious about extrapolating from the European evidence to conclusions about the effects codetermination would have in the United States.

F. The future of work

Technological change and automation will continue to reshape labor markets over the next several decades, with the effects of these transformations on worker welfare likely to be mixed (Acemoglu 2021). Some authors hypothesize that worker involvement in production/job design through codetermination will help ensure that automation is handled in ways appropriately sensitive to the welfare of workers (Autor, Mindell, and Reynolds 2019; McKay, Pollack, and Fitzpayne 2019). There is important historical precedent for this view: In the 1970s, Swedish unions’ campaign for broad workplace codetermination rights was motivated by a perception that new technologies were being implemented in ways that eroded job quality and that workers needed a voice in production and workplace design in order to correct this trend (Sandberg et al. 1992).

On the one hand, by boosting wages or increasing firing costs, codetermination could raise the price of workers relative to robots and thereby unintentionally accelerate the substitution of humans with machines. Indeed, there is evidence that wage increases among low-paid workers spur increased automation of routine tasks (Dechezleprêtre et al. 2021), and union density is cross-sectionally correlated with higher automation (Acemoglu and Restrepo 2022). We have noted that codetermination does not appear to substantially boost wages, but it does reduce involuntary separations—plausibly in part by increasing firing costs—which leaves room for this mechanism to operate (although higher firing costs may also lead to a shift from labor-saving process innovations to product innovations; Manera and Uccioli 2021). On the other hand, codetermination might do exactly what Swedish unions hoped it would: give workers the power to persuade employers to build up their human capital and assign them to more sophisticated tasks that complement advanced technologies rather than keeping them on routine manual tasks and replacing them when those manual tasks are automated.

Unfortunately, there is very little concrete evidence to help us adjudicate between these potential channels. A pair of studies using data from the European Company Survey provides suggestive support for the latter hypothesis. The first study shows that the presence of employee representation in an establishment is correlated with the adoption of more sophisticated and automation-resistant job designs (Belloc et al. 2021). The second shows that codetermination is positively associated with take-up of advanced technologies, arguably because those technologies are complementary to workers’ newly acquired sophisticated skills (Belloc, Burdín, and Landini 2022).

In a similar vein, evidence from the German manufacturing sector, where works councils and board-level representatives are very influential, suggests that automation does not threaten the job security of incumbent workers and instead causes them to transition to higher-skilled and higher-paying tasks within their firms (though automation does negatively affect the outcomes of young workers and labor market entrants; see Dauth et al. 2021).5 These findings stand in contrast to other studies with similar methodologies showing substantial negative effects of automation on job security and employment in the United States (Acemoglu and Restrepo 2020), perhaps owing to the weakness of employee representation in the U.S. relative to Germany and a corresponding failure of the upskilling mechanism to materialize there.

Overall, the (small amount of) existing evidence suggests that worker representation does not obstruct automation or the adoption of advanced technologies, but it does encourage automating firms to upskill rather than replace their incumbent workers. (Importantly, the fact that worker representation protects incumbent workers in these cases does not necessarily mean it improves overall welfare.)

G. A puzzle

Before concluding our discussion, we turn to an interesting puzzle raised by the evidence we have surveyed so far. If codetermination has weakly positive impacts on both workers and firms, and if directors and managers mostly approve of the institution, why are codetermination laws necessary? Why don’t American firms voluntarily adopt shared governance arrangements?

First and most obviously, the National Labor Relations Act imposes significant legal barriers to voluntary codetermination arrangements in the United States (Liebman 2017). Even when a jurisdiction’s corporate law does not explicitly prohibit codetermination, by enshrining owner/shareholder control as the default form of firm governance, established legal frameworks can make experimentation with alternative governance systems difficult and risky (Anderson 2017).

Second, information asymmetries or “prisoner’s dilemma” dynamics may block unilateral voluntary adoption of worker participation even if the institution is socially beneficial (Levine and Tyson 1990). For instance, firms that voluntarily adopt shared governance may thereby signal to the stock market that workers have gained the upper hand in their internal labor relations, resulting in a stock price decline (Hayden and Bodie 2021); alternatively, mild wage compression induced by codetermination may cause talented workers to leave codetermined firms for noncodetermined ones (Burdín 2016).

These explanations are consistent with the observation that in Europe, where worker participation in firm governance is normalized by formal codetermination laws, we do observe widespread voluntary adoption of worker participation in firms without formal codetermination. Jäger, Noy, and Schoefer (2021) discuss evidence from the European Company Survey showing that firms without formal worker representation still frequently involve their workers in decision-making, and they claim that informal worker involvement has a considerable impact on the outcomes of decision-making.

H. Conclusion

Quantitative studies of the economic impacts of codetermination produce results consistent with our qualitative characterization of the institution. Board-level and shop-floor codetermination do not significantly shift core economic outcomes; both forms of codetermination may cause slight increases in job quality and job security, and strong forms of shop-floor codetermination may also slightly boost wages and productivity. The weakly positive impacts of codetermination on firm performance are arguably consistent with the absence of voluntarily adopted codetermination arrangements in many contexts. Finally, potential interactions between codetermination and other European labor market institutions, or between codetermination and emerging technologies, place important limits on our ability to extrapolate from the existing evidence.

V. Overall conclusion

According to the evidence, existing codetermination arrangements are mild, mostly benign institutions with nonexistent or small positive economic impacts. European-style codetermination institutions, especially minority board-level representation, convey very little authority to workers, and are hence unlikely to significantly shift power from employers to workers. It remains possible that stronger codetermination arrangements, such as German-style works councils, parity board-level codetermination, or a bicameral governance system where decisions require joint approval by shareholder- and worker-elected bodies, provide a larger boost to worker power (as speculated by a group of academics in The Guardian; see Fraser et al. 2020). However, empirical evidence on the economic impacts of such strong shared governance institutions is scant (Jäger, Noy, and Schoefer 2021).

Although existing codetermination arrangements do not significantly shift power within workplaces, they do appear to increase information-sharing and worker-management cooperation, which may explain the evidence for small positive impacts of these institutions on worker welfare and firm performance. Notably, the purpose of codetermination laws need not be to dramatically restructure workplace hierarchies or influence major economic decisions. As Jäger, Noy, and Schoefer (2021) note, proponents of codetermination often emphasize the intrinsic importance of fostering open and democratic workplaces by giving workers formal channels to express a voice in firm governance. Codetermination laws may successfully deliver on this front even if they do not have large effects on measurable economic outcomes.

Acknowledgments

We thank the editor, Lawrence Mishel, and two reviewers, Matthew Bodie and Gabriel Burdín.

About the authors

Simon Jäger (sjaeger@mit.edu) is an assistant professor of economics at MIT, Shakked Noy (snoy@mit.edu) is a predoctoral research fellow at MIT, and Benjamin Schoefer (schoefer@berkeley.edu) is an assistant professor of economics at UC Berkeley.

Notes

1. Chamber of Commerce brief in First National Maintenance Corp. v. NLRB: 452 US 666 (1981).

2. Some of our quotes from Finnish worker representatives in this section, including this one, are drawn from interviews or surveys conducted as part of the research for Harju, Jäger, and Schoefer (2021) that were not specifically quoted in the final version of the paper.

3. This statistic encompasses all forms of worker representation (including board-level representation), but shop-floor representation is much more widespread among surveyed firms than board-level representation.

4. A notable exception may be Germany, where works councils have the authority to veto all “unwarranted” dismissals and force their employer to take the issue to an employment court. However, even in Germany, only 28% of managers in the European Company Survey say that worker representatives wield “moderate” or “great” influence over dismissal decisions.

5. Similarly, Hirvonen, Stenhammer, and Tuhkuri (2021) provide quasi-experimental evidence that firms’ investment in new technologies increases firms’ employment in Finland (see Harju, Jäger, and Schoefer 2021 for an overview of Finnish codetermination institutions).

References

Acemoglu, Daron. 2021. “Harms of AI.” National Bureau of Economic Research.

Acemoglu, Daron, and Pascual Restrepo. 2020. “Robots and Jobs: Evidence from U.S. Labor Markets.” 128 Journal of Political Economy 2188.

Acemoglu, Daron, and Pascual Restrepo. 2022. “Demographics and Automation.” 89 Review of Economic Studies 89 1.

Addison, John. 2009. The Economics of Codetermination: Lessons from the German Experience. Palgrave Macmillan.

Addison, John, Lutz Bellmann, Claus Schnabel, and Joachim Wagner. 2004. “The Reform of the German Works Constitution Act: A Critical Assessment.” 43 Industrial Relations 392.

Addison, John, Claus Schnabel, and Joachim Wagner. 2001. “Works Councils in Germany: Their Effects on Establishment Performance.” 53 Oxford Economics Papers 659.

Alestalo, Matti, and Stein Kuhnle. 1986. “The Scandinavian Route: Economic, Social, and Political Developments in Denmark, Finland, Norway, and Sweden.” 16 International Journal of Sociology 3.

Anderson, Elizabeth. 2017. Private Government: How Employers Rule Our Lives (and Why We Don’t Talk about It). Princeton University Press.

Aumayr, Christine, Demetriades Stavroula, David Foden, Vera Scepanovics, and Felix Wolf. 2011. “Employee Representation at Establishment Level in Europe.” Eurofound Report.

Autor, David, David Mindell, and Elisabeth Reynolds. 2019. The Work of the Future: Shaping Technology and Institutions. MIT Work of the Future.

Beal, Edwin F. 1955. “Origins of Codetermination.” 8 Industrial and Labor Relations Review 483.

Belloc, Filippo, Gabriel Burdín, Luca Cattani, William Ellis, and Fabio Landini. 2021. Coevolution of Job Automation Risk and Workplace Governance. Institute of Labor Economics.

Belloc, Filippo, Gabriel Burdín, and Fabio Landini. 2022. “Robots, Digitalization, and Worker Voice.” Working Paper. https://dx.doi.org/10.2139/ssrn.4009136.

Bergene, Ann Cecilie, and Per Bonde Hansen. 2016. “A Historical Legacy Untouched by Time and Space? The Hollowing-Out of the Norwegian Model of Industrial Relations.” 6 Nordic Journal of Working Life Studies 5.

Bewley, Truman. 2002. Why Wages Don’t Fall During a Recession. Harvard University Press.

Bjorheim, Lars. 1974. “The Development of Industrial Democracy in Norway.” Paper presented to the International Conference on Manpower Training and Development, Oslo, Norway, August 1974.

Blandhol, Christine, Magne Mogstad, Peter Nilsson, and Ola Vestad. 2020. “Do Employees Benefit from Worker Representation on Corporate Boards?” National Bureau of Economic Research.

Burda, Michael, and Jennifer Hunt. 2011. “What Explains the German Labor Market Miracle in the Great Recession?” National Bureau of Economic Research.

Burdín, Gabriel. 2016. “Equality Under Threat by the Talented: Evidence from Worker-Managed Firms.” 126 Economic Journal 1372.

Burdín, Gabriel, and Andrés Dean. 2009. “New Evidence on Wages and Employment in Worker Cooperatives Compared with Capitalist Firms.” 37 Journal of Comparative Economics 517.

Conchon, Aline. 2011. Board-Level Employee Representation Rights in Europe: Facts and Trends. European Trade Union Institute.

Dauth, Wolfgang, Sebastian Findeisen, Jens Südekum, and Nicole Woessner. 2021. “The Adjustment of Labor Markets to Robots.” 19 Journal of the European Economic Association 3104.

Dechezleprêtre, Antoine, David Hémous, Morten Olsen, and Carlo Zanella. 2021. “Induced Automation: Evidence from Firm-Level Panel Data.” Working Paper. https://dx.doi.org/10.2139/ssrn.3835089.

European Trade Union Institute (ETUI). 2020. “National Industrial Relations.” https://www.worker-participation.eu/National-Industrial-Relations/Countries.

Fraser, Nancy, Chantal Mouffe, Saskia Sassen, Jan-Werner Müller, Dani Rodrik, Thomas Piketty, Gabriel Zucman, Ha-Joon Chang, et al. 2020. “Humans Are Not Resources: Coronavirus Shows Why We Must Democratise Work.” Guardian, May 15, 2020.

Freeman, Richard, and Edward Lazear. 1995. “An Economic Analysis of Works Councils.” In Works Councils: Consultation, Representation, Cooperation in Industrial Relations, edited by Joel Rogers and Wolfgang Streeck. National Bureau of Economic Research Comparative Labor Markets Series.

Friedman, Milton. 1970. “The Social Responsibility of Business Is to Increase its Profits.” New York Times Magazine, September 13, 1970.

Gold, Michael, Norbert Kluge, and Aline Conchon. 2010. ‘In the Union and on the Board’: Experiences of Board-Level Employee Representatives across Europe. European Trade Union Institute.

Gregoric, Aleksandra, and Marc Steffen Rapp. 2019. “Board-Level Employee Representation (BLER) and Firms’ Responses to Crisis.” 58 Industrial Relations 376.

Hansmann, Henry, and Reinier Kraakman. 2000. “The End of History for Corporate Law.” 89 Georgetown Law Journal 439.

Harju, Jarkko, Simon Jäger, and Benjamin Schoefer. 2021. Voice at Work. National Bureau of Economic Research.

Harlin, Julia. 1982. “First National Maintenance Corp. v. NLRB: The Supreme Court Narrows Employers’ Section 8(A)(5) Duty to Bargain.” 39 Washington and Lee Law Review 285.

Haug, Ralph. 2004a. “The History of Industrial Democracy in Sweden: Industrial Revolution to 1980.” 21 International Journal of Management 7.

Haug, Ralph. 2004b. “Industrial Revolution to 1980: The History of Industrial Democracy in Denmark and Norway.” 21 International Journal of Management 135.

Hayden, Grant M., and Matthew T. Bodie. 2021. “Codetermination in Theory and Practice.” 73 Florida Law Review 1.

Herzog-Stein, Alexander, Fabian Lindner, and Simon Sturn. 2018. “The German Employment Miracle in the Great Recession: The Significance and Institutional Foundations of Temporary Working-Time Reductions.” 70 Oxford Economic Papers 206.

Hirsch, Boris, and Steffen Mueller. 2020. “Firm Wage Premia, Industrial Relations, and Wage Sharing in Germany.” 73 Industrial and Labor Relations Review 1119.

Hirvonen, Johannes, Aapo Stenhammer, and Joonas Tuhkuri. 2021. “New Evidence on the Effect of Technology on Employment and Skill Demand.” Working Paper. https://economics.mit.edu/files/22239.

Jäger, Simon, Shakked Noy, and Benjamin Schoefer. 2021. “What Does Codetermination Do?” ILR Review, n.p. https://doi.org/10.1177/00197939211065727.

Jäger, Simon, Benjamin Schoefer, and Jörg Heining. 2021. “Labor in the Boardroom.” 136 Quarterly Journal of Economics 669.

Jensen, Michael, and William Meckling. 1979. “Rights and Production Functions: An Application to Labor-Managed Firms and Codetermination.” 52 Journal of Business 469.

Jirjahn, Uwe, and Stephen Smith. 2018. “Nonunion Employee Representation: Theory and the German Experience with Mandated Works Councils.” 89 Annals of Public and Cooperative Economics 201.

Keskinen, Maija. 2017. “Workplace Cooperation and Firm Performance: Evidence from Finland.” Master’s thesis, Aalto University.

Knudsen, Herman. 2006. “Workplace Representation in Denmark: Structure and Role.” Paper presented to the HIVA-EZA Conference, “A Boost for Workplace Democracy in Europe?” Leuven, Belgium, November 9–10, 2006.

Kummer, Karl. 1960. “Works Councils in Austria.” 81 International Labour Review 110.

Levine, David, and Laura D’Andrea Tyson. 1990. “Participation, Productivity, and the Firm’s Environment.” In Paying for Productivity: A Look at the Evidence, edited by Alan S. Blinder. Brookings Institution.

Levinson, Klas. 2000. “Codetermination in Sweden: Myth and Reality.” 21 Economic and Industrial Democracy 457.

Liebman, Wilma. 2017. “Does Federal Labor Law Preemption Doctrine Allow Experiments with Social Dialogue?” 12 Harvard Law and Policy Review, n.p.

Manera, Andrea, and Martina Uccioli. 2021. “Employment Protection and the Direction of Technology Adoption.” Working Paper. https://dx.doi.org/10.2139/ssrn.3754464.

McGaughey, Ewan. 2016. “The Codetermination Bargains: The History of German Corporate and Labor Law.” 23 Columbia Journal of European Law 1.

McKay, Conor, Ethan Pollack, and Alastair Fitzpayne. 2019. Automation and a Changing Economy Part II: Policies for Prosperity. Aspen Institute Future of Work Initiative.

Mueller, Steffen, and Jens Stegmaier. 2017. “The Dynamic Effects of Works Councils on Labour Productivity: First Evidence from Panel Data.” 55 British Journal of Industrial Relations 372.

Page, Rebecca. 2018. Codetermination in Germany: A Beginner’s Guide. Arbeitspapier, Hans-Böckler-Stiftung.

Paster, Thomas. 2012. “Do German Employers Support Board-Level Codetermination? The Paradox of Individual Support and Collective Opposition.” 10 Socio-Economic Review 471.

Rehder, Britta. 2003. Betriebliche Bündnisse für Arbeit in Deutschland: Mitbestimmung und Flächentarif im Wandel. Campus Verlag.

Rinne, Ulf, and Klaus Zimmermann. 2012. “Another Economic Miracle? The German Labor Market and the Great Recession.” 1 IZA Journal of Labor Policy, n.p. https://doi.org/10.1186/2193-9004-1-3.

Sandberg, Ake, Gunnar Broms, Arne Grip, Lars Sundström, Jesper Steen, and Peter Ullmark. 1992. Technological Change and Codetermination in Sweden. Temple University Press.

Scherrer, Christoph. 1983. “Co-Determination: The German Workplace Experience.” 2 Against the Current 47. https://www.marxists.org/history/etol/newspape/atc/v02n02.pdf.

Schnabel, Claus. 2020. “Betriebliche Mitbestimmung in Deutschland: Verbreitung, Auswirkungen und Implikationen (Workplace Participation in Germany: Distribution, Effects, and Implications).” 21 Perspektiven der Wirtschaftspolitik, n.p.

Silvia, Stephen. 2013. Holding the Shop Together: German Industrial Relations in the Postwar Era. Cornell University Press.

Silvia, Stephen. 2018. “The United Auto Workers’ Attempts to Unionize Volkswagen Chattanooga.” 71 ILR Review 600.

Silvia, Stephen. 2020. “The UAW Attempts to Organize Volkswagen Chattanooga: The Sequel.” Paper presented at the 72nd Annual Meeting of the Labor Employment Research Association, June 12–16, 2020.

Sippola, Markku. 2012. “Local Bargaining and Codetermination: Finnish Experience in Comparative Perspective.” 18 European Journal of Industrial Relations 53.

Stansbury, Anna, and Lawrence Summers. 2020. The Declining Worker Power Hypothesis: An Explanation for the Recent Evolution of the American Economy. National Bureau of Economic Research.

Strine, Leo, Jr., Aneil Kovvali, and Oluwatomi Williams. 2021. Lifting Labor’s Voice: A Principled Path toward Greater Worker Voice and Power within American Corporate Governance. University of Pennsylvania, Institute for Law and Economics.

Thelen, Kathleen Ann. 1991. Union of Parts: Labor Politics in Postwar Germany. Cornell University Press.

Trampusch, Christine. 2006. “Industrial Relations and Welfare States: The Different Dynamics of Retrenchment in Germany and the Netherlands.” 16 Journal of European Social Policy 121.

Van het Kaar, Robbert. 1997. Amendment of the Dutch Works Councils Act: A Few Surprises. European Foundation for the Improvement of Living and Working Conditions. https://www.eurofound.europa.eu/publications/article/1997/ amendment-of-the-dutch-works-councils-act-a-few-surprises.

Van het Kaar, Robbert. 2007. “Corporate Governance and Employee Board-Level Representation in the Netherlands.” In The Forgotten Resource: Corporate Governance and Employee Board-Level Representation. The Situation in France, the Netherlands, Sweden, and the UK. Hans Böckler Stiftung.

Van Leeuwen, Marco. 1997. “Trade Unions and the Provision of Welfare in the Netherlands, 1910–1960.” 50 Economic History Review 764.

Visser, Jelle. 2021. “OECD/AIAS ICTWSS Database: Detailed Note on Definitions, Measurement, and Sources.” Organization for Economic Cooperation and Development.

Votinius, Jenny. 2012. “Employee Representation at the Enterprise: Sweden.” 81 Bulletin of Comparative Labour Relations 51.

Weipert, Axel. 2012. “Vor den Toren der Macht. Die Demonstration am 13. Januar 1920 vor dem Reichstag (At the Gates of Power: The Demonstration on January 13, 1920, in Front of the Reichstag).” In JahrBuch für Forschungen zur Geschichte der Arbeiterbewegung (Yearbook for Research on History of the Labor Movement). Friedrich-Ebert-Stiftung.

Wheeler, Jeff. 2002. “Employee Involvement in Action: Reviewing Swedish Codetermination.” 26 Labor Studies Journal 71.

Winkler, Henrich August. 1993. Weimar 1918–1933: Die Geschichte der ersten deutschen Demokratie (Weimar 1918–1933: The History of the First German Democracy). Beck Paperback.

Zahn, Rebecca. 2015. “German Codetermination without Nationalization, and British Nationalization without Codetermination: Retelling the Story.” 36 Historical Studies in Industrial Relations 1.

The Organization for Economic Cooperation and Development (OECD) creates and reflects an intellectual center of gravity for economic policymakers among the advanced democratic industrial countries. Ultimately, its authority rests upon the credibility of its analysis, though its economic thinking has reflected and reinforced the paradigm shifts between different schools of economic thought.

Such influence has been especially evident regarding employment analysis and recommendations. The original OECD Jobs Strategy in 1994 marked the high point of an era of market fundamentalism. In the words of one ex-OECD staff member, “The OECD is seen as stressing the primacy of markets and thus of market-based solutions; institutions are generally viewed as hindrances, and deregulation is favoured over regulation.” The 1994 strategy argued that high unemployment, especially in Europe, was essentially a structural problem and that economies were operating close to the “natural rate” of unemployment, which could not be reduced through monetary or fiscal policy. The solution was to make labor markets “flexible” through the reduction of employment protection: Wages would be allowed to fall, and employment thus to rise, by reducing minimum wages and weakening collective bargaining systems.

But the most recent version of the OECD Jobs Strategy in 2018 came to a very different conclusion: “countries with policies and institutions that promote job quality, job quantity and greater inclusiveness perform better than countries where the focus of policy is predominantly on enhancing (or preserving) market flexibility.” The subsequent OECD report on collective bargaining released in 2019 argued: “Collective bargaining is a key institution to promote rights at work. At the same time, collective bargaining and workers’ voice are unique instruments to reach balanced and tailored solutions to the challenges facing OECD labour markets.” The claims that labor market, or job, flexibility should be a paramount goal of economic policy to minimize unemployment should correspondingly fade.

This paper explores the developments along that journey step-by-step:

• Calls for labor market flexibility had appeared in OECD documents since 1980, but the 1994 Jobs Strategy was influential in reinforcing the drive to remove what had come to be referred to as labor market rigidities in many OECD countries in the 1990s, namely by reducing minimum wages, decentralizing collective bargaining, weakening employment protection legislation, and reducing unemployment benefits.
• Over the following decade, despite assertions to the contrary, there was little empirical evidence to indicate positive employment results from these reforms.
• Divergence appeared between the European Union and the OECD over the support for a social-market as opposed to a free-market economy.
• The OECD revision of the Jobs Strategy in 2006 recognized that the impact of labor market institutions on employment performance depends on the economic and social setting. It was argued later that there are several “roads to Rome.”
• Growing income inequality, previously ignored by mainstream economists, was seen to have negative economic as well as social effects. Two major OECD reports—“Growing Unequal” and “Divided we Stand”—charted the rise in income inequality across OECD countries.
• The OECD acts in concert with the International Monetary Fund (IMF) and the World Bank, and in 2011 the IMF Research Department argued that inequality leads to imbalanced growth. Subsequent IMF research found that half of the rise in inequality in industrialized countries since 1980 was due to the decline of union density or bargaining coverage.
• The World Bank 2013 World Development Report on “Jobs” described a “plateau effect” of labor institutions on employment; specifically, it found only a small impact of minimum wages and employment protection on employment levels.
• As the effects of the Great Recession persisted, the costs of lightly regulated financial markets became clearer and the broader impacts of deregulated markets and free-market ideology were cast in a new light. During that time, the OECD launched a “New Approaches to Economic Challenges” program designed to revisit past policy prescriptions.
• Trade unions began to argue that the OECD’s labor market policy focus on education and skill levels was a necessary but not a sufficient condition for reducing inequality. More important were minimum wages and collective bargaining.
• Against the background of growing populism, a backlash against globalization and the digitalization of employment, and debates around the gig economy, the OECD revised the Jobs Strategy in 2018. The new strategy made 20 specific policy recommendations and for the first time recognized a positive role of collective bargaining, concluding that “well-designed collective bargaining systems are also found to promote labour market resilience.”
• The OECD’s 2019 report, “Negotiating Our Way Up: Collective Bargaining in a Changing World of Work,” concluded: “The need for co-ordination and negotiation mechanisms between employers and workers is heightened in the changing world of work.” Coordinated bargaining was seen to be superior in labor market outcomes, including for workers considered “marginal.”
• The OECD hosted the “Global Deal,” a multistakeholder partnership initiated by the Swedish government designed to benefit from, and contribute to, a platform that highlights the value of social dialogue and strengthens existing cooperation structures. Tacitly the OECD came full circle and recognized the advantages of the Nordic approach to greater cooperation and broader economic consensus, an approach that relies on collective bargaining and social dialogue.
• The global Covid-19 pandemic, beginning in 2020, has underlined the risks of social and economic inequality in the labor market and beyond.

Introduction

The Organization for Economic Cooperation and Development (OECD) creates and reflects an intellectual center of gravity for economic policymakers among the advanced democratic industrial countries. Krugman (2010) has described the organization as “conventional wisdom central.” It operates as a coordinating body and has been called a think tank for its member governments. Its leadership has been at pains to point out that it is a “do tank” (Gurria 2020), as it frequently issues recommendations. While there has always been some heterogeneity across different departments within the OECD, the power to establish this center of gravity lies within the Economics Department. That power in turn is derived from the national government treasury departments that nominate representatives to the OECD Economic Policy Committee and from the fact that the chair of the committee normally comes from the United States. These sources of power can be highly influential in determining broadly what is acceptable in economic policy and what is not. While the United States has often happily ignored the OECD views of American economic policy, other countries often cannot afford to do so without incurring the wrath of bond markets.

Ultimately, the OECD’s authority rests upon the credibility of its analysis, though its economic thinking has reflected and reinforced the paradigm shifts between different schools of economic thought. Such influence has been especially evident regarding employment analysis and recommendations. The original OECD Jobs Strategy in 1994 marked the high point of an era of market fundamentalism. While the strategy’s recommendations were often couched as euphemisms, the underlying messages were clear to decipher. In the words of one ex-OECD staff member evaluating the original Jobs Strategy 10 years after its launch: “The OECD is seen as stressing the primacy of markets and thus of market-based solutions; institutions are generally viewed as hindrances, and deregulation is favoured over regulation” (Casey 2004). The 1994 strategy argued that high unemployment, especially in Europe, was essentially a structural problem and that economies were operating close to Friedman’s (1968) “natural rate” of unemployment, which could not be reduced through monetary or fiscal policy. The solution was to make labor markets “flexible” through the reduction of employment protection: Wages would be allowed to fall, and employment thus to rise, by reducing minimum wages and weakening collective bargaining systems.

The most recent version of the Jobs Strategy adopted by the OECD Ministerial Council in 2018 came to a very different conclusion: “countries with policies and institutions that promote job quality, job quantity and greater inclusiveness perform better than countries where the focus of policy is predominantly on enhancing (or preserving) market flexibility” (OECD 2018). The subsequent OECD report on collective bargaining released in 2019 argued: “Collective bargaining is a key institution to promote rights at work. At the same time, collective bargaining and workers’ voice are unique instruments to reach balanced and tailored solutions to the challenges facing OECD labour markets” (OECD 2019). The claims that labor market, or job, flexibility should be a paramount goal of economic policy to minimize unemployment should correspondingly fade.

This paper explores the developments along that journey:

• Calls for labor market flexibility had appeared in OECD documents since 1980, but the 1994 Jobs Strategy was influential in reinforcing the drive to remove what had come to be referred to as labor market rigidities in many OECD countries in the 1990s, namely by reducing minimum wages, decentralizing collective bargaining, weakening employment protection legislation, and reducing unemployment benefits.
• Over the following decade, despite assertions to the contrary, there was little empirical evidence to indicate positive employment results from these reforms.
• Divergence appeared between the European Union and the OECD over the support for a social-market as opposed to a free-market economy.
• The OECD revision of the Jobs Strategy in 2006 recognized that the impact of labor market institutions on employment performance depends on the economic and social setting. It was argued later that there are several “roads to Rome” (OECD 2018).
• Growing income inequality, previously ignored by mainstream economists, was seen to have negative economic as well as social effects. Two major OECD reports—“Growing Unequal” (2008) and “Divided we Stand” (2011)—charted the rise in income inequality across OECD countries.
• The OECD acts in concert with the International Monetary Fund (IMF) and the World Bank, and in 2011 the IMF Research Department argued that inequality leads to imbalanced growth. Subsequent IMF research (2015) found that half of the rise in inequality in industrialized countries since 1980 was due to the decline of union density or bargaining coverage—itself the result in part of the drive for labor market flexibility encouraged by the earlier version of the Jobs Strategy and IMF structural adjustment programs.
• The World Bank 2013 World Development Report on “Jobs” described a “plateau effect” of labor institutions on employment; specifically, it found only a small impact of minimum wages and employment protection on employment levels.
• As the economic and financial crisis persisted following the 2008 Lehman Brother’s collapse, the costs of lightly regulated financial markets became clearer and the broader impacts of deregulated markets and free-market ideology were cast in a new light. During that time, the OECD launched a “New Approaches to Economic Challenges” program designed to revisit past policy prescriptions.
• Trade unions began to argue that the OECD’s labor market policy focus on education and skill levels was a necessary but not a sufficient condition for reducing inequality. More important were minimum wages and collective bargaining.
• Against the background of growing populism, a backlash against globalization and the digitalization of employment, and debates around the gig economy, the OECD launched a further revision of the Jobs Strategy. The new Jobs Strategy, published in 2018, made 20 specific policy recommendations and for the first time recognized a positive role of collective bargaining, concluding that “well-designed collective bargaining systems are also found to promote labour market resilience.”
• The OECD subsequently published in 2019 “Negotiating Our Way Up: Collective Bargaining in a Changing World of Work.” The report concluded: “The need for co-ordination and negotiation mechanisms between employers and workers is heightened in the changing world of work.” Coordinated bargaining was seen to be superior in labor market outcomes, including for workers considered “marginal” (young workers and women).
• The OECD hosted the “Global Deal,” a multistakeholder partnership initiated by the Swedish government designed to benefit from, and contribute to, a platform that highlights the value of social dialogue and strengthens existing cooperation structures. Tacitly the OECD came full circle and recognized the advantages of the Nordic approach to greater cooperation and broader economic consensus, an approach that relies on collective bargaining and social dialogue.
• The global Covid-19 pandemic, beginning in 2020, underlined the risks of social and economic inequality in the labor market and beyond.

From Keynesianism to structural adjustment policies, 1970–1994

Until the oil shocks of the 1970s, the predominant strand of OECD thinking was Keynesian. And as far as labor markets were concerned, its thinking was influenced by the Nordic model developed by Gösta Rehn and Rudolf Meidner when they were working at the Swedish trade union center, the LO, in the 1950s. Rehn went on in 1962 to become the director of what later became the OECD Department for Labor and Social Affairs.

Following the oil shocks of the 1970s the OECD’s conventional wisdom shifted to be more in line with the free-market thinking of the Thatcher and Reagan governments. Unemployment in the OECD area soared in the 1980s, but many governments saw this as a price worth paying for dampening inflation. The notion of a natural rate of unemployment determined by the structural features of economies was accepted, and macroeconomic policy, it was argued, could have little impact on employment—but government deficits would crowd out private investment and lead to inflation. A variant of the natural rate of unemployment—the nonaccelerating inflation rate of unemployment (NAIRU)—became the core of macroeconomic analysis in the OECD, the IMF, central banks, and finance ministries during the 1980s and 1990s. Economists made calculations for OECD countries of levels of unemployment that it was argued were “structural,” below which inflation would accelerate.

The view that many OECD labor markets, especially in continental Europe, were becoming sclerotic due to too much regulation to protect workers and set wages through minimum wages and collective bargaining became the conventional wisdom. Slow employment growth in Europe was contrasted with faster employment growth in the United States.

The new leitmotif of the OECD became structural reforms, especially reform of labor markets and labor market institutions. The 1983 OECD statement on structural adjustment program (OECD 1983) called for wage flexibility so that wages would reflect productivity, and policies that impeded this, including minimum wages and, in part, collective bargaining agreements, should be reformed. Employment protection legislation should be weakened to encourage management to hire more workers, secure in the knowledge that they could be fired. Internal work rules that could restrict management’s drive for cost cutting should be lifted. Although it was rarely made explicit, the OECD Economics Department saw the work bargain solely as a monetary exchange, thus legitimizing management prerogative and exploitation. The market and institutional failures that required labor market regulation to be introduced in the first place were never discussed.

Business, and especially American business, enthusiastically encouraged this drive. Barkin (1987) noted:

Management’s drive for the removal of contractual and governmental restraints on their control of the workforce is rationalized in Western Europe as necessary to achieve greater internal and external competitiveness. In support of this view the OECD substituted the advocacy of a flexible manpower policy (including wage policy) under the euphoric title of positive adjustment policy for the prior programme of an active manpower policy promoted during the 60s and early 70s.

A commission, serviced by the Employment and Social Department of the OECD and chaired by London School of Economics Director Ralph Dahrendorf, published a report in 1984 that sought to humanize the flexibility debate, arguing that reforms should not be instrumentalized against one group in society, namely workers. There is no sign that this argument had an impact on the mainstream recommendations of the OECD. Indeed, the conventional wisdom shifted little for two decades. Freeman (2005) commented:

Today, there is a new orthodoxy that makes the deregulation of labor market institutions and increased employment and wage flexibility in the labor market the keys to economic success. International agencies, such as the OECD and the IMF, and many economists blame unemployment and sluggish economic growth on unions and state regulations of pay and employment that purportedly reduce market flexibility. They recommend that governments weaken labor market institutions in favor of market driven solutions. They called for reductions in the pay of low wage workers to create additional demand for them and tax breaks for the highly paid to induce them to work more or harder.

The 1994 Jobs Study and Jobs Strategy

The original 1994 OECD Jobs Strategy, set in motion at the OECD ministerial meetings in 1992, codified this new orthodoxy as far as employment policy was concerned. The communique of the employment and labor ministers meeting in January 1992 set the tone, concluding: “Flexible and efficient labour markets are key to achieving non-inflationary economic and employment growth” (OECD 1992). The Jobs Study from the OECD secretariat and the resulting Jobs Strategy were published and adopted two years later at the 1994 Ministerial Council of Economic and Finance Ministers (OECD 1994a).

The strategy had originally nine recommendations, to which a 10th (increasing product market competition) was added the following year:

1. Fashion macroeconomic policy to encourage noninflationary growth.
2. Improve frameworks to enhance the creation and diffusion of new technology.
3. Increase flexibility of working time.
4. Eliminate impediments to the creation and expansion of enterprise.
5. Make wages and labor costs more flexible to reflect local conditions and individual skill levels.
6. Reform employment security provisions that inhibit the expansion of employment in the private sector.
7. Strengthen active labor market policies and reinforce their effectiveness.
8. Improve labor force skills and competency through reforming education and training systems.
9. Reform unemployment and related benefit and tax systems so that equity is not pursued at the expense of efficient labor markets.
10. Enhance product market competition to reduce monopolistic tendencies (added subsequently).

Of the 10 policy recommendations, one referred to macroeconomic policy and nine to microeconomic policy, reinforcing the underlying view that unemployment was a structural problem caused by insufficiently flexible labor markets. Of the nine structural policy recommendations, those concerning technology, improved skill levels, entrepreneurship, and increased competition were hardly controversial; they were described by Freeman (2005) as “boiler plate platitudes.” Four were recommendations to deregulate labor markets in continental Europe to bring them into line with the U.S. model.

The recommendation on wage setting specified that governments should “make wages and labour costs more flexible by removing restrictions that prevent wages from reflecting local conditions and individual skill level, in particular for younger workers,” and the recommendation on employment protection called upon governments to “reform employment security provisions that inhibit the expansion of employment in the private sector” (OECD 1994a). The strategy focused on provisions regulating dismissals and redundancies and those governing temporary employment contracts. On unemployment benefits, the OECD recommended “cutting unemployment benefit levels and duration of payment, tightening eligibility and enforcing work requirements, and restricting entry to and the generosity of early pensions” (Casey 2004).

The recommendations bore a strong similarity to the policies of the Thatcher government for “flexibilising” labor markets in the U.K. in the 1980s. These have been summarized by Blanchflower and Freeman (1993) as “industrial relations laws that weakened union power; measures to enhance self-employment; privatization of government run or owned businesses; reduction in the value of unemployment benefits and other social receipts relative to wages; new training initiatives; tax breaks to increase use of private pensions; lower marginal taxes on the individuals; elimination of wage councils that set minimum wages.” The Jobs Strategy followed rather than predated these reforms.

Nickell (2017) argued that the recommendation on active labor market policies, which called for strong job search conditionality to be attached to receipt of unemployment benefits, reflected a “Nordic approach” to activation policies. However, the strategy did not recommend the Nordic institutions of strong unions and high levels of expenditure on labor market policies.

Though the Jobs Study and Jobs Strategy were backed up by two volumes of “evidence and explanations” by the OECD (1994b), they offered little external empirical evidence to support the employment case for deregulating labor markets. As is discussed below, over the succeeding two decades no strong empirical evidence has emerged to support the claims for flexibilizing labor markets to obtain positive employment effects—among neoliberal economists the claim was simply taken as self-evident. Freeman (2005) notes:

Adherents to the new orthodox view search the data for specifications…[or] measures that support their priors, while barely noticing evidence that goes against them. If results are inconsistent with the priors, they assume that something is wrong with their empirical specification or measures, rather than question the validity of their case.

Follow-up, the 1998 ‘assessment’ of the Jobs Strategy, and ‘Going for Growth’

The OECD follow-up to the Jobs Strategy included a series of thematic reviews but also detailed country recommendations processed through the OECD’s Economic Development Review Committee (EDRC), which is responsible for producing annual or biennial economic surveys of member states. As with the Economic Policy Committee, the EDRC is made up of officials from the treasury, economics, and finance ministries. It uses a peer review process in which two countries comment on a draft of recommendations, prepared by the secretariat, to the country being surveyed. The EDRC’s work is the most closed and least transparent of all OECD committees and working groups.

The country-specific recommendations from the Jobs Strategy follow-up primarily focused on continental European countries.1 In the 1990s the only wage-setting recommendation directed at the United States was that it weaken minimum wage laws.

Regarding the recommendations to governments for increasing flexibility by wage-setting institutions, the follow-up included detailed proposals beyond the broad policy guidance. The recommendations included:

• Refocusing collective bargaining at the sectoral level to the provision of framework agreements that leave firms with more leeway to adjust wages to local conditions.
• Introducing opening clauses for local bargaining parties to renegotiate sectoral agreements.
• Phasing out administrative extensions of agreements that were considered to rigidify wage-setting arrangements.
• Reassessing the role of statutory minimum wages and either switch to better-targeted redistributive instruments or minimize their adverse effects by introducing sub-minimum wages differentiated by age or region and/or indexing them to prices instead of average earnings.

The five-year assessment of the Job Strategy in 1999 carried out by the OECD secretariat was largely self-congratulatory. It argued that many OECD countries had sought to implement the strategy’s recommendations, and those that had had enjoyed better employment performance. However, attempts to quantify the impact on employment of some of the key recommendations on labor market regulation produced insignificant results. The 2004 OECD Employment Outlook examined employment protection legislation (EPL) and acknowledged that, “The net impact of EPL…on aggregate unemployment is therefore ambiguous a priori, and can only be resolved by empirical investigation. However, the numerous empirical studies of this issue lead to conflicting results, and moreover their robustness has been questioned” (OECD 2004, Chapter 2). Nevertheless, reflecting the organization’s “priors,” the structural policy recommendations became a high-profile mantra for the OECD. The biennial “Going for Growth” report first published in 2005 repeated the self-reinforcing methodology of the initial Jobs Strategy assessment.

The influence of the OECD and the self-reinforcing relationship with finance ministry officials went beyond Europe. Jackson (2007) commented from a Canadian perspective:

…OECD processes have been influential and important in terms of defining the “conventional wisdom” that drives economic policy advice. Seen through the prism of published country reviews and based on information provided by senior Canadian government officials and interviews, the OECD strongly influenced the main themes of Canadian economic and labor market policy over the 1990s: very large cuts to the deficit achieved by cuts in social spending; deep cuts to the unemployment insurance programme; deregulation and privatisation; the pursuit of greater labor market flexibility; formal targets for low inflation; and a major focus on debt reduction and tax cuts as opposed to reinvestment in social programmes after the elimination of the federal deficit.

Factors behind the 2006 review of the Jobs Strategy: Different ‘roads to Rome’

In the early 2000s, thinking in the OECD began to shift. Two-thirds of the OECD members at that time were also members of the European Union, and the European Commission’s European Employment Strategy, developed over the same period as the Jobs Strategy and having many similarities, was more influenced by “social market theories whereby the state intervenes to moderate the negative effects of market relationships and to enhance the efficiency of market performance” (Casey 2004). The social dimension of European integration and instigation of what was known as the European Social Dialogue—negotiations between employers and trade unions at the European level—that was instigated by European Commission President Jacques Delors was an attempt to build agreement around a more social model of labor market reform. Already in 1997 a joint seminar between the OECD and the European Commission to discuss the implementation of the two strategies concluded that:

A number of member countries, notably in EU, have however been reluctant to implement the recommendations relating to labour market flexibility. As the OECD itself acknowledges this is due to concern the policies to achieve greater flexibility in the labour market would be at odds with the objectives concerning equity and social cohesion. The trade-off posed is clearly a difficult one. (European Commission 1999)

Labor and employment ministers meeting in 2003 called on the OECD to “reassess the Jobs Strategy in the light of more recent experience and future challenges.” The 2004 OECD Employment Outlook prepared by the Department of Employment, Labor, and Social Affairs began the work by including detailed chapters on two of the flexibility recommendations of the Jobs Strategy: the impact of employment protection legislation, and wage setting. The Outlook concluded that “the evidence of the role played by employment protection legislation on aggregate employment and unemployment rates remains mixed” (OECD 2004). It expressed concern that in some countries such as Spain temporary contracts that replaced permanent jobs produced labor market duality between those with permanent contracts and those with temporary contracts, and it recognized that job insecurity itself was a problem. The Outlook examined the Job Study diagnosis that excessively high aggregate wages and wage compression hindered employment creation and found that the “evidence is somewhat fragile.” It concluded that the effect of collective bargaining on employment “is contingent upon other institutional policy factors that need to be clarified to provide robust policy advice.”

In 2004, at an OECD seminar organized by the Trade Union Advisory Committee to the OECD (TUAC), Washington Center for Economic and Policy Research (CEPR), and the European Trade Union Institute (ETUI), the authors of a set of empirical surveys on the effect of labor market regulations on employment (Baker et al. 2004) presented their findings. The report of the meeting2 noted that the results “showed no statistically significant relationship between labour market protection and unemployment.” The authors concluded that there was a “yawning gap between the confidence with which the case for labour market deregulation has been asserted and the evidence that the regulating institutions are the culprits” (Baker et al. 2004). While the Economics Department director acting as discussant in the meeting found the results “uninteresting,” the response from governments present was more nuanced. The British treasury official who at the time was also chair of the Economic Committee Working Party on Structural Policy noted that the introduction of the minimum wage in the U.K. had been effective in raising wages of less-skilled workers without raising unemployment.

Other research in the 2000s that focused on the emerging market economies in Central and Eastern Europe similarly failed to find evidence of the positive effects of labor market deregulation. Avdagic (2015), examining the effect of employment protection legislation on aggregate and youth unemployment in advanced economies and Central and Eastern Europe during 1980–2009, concluded: “The results offer no clear support for the argument that EPL is a cause of unemployment.…[T]he findings on the whole indicate that government efforts to tackle unemployment by deregulating EPL alone may well be futile.”

Heimberger (2020) subsequently carried out a meta-analysis of 75 studies across a range of countries examining the relationship between employment protection legislation and unemployment, concluding:

We cannot reject the hypothesis that, on average, the genuine empirical effect of EPL is zero. Notably, this main finding would be consistent with an explanation according to which the effects of employment protection are not universal, as increased employment protection may have different effects on unemployment in different countries or time periods.

Some studies have discovered potentially positive welfare-improving effects of employment protection, depending on institutional settings. Belot, Boone, and van Ours (2007) found “new results on the welfare effects of employment protection”:

Using data from 17 OECD countries, we show that there exists an inverse U shape relationship between employment protection and economic growth. Using a simple theoretical model with non-contractible specific investments, we show that over some range increasing employment protection does indeed raise welfare. We also show that the optimal level of employee protection depends on other local market features, such as the bargaining power of workers and the existence of wage rigidities like the minimum wage.

Vergeer and Kleinknecht (2011) analyzed the impact of labor market deregulation on productivity in 19 OECD countries for a longer period, 1960–2004. They concluded:

…wage cost saving flexibilization of labour markets has a negative impact on labour productivity growth. A one percentage point change in growth rates of real wages leads to change in Labour productivity growth by 0.31 to 0.39 percentage points. This cannot solely be explained by hiring low-productive labour. Flexibilization of Labor markets leads to a labour-intensive growth path that is problematic with an ageing population in Europe.

Regarding collective bargaining, the OECD’s 2006 reassessment of the Jobs Strategy, drawing on Bassanini and Duval (2006), concluded that:

…high corporatism bargaining systems tend to achieve lower unemployment than do other institutional setups. Nevertheless, the evidence concerning the impact of collective bargaining structures on aggregate employment and unemployment continues to be somewhat inconclusive. The overall non-robustness of results across studies probably reflects, at least in part, the difficulty of measuring bargaining structures and practices, as well as the fact that the same institutional set-up may perform differently in different economic and political contexts. One exception to this pattern is the robust association between higher centralisation/coordination of bargaining and lower wage dispersion…. The empirical evidence concerning a negative impact of minimum wages on unemployment is mixed, with some studies finding evidence for significant effects particularly for youth while others do not detect any effects. The Bassanini and Duval research found no significant impact of the minimum wage on the aggregate unemployment rate. (OECD 2006)

The “Reassessed Jobs Strategy” in the 2006 Employment Outlook (OECD 2006) represented a shift in thinking by reflecting the uncertainty of the empirical evidence about the effects of labor market institutions and accepting that different national and institutional settings are key to understanding good employment performance. Acknowledging this a decade later the OECD said:

The 2006 Reassessed Job Strategy placed more emphasis on promoting labour force participation and improving job quality. The main message was that there are several roads to Rome, i.e., good labour market performance is consistent with more market reliant models that emphasise labour and product market flexibility, but also with models that involve a stronger role of public policies, generally coupled with strong social dialogue and a combination of stronger protection for workers with flexibility for firms. (OECD 2018)

Coats (2018) noted:

…the OECD, in its reassessment of the Jobs Study…stepped back from an unqualified endorsement of the Anglo-Saxon model of labour market flexibility. In part, this was because the evidence showed more than one route to strong jobs growth. For example, in the Nordic countries and Netherlands (to some extent), strong collective bargaining institutions, social dialogue, generous out-of-work benefits, rigorous job search requirements and investment in human capital constituted a package of policies that were just as successful at creating jobs as the deregulated Anglo-Saxon model. A judicious mix of flexibility and security seemed to have been achieved, to which the neologism “flexicurity” was applied.

Watt (2006) noted: “analysis of the policy recommendations and the underlying evidence presented in the Employment Outlook suggests that since 1994 the OECD has moved a considerable way on a number of key policy issues.”

Lansley and Reed (2010), drawing on international experience for recommendations for U.K. employment policies for British trade unions in the light of the Great Recession, concluded:

…the fundamental arguments of the anti-regulationist school and its belief in self-regulating markets simply don’t stand up to the economic experience of the last two decades. The free-market experiment has been plagued by instability, not the stability it predicted. The empirical evidence provides no backing for those calling for the weakening or abolition of the minimum wage and cutbacks in current and planned labour market interventions. While badly thought-out regulation can be harmful, the evidence is that it is possible to achieve successful economic outcomes (low unemployment, high employment participation and growth) with strong social and workplace protection. More regulation does not necessarily mean poorer economic performance while increased regulation of the appropriate kind can actually improve performance in the right circumstances. Indeed, the OECD once the champion of the orthodox view has accepted the case for intervention in recent years.

Part of what drove the selling of the Anglo-Saxon model of labor market flexibility was the perceived success of the U.S. labor market in providing low unemployment rates. But, even in the U.S., where most workers do not have contract protection under law, studies on the recent and limited protections workers have from at-will dismissal gave a mixed view on employment but a mostly positive view on workers’ wages. Women and workers of color, in particular, had higher wages with laws protecting against unjust or unreasonable cause protections from firing (Autor 2003; Kugler and Saint-Paul 2004; Miles 2000; Hoyt 2018). Showing improvements from the U.S. extreme of employer flexibility on an important labor market metric.

Within the OECD a break in consensus was allowing a subtle shift away from the neoliberal paradigm. There had been some opposition within the OECD to the structural policy hegemony of the Economics Department, and successive Employment Outlooks from what had become the Department of Employment, Labor, and Social Affairs produced research results that did not support the Economics Department’s enthusiastic drive for deregulation. Efforts to quantify the adverse economic effects of labor market regulations produced inconclusive results.

Nevertheless, neoliberal economists within and outside the OECD have continued to advocate flexibilizing labor markets. For instance, the “Going for Growth” report, still one of the flagship publications of the Economics Department, identifies five structural reform priorities for each OECD country, and until 2011 the benchmark for identifying these was a comparison with gross domestic product (GDP) per capita in the U.S. The 2010 “Economic Survey of Slovakia” gives the flavor of typical recommendations: It called on the government to undertake reforms “of a bold nature” and argued that reforming the wage determination system “[so as to allow] job seekers to price themselves into the market and employment will help to reduce poverty risks, reduce social expenditure pressures, limit the economic costs of fiscal consolidation, help lower entry barriers for innovative entrepreneurs and increase efficiency of active labour market measures” (OECD 2010).

The elephant in the room and the elephant curve: Rising income inequality

If there was growing uncertainty and disagreement within the OECD, the IMF, and the World Bank over the past two decades about the efficiency and employment effects of labor market institutions, there was growing consensus that rising inequality, primarily in advanced economies, was a major concern for economic as well as social and political reasons.

At the beginning of the current century the conventional view of most economists in the OECD and the international financial institutions (IFIs) would have been that distributional issues were questions for political and social decision-makers at the national level and only relevant to economists if measures to address inequality had a significant economic cost, which it was often argued was the case. Lucas (2003), author of the rational markets’ theorem, wrote: “Of the tendencies that are harmful to sound economics, the most seductive and in my opinion the most poisonous, is to focus on questions of distribution.”

At the time this would have been the consensus view of many economists within the OECD. To the extent that distributional issues were raised, it was frequently pointed out that globalization was reducing inequality between the developing countries and the OECD countries, explained almost entirely by the rapid industrialization of China and India since 1990—which, it was argued, was due to the success of market reforms. Rising inequality within countries was ignored.

The global picture of rising incomes in poor countries, exploding incomes of the top 1%, and stagnant or falling incomes of the middle class in OECD countries was depicted in 2016 by Milanovic (2016) of the World Bank in an “elephant curve,” which showed that those who had lost out over the period had been low- and middle-income groups in developed countries.

Complacency concerning inequality as well as the enthusiasm for the promotion of the light regulation of markets in general was shaken by the near meltdown of the global economy following the 2008–2009 financial crisis and the Great Recession. Stagnation of middle-class incomes in the United States was seen as one of the factors behind the growth of unsustainable subprime lending.

As noted above, the thinking on the employment and social side of the OECD concerning labor markets and in particular income inequality had already begun to shift in the 2000s, as reflected by the publication of a series of reports on income inequality—notably “Growing Unequal” in 2008 (OECD 2008) and “Divided We Stand: Why Inequality Keeps Rising” in 2011 (OECD 2011). Rising income inequality began to be regarded as an economic as well as a social and political problem. In the foreword to a 2015 OECD publication summarizing this work, OECD Secretary General Angel Gurria wrote:

Inequality is bad and getting worse. In the 1980s, the richest 10% of the population in OECD countries earned seven times more than the poorest 10%. They now earn nearly ten times more. When you include property and other forms of wealth, the situation is even worse: in 2012, the richest 10% controlled half of all total household wealth and the wealthiest 1% held 18%, compared to only 3% for the poorest 40%. The poorest members of society suffer immediately from inequality, but in the longer term, the whole economy is also damaged. OECD figures show that the rise in inequality observed between 1985 and 2005 in 19 OECD countries knocked 4.7 percentage points off cumulative growth between 1990 and 2010. (Keely 2015)

In 2017 Gurria stated in a speech to employers:

Inequalities harm growth. They erode trust in governments, in business, in modern capitalism and in democracy. They also contribute to a polarised and dangerous environment where populism, protectionism, and exclusive nationalism tend to grow and spread. We urgently need to reverse these trends. (Gurria 2017)

The shift in rhetoric was not limited to the OECD. Then-IMF Managing Director Christine Lagarde stated in a speech to the 2012 annual meetings of the IMF and World Bank: “Excessive inequality is corrosive to growth; it is corrosive to society. I believe the economics profession and the policy community have downplayed inequality for too long” (Lagarde 2012). These statements were made at the time when the Occupy Wall Street protests had reverberations around the world, with public demonstrations erupting against economic inequality and the injustice of wealth and income being concentrated in the top 1%.

The picture painted by the OECD has become familiar in retrospect, but at the time it was significant and unusual, coming as it did from an organization that was seen as being highly orthodox in its analysis and policy prescriptions. Significantly, the U.S. launch of “Divided We Stand” was delivered in front of a trade union audience at an event held at the AFL-CIO and chaired by its president, Rich Trumka. It provoked Daniel Mitchell of the Cato Institute to decry Angel Gurria as “an International bureaucrat pushing socialism…using your tax dollars to push for class warfare” (Mitchell 2011). The analysis of household income data in “Divided We Stand” showed that in most OECD countries the incomes of the top 10% had grown faster than the bottom 10% over the previous two decades, resulting in an overall rise in inequality. The Gini coefficient3 rose over the same period in 17 out of 24 OECD countries for which long-term data series were available—the OECD average had risen almost 10%, from 0.29 to 0.316—and the rise in household inequality was primarily attributed to changes in the distribution of wages and salaries (OECD 2011). The highest levels of inequality in the larger OECD economies were in the English-speaking countries, notably the U.K. and the U.S., where inequality had first begun to rise in the 1980s after falling in the postwar decades. However, increases were also seen in traditionally low-inequality countries—the exceptions being Turkey, Greece, France, Hungary, and Belgium. In subsequent analyses of tax data for a more limited group of countries, in which the OECD drew on data published in Piketty’s (2014) bestselling 700-page study “Capital in the Twenty-First Century,” the OECD noted:

…from 1975 up to the crisis, the top percentile managed to capture a very large fraction of the growth in pre-tax incomes, especially in English speaking countries: around 47% of total growth went to the top 1% in the United States, 37% in Canada, and above 20% in Australia and the United Kingdom. By contrast, in Nordic countries, but also in France, Italy, Portugal, and Spain it was the bottom 99% of the population which benefited from more growth, receiving about 90% of the increase in total pre-tax income between 1975 and 2007. (OECD 2014a)

OECD work that began with “Divided We Stand” also found unconvincing that the focus of concern should be “inequality of opportunity” rather than “inequality of outcome” and that cross-generational mobility would offset inequality over time. It broadly confirmed the succinct argument of Atkinson (2015) that, “If we are concerned about inequality of opportunity tomorrow, we should be concerned about inequality of outcome today.” Inequality is transmitted through generations in part due to lack of access to education—high inequality is self-reinforcing. Moreover, high-inequality countries tend to display low intergenerational mobility, reflected in what came to be known as the “Gatsby Curve,” made famous by Alan Krueger (2012). The OECD concluded that, “One of the main objectives of social policy is to break the cycle of disadvantage across generations and prevent the development of a self-replicating underclass” (OECD 2008, 216).

Reports prepared for OECD committees in 2011 had highlighted the impact of regulatory reform and changes in labor market institutions alongside technological change on rising wage inequality. However, when it came to going public in 2012 in “Divided We Stand” on the causes of the rise in inequality, the OECD remained within a conventional analytic comfort zone. It dismissed the effects of globalization and financialization, arguing “neither rising trade integration nor financial openness had a significant impact on either wage inequality or employment trends within OECD countries” (OECD 2011, 29). While raising the possible effects of deregulation policies, including the weakening of unions that the OECD itself had been recommending for two decades, the report concluded that the effects were mitigated by the effects that deregulation, including wage suppression, could have by raising employment levels and thereby reducing inequality “amongst workers and jobless individuals.” Skill-biased technological change was seen as the key driver of inequality. The report noted that taxation systems had become less effective in redistribution and that “redistribution strategies based on government taxes and transfers alone would be neither effective nor financially sustainable” (OECD 2011, 40). A major feature of the report was therefore to highlight “the central role of education. The rise in the supply of skilled workers considerably offset the increase in wage dispersion associated with technological progress, regulatory reforms and institutional changes” (OECD 2011, 31). The overall policy conclusion was that economic strategy “should rest on three main pillars: more intensive human capital investment; inclusive employment promotion; and well-designed tax/transfer redistribution policies” (OECD 2011, 41).

Subsequent OECD work in 2014 quantified the negative impact of rising inequality on economic growth across a range of OECD countries. The main transmission mechanism was found to be the impact on human capital accumulation in low-income households and the persistence of this across generations. The summary of the work found that rising income inequality “has a negative and statistically significant impact on subsequent growth. In particular, what matters most is the gap between low-income households and the rest of the population. In contrast, no evidence is found that those with high incomes pulling away from the rest of the population harms growth” (Cingano 2014).

A series of reports from other institutions including the IMF (referred to below) around the same time broadened the list of causal factors that led inequality to reduce economic growth. Stiglitz (2012) found that a rising concentration of income at the top of the distribution reduces welfare by allowing top earners to manipulate the economic system in their favor.

The OECD was also reluctant to tackle the issue of rising functional inequality and the decline of the share of wages in national income. At the 2011 Washington launch event of “Divided we Stand,” Richard Trumka, in broadly welcoming the work, noted that the report “seems to focus primarily on the distribution of wages as a principal driver of income inequality. You do not seem to view the more fundamental fall in share of wages in national income as a major cause of rising inequality. I think both the fall in wage shares and the rise in inequality are important causes of rising inequality and both have implications for policy” (Trumka 2011). Until the 1980s the share of wages in national income in industrialized countries was constant to such an extent that economists treated it as “stylized fact,” and different theories of the functional distribution of income were built around it. Neoclassical theory postulated that income distribution is determined by the marginal productivity of factors of production. Neo-Keynesian theory postulated that income distribution was determined by technological progress (Kaldor 1955). The data changed significantly after the Thatcher-Reagan reforms. The International Labor Organization (ILO) found that starting in the early 1990s wage shares had fallen in three-quarters of the 69 countries it studied; on average wage shares in industrialized countries fell by an average of nine percentage points of GDP over three decades (ILO 2011). In explaining the decline in wage shares the ILO found that “financialization” and “globalization” both played important roles in weakening the bargaining power of workers compared to business, as did declining union strength and the weakening of labor market protection. In contrast, the IMF and the OECD argued that technological progress was the main cause of the declining wage share, with capital substituting labor through automation (IMF 2017). As seen within the declining wage share, the rise in inequality, it was argued, was primarily due to skill-biased technological change, whereby high-skilled workers enjoy a wage premium.

Despite the reluctance of the OECD at that time to publicly accept that labor market flexibility was behind much of the rise in inequality, elsewhere there was increasing acceptance of the evidence that wage-setting mechanisms such as collective bargaining and wide trade union membership reduce income inequality, and also that the weakening of these institutions in industrialized countries through changing power relationships between workers and business over the past three decades has been an important cause of rising personal income inequality and a driving factor in explaining the fall in the wage share. Deakin, Malmberg, and Sarkar (2014), examining the effects of weakening labor laws in six OECD countries from 1970 to 2010, found that “worker protective labour laws are associated with a higher labour share and therefore, in broad terms, with improved income distribution—an outcome driven by laws on working time and employee representation.” Guschanski and Onaran (2017) found “a robust effect of institutional factors such as union density and minimum wages on the wage share, lending strong support to the political economy approach to functional income distribution.”

The shifting views of the IMF and World Bank

The advocacy of labor market deregulation reflected by the OECD Economics Department had been made with equal strength at the international financial institutions during the 1990s and early 2000s (IMF 2003). To some commentators, this stance was influenced by political influences rather than empirical evidence. According to Freeman (2005), the fact that trade unions and other institutions could resist structural adjustment programs imposed by the IFIs “led IMF-associated economists to stress the dangers of insufficient labour market flexibility in economic crises even when those crises arise from problems far removed from the labour market.”

However, different views came from within the IMF. An IMF 2011 Research Department report challenged fundamentally the traditionally benign view the IFIs had held toward income distribution. It concluded that higher inequality is associated with lower and less sustainable growth in the medium term even in advanced economies (Berg and Ostry 2011). A 2015 study by Jaumotte and Buitron examining the rise in inequality in advanced countries between 1980 and 2010 found that:

…the rise of inequality in the advanced economies included in this study has been driven by the upper part of the income distribution, owing largely to the increase in income shares of top 10 percent earners. We find evidence that the decline in union density—the fraction of union members in the workforce—is strongly associated with the rise of top income shares.…On average, the decline in union density explains about 40 percent of the 5-percentage point increase in the top 10 percent income share. This contribution rises to over 50 percent when controlling for sectoral employment shifts over the sample period.

The study was widely quoted in the press, including in the U.S., where the Los Angeles Times reported:

The IMF’s analysis undermines the accepted wisdom that lower union membership affects chiefly low- and moderate-income workers. The fund’s analysts…find instead that the impact of declining unionization is felt across the entire income spectrum. The trend not only reduces the welfare of the lower income worker, they find; it makes the rich richer. (Hiltzik 2015)

The Economic Policy Institute had been publishing regular briefs drawing attention to the link between increasing inequality and declining union membership.

In a 2013 assessment of the IMF’s advice on labor markets in advanced economies in the Great Recession, IMF then-Chief Economist Oliver Blanchard together with IMF researchers Florence Jaumotte and Prakash Loungani cautioned: “…the implications of alternative structures of collective bargaining are poorly understood, suggesting that the IMF should tread carefully in its policy advice in this area. Moreover, trust among the social partners appears to be just as important in bringing about macro flexibility as the structure of collective bargaining.” The IMF was criticized by trade unions at the time4 for not reflecting the reality of IMF country-level recommendations. The International Trade Union Confederation (ITUC) general secretary, writing to the IMF managing director, noted that “nothing we have seen from recent IMF reports on European countries indicates that this is being put into practice.”

From these exchanges and starting in 2013, Blanchard organized with the international trade union organizations—the ITUC and the TUAC—a series of workshops over the following two years on collective wage bargaining. The workshops brought together labor market academics, trade union representatives, and IMF staff. Conclusions of the meetings were not published by the IMF, but in the opinion of one of the current authors the discussions contributed to more caution in the IMF’s advocacy of labor market flexibility at the global level even if not reflected in national programs.

The conventional wisdom on labor market institutions shifted at the same time in parts of the World Bank. The title of the World Bank’s 2013 World Development Report was “Jobs,” and a chapter devoted to “Labour Policies Revisited” found that there was a “plateau effect” of the impact of labor regulations on employment and efficiency. At the edges of the plateau, where regulations were highly stringent or highly lax, the regulations might have a significant impact on economic efficiency and employment. But in between across most national settings, “estimated effects prove to be relatively modest in most cases—certainly more modest than the intensity of the debate would suggest” (World Bank 2012).

The impact of the financial crisis and the Great Recession

The Great Recession and rise in unemployment in most OECD countries following the 2008–2009 financial crisis was the key backdrop to the discussion of inequality and the role of labor market institutions over the following years, but the downturn had conflicting effects on the discussion of labor regulation at the international organizations. The ILO had dubbed the rise in inequality the “crisis before the crisis.” There was also growing recognition that the depression of middle-class incomes in the U.S. had led to credit-driven growth encouraged by lightly regulated financial markets, with conflicted governance, that proved to be unsustainable. There was also increasing awareness that the role of trade unions in social partnership arrangements such as in Germany had demonstrable positive outcomes in terms of moderating the impact of the crisis on employment. The debate on how to share the cost of the crisis had an impact on the political debate. However, there was also renewed attention given to the relationship between employment regulation and unemployment (Heimberger 2020). The causes of the Great Recession lay in financial markets, not labor markets, but the notion to never waste a crisis took hold among neoliberal commentators, and they employed it to promote the deregulation of labor market institutions. In Europe during the sovereign debt crisis beginning in 2010, unemployment rose to over 25% in Spain and Greece and to 16% in Portugal. The countries seeking external assistance to support their banking systems—Greece, Portugal, Cyprus, and Ireland—were forced to accept draconian austerity and structural reforms by the troika of the European Commission, the European Central Bank, and the IMF. Greece was obliged as part of the support package to adopt brutal cuts in minimum wages and the weakening of collective bargaining systems to bring about internal deflation of incomes. Despite the IMF Research Department’s own evidence of the impact of past policy in weakening labor market institutions, IMF country programs continued to promote labor market deregulation. A series of working papers from IMF authors published in 2012 that purported to show that labor market reforms were strongly associated with declines in unemployment were criticized by ILO authors for having serious data flaws (Aleksynska 2014).

The trade union movement made its own prescriptions for dealing with the crisis with its study, “Exiting from the Crisis: Towards a Model of More Equitable and Sustainable Growth” (Coats 2011). TUAC and AFL-CIO Chief Economist Ron Blackwell played a key role in putting the document together and recruiting the Nobel laureate Joseph Stiglitz to provide a preface. Stiglitz’s authority helped the argument that inequality had to be addressed as part of the path forward. The book challenged the narrow focus on GDP per capita as an economic yardstick and pushed for a rethinking of the Washington consensus—the package of neoliberal policy reforms pressed on many developing countries by the international financial institutions since the 1980s.

The Pittsburgh G20 Summit in 2009 made two key innovations that had significant implications for global employment policy as well as economic governance. The first was the decision to continue to focus on employment policy, in light of which the U.S. labor secretary was asked “to invite our Employment and Labour Ministers to meet as a group in early 2010 consulting with labour and business” (G20 2009). This set in motion for the first time the G20 labor and employment ministerial meetings; prior to that the only G20 ministerial meetings had been of finance ministers.

The second innovation was the statement by leaders declaring “the G20 to be the premier forum for our international economic cooperation.” This gave a degree of permanence to the G20 leaders’ meetings and, with the institution of the labor and employment ministerial meetings, ensured that a listening session with labor union leaders of the G20 countries would be included. That innovation helped encourage language in the G20 to respect tripartite meetings between government, business, and trade unions at the national level.

But the push for labor market reform persisted. In the Mutual Assessment Process (MAP) of the G20 Framework Working Group (FWG) of finance ministry and central bank officials, participants argued that structural reforms of labor and product markets could lift growth by 2% over five years. Using “Going for Growth” methodology, they offered governments a way to remedy disappointing growth when fiscal policies were set on a track of deficit reduction and monetary instruments were exhausted.

In assessing government commitments to lift growth, the OECD identified some 700 proposed reforms, about a quarter of which were labor market reforms. At the Brisbane G20 Summit in November 2014, leaders agreed that “against the backdrop of a disappointingly weak cyclical recovery from deep recession, weakened productive capacity in key economies and a legacy of vulnerabilities from the financial crisis, we need to pursue an integrated approach to boost growth” (G20 2014). Leaders “set an ambitious goal to lift the G20’s GDP by at least an additional two per cent by 2018.” The main mechanism for achieving this goal was the implementation of “structural reforms to lift growth and private sector activity, recognising that well-functioning markets underpin prosperity.” Analysis by the IMF-OECD had indicated “that our commitments, if fully implemented, will deliver 2.1 per cent.” As the growth trajectories were more and more off target, the action plan was subsequently and quietly forgotten. In the words of one participant: “In retrospect the 2014 G20 structural reform initiative looks like a last hurrah for the deregulatory labor market strategy driven by the OECD Economics Department. However, at the time it was very worrying [though] the wind was changing already within the OECD and the economics profession (Pursey 2021).

The 2015 G20 meetings of labor and employment ministers, held in Turkey, discussed as a priority inequality and the decline of the labor share. The final declaration abandoned the deregulatory framework and stated: “In order to address rising inequalities and declining labour income shares, we agree to undertake a mix of policies appropriate to our national circumstances including improving wage-setting mechanisms, institutions for social dialogue, social protection systems and employment services.” The ministers endorsed an annexed set of “G20 Policy Priorities on Labour Income Share and Inequalities” that gave support to collective bargaining systems. Two important commitments of the principles were: “Strengthening labour market institutions [social dialogue, collective bargaining, wage-setting mechanisms, labor legislation] based on respect for the Fundamental Principles and Rights at Work; and reducing wage inequality, through policy tools such as minimum wages and the promotion and coverage of collective agreements, ensuring fair wage scales and that work pays” (G20 2015).

The populist backlash and the geography of discontent

Starting with the Occupy Wall Street movement, something began to change in OECD countries regarding the toleration of inequality. There were explosions of anger against governments and the “elite” from both the left and right, with paradoxical political implications. The Brexit vote in the United Kingdom in 2016, the election of Donald Trump in the United States the same year, the growth of nationalist and anti-immigration parties in northern Europe and, in 2018, the “yellow vest” insurrection in France each can be interpreted as a populist reaction to the rising inequality, stagnant median incomes, and economic insecurity that followed the Great Recession. They reflected a growth of relative deprivation, where significant segments of populations felt that they and their families had lost out—and they feared a future of even greater insecurity. These sharpening divisions appeared after three decades of the weakening of trade unions, whose economic role was to act as a brake on rising inequality and whose political role was to provide voice to those feeling unjustly treated and to negotiate solutions to grievances. Brexit, Trump, nationalism, and street violence all represent bad answers to an important question—how to re-forge agreement on distributive justice for those who have lost out (or so feel) from globalization, technological innovation, and responses to climate change.

Concern over the political fallout of the Great Recession as well as the organization’s failure to anticipate the risks of deregulated markets and rising inequality led the OECD to launch a program of “New Approaches to Economic Challenges” (NAEC) that was intended to assess what had gone wrong in previous OECD models and policy recommendations. One of the key messages from the NAEC work to the OECD Ministerial Council in 2014 was that “the last three decades have seen a rise of inequality, which can effect economic growth, weaken social cohesion and sap trust in markets and institutions. To address the growing concerns linked to increasing inequality, policy makers are advised to support a move to a more inclusive and sustainable economic approach” (OECD 2014b). The OECD “Inclusive Growth Initiative” recognized the need to act across a range of different policy areas and called for a “’whole of government’ approach to make sure that financial, fiscal or monetary decisions, among others do not undermine social cohesion or social progress” (OECD 2017b).

In successive submissions to the OECD, its ministerial council meetings, and annual meetings with national representatives, TUAC had proposed a widening of government action to reduce inequality beyond the traditional remedies of improving skills and to operationalize the recommendations of the NAEC project. TUAC’s submission of December 2013 called for:

…a comprehensive strategy on tackling inequality and moving to inclusive growth including action to: address the growth of in work poverty through establishment of well set minimum wages; strengthen the coverage of collective bargaining by the social partners and adopt this as a government policy objective; undertake corporate governance reforms to contain all the excesses of top income compensation and encourage the setting of links of top pay to median incomes in the private sector; ensure the access of all to quality education and training systems; restore progressivity in the tax system; ensure that economic performance is judged by wider criteria than GDP per head. (TUAC 2013)

The unions took encouragement from the apparent shift in thinking by G20 labor and employment minsters, who at their July 2013 meeting had committed to move forward by:

…Implementing labour market and social investment policies that support aggregate demand and reduce inequality, such as broad based increases in productivity, targeted social protection, appropriately set minimum wages with respect to national wage setting systems, national collective bargaining arrangements and other policies to reinforce the links between productivity, wages, and employment. (G20 2013)

Economists such as Thomas Piketty and the late Anthony Atkinson were invited to make presentations at OECD meetings, and their work began to be used by the OECD to prepare the ground for shifting policy to more progressive taxation and to revisit analysis of labor market institutions. Atkinson (2015), focusing on the U.K., offered 15 proposals to reduce inequality; he recommended establishing a more favorable balance of power between labor and capital, allied with progressive taxation to help those in the lower deciles of the income distribution. He also promoted wider trade union representation to relink productivity growth and incomes.

The 2018 revision of the Jobs Strategy

A new effort to revise the OECD Jobs Strategy, begun in 2016, was influenced by both the popular backlash against globalization and the emerging debate about the future of work in light of technological change, the growth of nonregular work in many countries, and the rise of the “gig” economy. While the OECD appeared to initially flirt with the idea that gig work was a growing and natural progression in the work relationship brought on by technology, it began to take a stand that existing labor regulations must apply to “platform” work as well (O’Farrell 2016).

The 2017 OECD Employment Outlook, in summarizing the 2006 work revising the Jobs Strategy, argued that “The populist backlash against globalisation fundamentally challenges employment policy” (OECD 2017a). It noted:

The perception that the international economic system is rigged clearly challenges the democratic legitimacy of current policies and thus needs to be taken seriously. It also challenges the policy advice offered by international organisations like the OECD, which has long emphasised the economic benefits of global integration, but only recently adopted an inclusive growth approach that pays due attention to the distribution of those benefits across the population. (OECD 2017a, 9)

The OECD began detailed work on changes to collective bargaining systems and their impact on labor market outcomes. The 2017 Employment Outlook had included a taxonomy of collective bargaining systems based on three criteria: the formal level of bargaining (company, sector, or national); the space left for lower-level agreement; and the degree of coordination. It acknowledged that policy reforms had contributed to decentralization of bargaining and had “tested the system.” The outlook laid out the areas of work that needed to be carried out to inform the wider Jobs Strategy revision.

The initial narrative of the revised Jobs Strategy containing the key policy messages and analysis was adopted at the OECD Ministerial Council meeting in June 2018, and the full report was launched in December 2018 (OECD 2018). The OECD claimed that the Jobs Strategy represented a shift in thinking and policy from both the 2006 version and even more so from the original 1994 study. The TUAC backed this view, commenting:

The Narrative of the revised Job Strategy…takes a much broader approach than was previously the case. Policy objectives are no longer limited to the quantity of jobs but have been expanded to include their quality and inclusiveness. Moreover, the narrative recognises there is no necessary trade-off between the quantity of jobs on the one hand and their quality and inclusiveness on the other. Importantly, the OECD now concedes that flexibility has been over-rated in view of “new evidence that shows that countries with policies and institutions that promote job quality, job quantity [maximum employment rather than minimum unemployment] and greater inclusiveness perform better than countries where the focus of policy is predominantly on enhancing market flexibility.”

The revision also clearly rejects the argument that collective bargaining defends the interest of “insiders against outsiders” in the labor market—a long-held tenet of the OECD Economics Department (OECD 2018, 147).

However, Janssen (2019) notes that in also claiming that “flexibility in product and labour markets is essential to create high quality jobs in an ever more dynamic environment,” the strategy is offering two distinct narratives on labor market flexibility running in parallel. McBride and Watson (2019) wrote off the revision as “an attempt to restore legitimacy through ideological positioning rather than fundamental change in basic strategy.” This is hardly surprising in an international organization in which reaching agreement on change between different countries is a complicated process. TUAC’s evaluation prioritized the importance of the implementation process. In any case, the positive role ascribed to collective bargaining and the admitting of greater ambiguity in labor market deregulation and flexibility amounted to more than a cosmetic change.

This assessment was reinforced when in 2019 the OECD published its most comprehensive work on collective bargaining in 30 years under the title, “Negotiating Our Way Up: Collective Bargaining in a Changing World of Work” (OECD 2019). The central conclusions of the report were: (1) coordination in wage bargaining is a key ingredient for good labor market performance; (2) collective bargaining systems and workers’ voice arrangements also matter for job quality; and (3) collective bargaining and workers’ voice play an important role in preventing inequalities in a changing world of work, but they need to adapt.

Most importantly, the OECD came down firmly on the side of the need for countervailing power in the workplace:

Whether considering key issues such as wage inequality, job quality, workplace adaptation to the use of new technologies, or support for workers displaced by shifts in industries, collective bargaining and workers’ voice can complement public policies to produce tailored and balanced solutions. The alternatives to collective bargaining are often either state regulation or no bargaining at all since individual bargaining is not always a realistic option as many employees are not in a situation to effectively negotiate their terms of employment with their employer. (OECD 2019, 13)

Postscript: Labor markets and the Covid-19 pandemic

In the months following the publication of these OECD reports, the Covid-19 pandemic again raised the importance of reducing inequality in OECD countries and beyond. The pandemic and subsequent containment measures have had differential impacts on income groups, age groups, ethnic groups, and social groups. The economic prospects of young people, ethnic minorities, and women—the groups more likely to be employed in service sectors that have been most affected by closures, but which are also overrepresented in insecure and unprotected work (Reitsma et al. 2021; Rogers et al. 2020; Goldman et al. 2021; Williamson et al. 2020; Out et al. 2020)—have been hardest hit in most countries. Sectors with activities that allow teleworking and so are more likely to be performed from home have seen a smaller reduction in employment. Low-income workers are less able to work from home than high-income workers.

An IMF study comparing the distributional impact in the United States of the global financial crisis and the pandemic recession found that young workers, less-qualified workers, and low-income earners were hit hardest in both recessions, but women and Hispanic workers were more severely affected in the pandemic. The concentration of female employment in service sectors together with the difficulty of managing child care when schools and other facilities are closed have resulted in a disproportional impact on women (Shibata 2020).

A study in France (Inserm 2020) found that those living in dense and cramped accommodations have been infected disproportionately during the pandemic, while those in low-paid vulnerable jobs are most at risk from infection and have suffered most from economic hardship. A study in the U.K. of the impact of what was then expected to be the ending of jobs support schemes found that young and ethnic minority workers were twice as likely to lose their jobs when the furlough scheme came to an end (Crossley, Fisher, and Low 2021; Brewer et al. 2020).

Emerging and developing countries have also seen a rise in inequality because of the pandemic. IMF authors concluded that “the estimated effect from COVID-19 on the income distribution is much larger than that of past pandemics. It also provides evidence that the gains for emerging market economies and low-income developing countries achieved since the global financial crisis could be reversed” (Cugat and Narita 2020). The World Bank has warned of “new poverty” appearing in middle-income countries (World Bank 2020).

The growth of nonstandard work and platform work prior to the pandemic had already increased inequality, and since most platform workers do not have employment protection, unemployment benefits, or paid sick leave, economic risk was shifted onto their shoulders during the pandemic. Also, during the pandemic customers for platform workers’ services dried up overnight, and the workers found themselves without income or employment. In a global survey by an online search platform for app-based jobs, over half of the platform workers surveyed said they had lost their jobs, and more than a quarter had seen their hours cut in the first months of the pandemic (AppJobs Institute 2020).

The OECD has noted that the pandemic revealed the shortcomings of safety nets (OECD 2020), in essence admitting that nonregular work leaves many people out of reach of vital social protections for income and health. While before OECD reports made almost indiscriminate calls for lowering labor standards to increase labor market flexibility for employers, they now caution that irregular work can be a danger.

As in the financial crisis, it is already clear that youth are big losers economically in the pandemic. In addition to the disruption of education in most countries, unemployment among first-time job seekers has soared. There is much evidence that initial job market experience influences earnings capacity in the long term, and hence deep recessions lead to scarring effects on individuals and subsequently on economies and societies.

Meanwhile, the wealth of billionaires has increased during the pandemic in all the major economies (Jones and Romei 2020), and U.S. internet companies have seen their stock values rise rapidly. Profits at Amazon tripled to $6.3 billion since the beginning of the pandemic, while turnover at Amazon increased by 37% in the third quarter of 2020.

As economies reopen and pressure for a return to normal increases, there is a danger that a business-as-usual approach will return. But the pandemic has brought into sharp focus the broader importance of social justice. Rising income inequality has fragmented societies, and the countries and communities weathering these trends most effectively will be those with greater social cohesion and where policies are designed to be fair and seen to be fair. The new thinking exhibited in the recent OECD reports represents an opportunity for a positive shift in policy if translated into policy action.

About the authors

John Evans was until 2017 general secretary of the Trade Union Advisory Committee to the OECD, a post he held for more than 30 years. From 2012 until 2017 he combined his TUAC post with that of chief economist of the International Trade Union Confederation. He is currently a visiting fellow at the University of Greenwich in the U.K. and an academic visitor at St Peter’s College at the University of Oxford. He is working on a book project on the shifting views of economists toward income inequality and labor market institutions.

William Spriggs is a professor in, and former chair of, the Department of Economics at Howard University and serves as chief economist to the AFL-CIO. In his role with the AFL-CIO, he also chairs the Economic Policy Working Group for the Trade Union Advisory Committee to the OECD, and serves on the board of the National Bureau of Economic Research. He is currently president-elect of the Labor and Employment Relations Association and vice chair of the board of MDC Inc. (Durham, N.C.). He serves on the advisory boards of WorkRise (of the Urban Institute) and the Opportunity and Inclusive Growth Institute of the Federal Reserve Bank of Minneapolis. From 2009 to 2012, he was assistant secretary for the Office of Policy at the U.S. Department of Labor.

Acknowledgments

The authors are grateful to David Coats, Larry Mishel, Stephen Pursey, and Kari Tapiola for comments on earlier drafts and indebted to Dean Baker and Andrew Watt, who undertook the peer review of the draft.

Notes

1. For an analysis of the country-specific recommendations from 1994 to 1998, see Casey 2004.

2. Unpublished Trade Union Advisory Committee mimeo to the OECD.

3. A common measure of income inequality that ranges from 0 when income is the same for all to 1 when all income goes to one person.

4. See, for example, International Trade Union Confederation, Equal Times, April 2013.

References

Aleksynska, Mariya. 2014. “Deregulating Labour Markets: How Robust Is the Analysis of OECD Recent IMF Working Papers?” ILO Conditions of Work and Employment Series no. 47.

AppJobs Institute. 2020. “How Does the COVID-19 Outbreak Affect the Life of Freelancers & Gig Workers?” Quoted in OECD 2020.

Atkinson, Anthony B. 2015. “Inequality: What Can Be Done?” Harvard University Press.

Autor, David H. 2003. ”Outsourcing at Will: The Contribution of Unjust Dismissal Doctrine to the Growth of Employment Outsourcing.” Journal of Labor Economics 21: 1–41.

Avdagic, Sabina. 2015. “Does Deregulation Work? Reassessing the Unemployment Effects of Employment Protection.” British Journal of Industrial Relations, 2014, 53, no. 1 (March): 6–26.

Baker, Dean, Andrew Glyn, David Howell, and John Schmitt. 2004. Fighting Unemployment: The Limits of Free Market Orthodoxy. Oxford University Press.

Barkin, Solomon. 1987. “The Flexibility Debate in Western Europe: The Current Drive to Restore Management’s Rights over Personnel and Wages.” Relations Industrielles 42, no. 1: 12–43.

Bassanini, Andrea, and Roland Duval. 2006. “Employment Patterns in OECD Countries: Reassessing the Role of Policies and Institutions.” Organization for Economic Cooperation and Development, Economics Department Working Paper No. 486.

Belot, Michèle, Jan Boone, and Jan van Ours. 2007. “Welfare‐Improving Employment Protection.” Economica 74, no. 295: 381–96.

Berg, Andrew, and Jonathon Ostry. 2011. “Inequality and Unsustainable Growth: Two Sides of the Same Coin.” IMF Staff Discussion Note, April.

Blanchard, Olivier, Florence Jaumotte, and Prakash Loungani. 2013. “Labor Market Policies and IMF Advice in Advanced Economies During the Great Recession.” IMF Staff Discussion Note, March.

Blanchflower, David, and Richard Freeman. 1993. “Did the Thatcher Reforms Change British Labor Performance?” National Bureau of Economic Research Working Paper No. 4384.

Brewer, Mike, et al. 2020. “Jobs, Jobs, Jobs: Evaluating the Effects of the Current Economic Crisis on the UK Labour Market.” Resolution Foundation, London.

Casey, Bernard. 2004. “The OECD Jobs Strategy and the European Employment Strategy: Two Views of the Labour Market and the Welfare State.” European Journal of Industrial Relations 10, no. 3: 229–351.

Cingano, Federico. 2014. “Trends in Income Inequality and Its Impact on Economic Growth.” OECD Employment, Social, and Migration Working Paper No. 163.

Coats, David, ed. 2011. “Exiting from the Crisis: Towards a Model of More Equitable and Sustainable Growth: Report of a Trade Union Task Force.” European Trade Union Institute.

Coats, David. 2018. Fragments in the Ruins: The Renewal of Social Democracy. London, England: Policy Network.

Crossley, Thomas F., Paul Fisher, and Hamish Low. 2021. “The Heterogeneous and Regressive Consequences of COVID-19: Evidence from High Quality Panel Data.” Journal of Public Economics 193 (January).

Cugat, Gabriela, and Futoshi Narita. 2020. “How COVID-19 Will Increase Inequality in Emerging Markets and Developing Economies.” IMF Blog, October 29.

Deaken, Simon, Jonas Malmberg, and Prabirjit Sarkar. 2014. “How Do Labour Laws Affect Unemployment and the Labour Share of National Income? The Experience of Six OECD Countries, 1970–2010.” International Labour Review 153, no. 1.

European Commission. 1999. OECD/DEELSA seminar, “Wages and Employment.”

Freeman, Richard. 2005. “Labor Market Institutions Without Blinders: The Debate over Flexibility and Labor Market Performance.” National Bureau of Economic Research Working Paper No. 11286.

Friedman, Milton. 1968. “The Role of Monetary Policy.” American Economic Review 58, no. 1: 1–17.

G20. 2009. “Leaders’ Statement: The Pittsburgh Summit.” Pittsburgh, Pa., September 24–25.

G20. 2013. “Labor and Employment and Finance Ministers’ Communiqué.” Moscow, July 19.

G20. 2014. “G20 Leaders’ Communiqué.” Brisbane, Australia, November 16.

G20. 2015. “Labour and Employment Ministers’ Declaration.” Annex 1. Ankara, Turkey, September 4.

Goldman, Noreen, et al. 2021. “Racial and Ethnic Differentials in COVID-19-Related Job Exposures by Occupational Standing in the US.” Preprint paper.

Gurria, Angel. 2017. “Speech to the British Confederation of Industry Conference,” November 6.

Gurria, Angel. 2020. “Remarks at 60th Anniversary of the OECD.” Press release. Organization for Economic Cooperation and Development, December 14.

Guschanski, Alexander, and Ozlem Onaran. 2017. “The Political Economy of Income Distribution: Industry Level Evidence from 14 OECD Countries.” Greenwich Papers in Political Economy GP ERC 51.

Heimberger, Philipp. 2020. “Does Employment Protection Affect Unemployment? A Meta-Analysis.” Oxford Economic Papers, November 28.

Hiltzik, Michael. 2015. “IMF Agrees: Decline of Union Power Has Increased Income Inequality.” Los Angeles Times, March 25.

Hoyt, Eric. 2018. “Constraining Labor’s ‘Double Freedom’: Revisiting the Impact of Wrongful Discharge Laws on Labor Markets, 1979-2014.” Ph.D. Dissertation, University of Massachusetts, Amherst.

ILO (International Labour Organization). 2011. “ILO World of Work Report: Making Markets Work for Jobs.”

IMF (International Monetary Fund). 2003. “Unemployment and Labour Market Institutions: Why Reforms Pay Off.” IMF World Economic Outlook, April.

IMF (International Monetary Fund). 2017. “World Economic Outlook,” Chapter 3, April.

Inserm. 2020. “Premiers Résultats des Enquêtes de Santé Publique de l’Inserm sur la Covid-19: Facteurs de Risque Individuels et Sociaux.“ Press release, October 9.

Jackson, Andrew. 2007. “Crafting the Conventional Economic Wisdom: The OECD and the Canadian Policy Process.” In Rianne Mahon and Stephen McBride, eds., OECD and Transnational Governance. Vancouver, Canada: UBC Press.

Janssen, Ronald. 2019. “The Revised OECD Jobs Strategy and Labour Market Flexibility: A Double Handed Narrative.” European Trade Union Institute Transfer 25: 221–27.

Jaumotte, Florence, and Carolina Osorio Buitron. 2015. “Inequality and Labor Market Institutions.” IMF Staff Discussion Note, July.

Jones, Sam, and Valentina Romei. 2020. “Pandemic Makes World’s Billionaires—and Their Advisers—Richer.” Financial Times, October 23.

Kaldor, Nicholas. 1955. “Alternative Theories of Distribution.” Review of Economic Studies 23, no. 2: 83–100.

Keely, Brian. 2015. “Income Inequality: The Gap Between Rich and Poor.” OECD Insights.

Krueger, Alan B. 2012. “The Rise and Consequences of Inequality in the United States.” Speech to the Center for American Progress, January 12.

Krugman, Paul. 2010. “Oops at the OECD.” New York Times, September 10.

Kugler, Adriana, and Gilles Saint-Paul. 2004. “How Do Firing Costs Affect Worker Flows in a World with Adverse Selection?” Journal of Labor Economics 22: 553–84.

Lagarde, Christine. 2012. “Speech to the 2012 Annual Meetings of the IMF and World Bank.”

Lansley, Stewart, and Howard Reed. 2010. “The Red Tape Delusion: Why Labour Market Deregulation Won’t Solve the Jobs Crisis.” Trades Union Congress.

Lucas, Robert. 2003. “The Industrial Revolution: Past and Future.” Annual Report of the Federal Reserve Bank of Minneapolis, 5–20.

McBride, Stephen, and James Watson. 2019. “Reviewing the 2018 OECD Jobs Strategy: Anything New Under the Sun?” European Trade Union Institute Transfer 25: 149–63.

Milanovic, Branko. 2016. “Global Inequality: A New Approach for the Age of Globalization.” Cambridge, Mass.: Belknap Press of Harvard University.

Miles, Thomas J. 2000. “Common Law Exceptions to Employment at Will and U.S. Labor Markets.” Journal of Law, Economics and Organizations 16: 74-101.

Mitchell, Daniel J. 2011. “Using Your Tax Dollars to Push for Class Warfare.” New York Post, December 17.

Nickell, Stephen. 2017. “Remarks to High-Level Policy Forum on the New OECD Jobs Strategy.” Berlin, Germany, June 17.

OECD (Organization for Economic Cooperation and Development). 1983. “Positive Adjustment Policies: Managing Structural Change.”

OECD (Organization for Economic Cooperation and Development). 1992. “Labour Market Policies for the 1990s: Communique of the Meeting of Employment and Labour Affairs Committee Meeting at Ministerial Level.”

OECD (Organization for Economic Cooperation and Development). 1994a. “The OECD Jobs Study: Facts, Analysis, and Strategy.”

OECD (Organization for Economic Cooperation and Development). 1994b. “The OECD Jobs Study: Evidence and Explanations.”

OECD (Organization for Economic Cooperation and Development). 2004. “Employment Outlook 2004.”

OECD (Organization for Economic Cooperation and Development). 2006. “Employment Outlook 2006.”

OECD (Organization for Economic Cooperation and Development). 2008. “Growing Unequal? Income Distribution and Poverty in OECD Countries.”

OECD (Organization for Economic Cooperation and Development). 2010. “Economic Survey of Slovakia.”

OECD (Organization for Economic Cooperation and Development). 2011. “Divided We Stand: Why Inequality Keeps on Rising.”

OECD (Organization for Economic Cooperation and Development). 2014a. “Focus on Top Incomes and Taxation in OECD Countries: Was the Crisis a Game Changer?”

OECD (Organization for Economic Cooperation and Development). 2014b. “NAEC Executive Summary for the OECD Ministerial Council Meeting.”

OECD (Organization for Economic Cooperation and Development). 2017a. “Employment Outlook 2017.”

OECD (Organization for Economic Cooperation and Development). 2017b. “Opportunities for All: A Framework for Policy Action on Inclusive Growth.”

OECD (Organization for Economic Cooperation and Development). 2018. “Good Jobs for All in a Changing World of Work: The OECD Jobs Strategy.”

OECD (Organization for Economic Cooperation and Development). 2019. “Negotiating Our Way Up: Collective Bargaining in a Changing World of Work.”

OECD (Organization for Economic Cooperation and Development). 2020. “Employment Outlook 2020.”

O’Farrell, Rory. 2016. “The Gig Economy Will Not Abolish Working 9 to 5.” Medium, November 22.

Out, Akaninyene, et al. 2020. “One County, Two Crises: What Covid-19 Reveals About Health Inequalities Among BAME Communities in the United Kingdom and the Sustainability of Its Health System.” International Journal for Equity in Health 19, no. 189.

Piketty, Thomas. 2014. Capital in the Twenty-First Century. Harvard University Press.

Pursey, Stephen. 2021. “G20 Structural Reforms and OECD Jobs Strategies.” International Labour Office. Mimeo.

Reitsma, Marissa B., et al. 2021. “ Racial/Ethnic Disparities in COVID-19 Exposure Risk, Testing, and Cases at the Subcounty Level in California.” Health Affairs 40, no. 6: 870–8.

Rogers, Tiana N., et al. 2020. “Racial Disparities in COVID-19 Mortality Among Essential Workers in the United States.” World Medical Health Policy, August 5.

Shibata, Ippei. 2020. “The Distributional Impact of Recessions: The Global Financial Crisis and the Pandemic Recession.” International Monetary Fund Working Paper WP/20/96.

Stiglitz, Joseph. 2012. “The Price of Inequality.” Allen Lane.

TUAC (Trade Union Advisory Committee). 2013. “Growing More Equal: A Policy Agenda for Reducing Inequality and Promoting Inclusive Growth.” Paper for Consultations with the OECD Liaison Committee.

TUAC (Trade Union Advisory Committee). 2018. “Assessment of the OECD’s Revised Job Strategy: Good on Objectives, Ambiguous on Policy.”

Trumka, Richard. 2011. Speech to Launch Event for “Divided We Stand,” OECD (author’s notes).

Vergeer, Robert, and Alfred Kleinknecht. 2011. “The Impact of Labour Market Deregulation on Productivity: A Panel Data Analysis of 19 OECD Countries (1960-2004).” Journal of Post-Keynesian Economics, Winter 2010–2011.

Watt, Andrew. 2006. “Assessing the Reassessment of the OECD Jobs Strategy: Eppur si Muove?” Etui.

Williamson, Elizabeth, et al. 2020. “OpenSAFELY: Factors Associated with COVID-19-Related Hospital Deaths in Linked Electronic Health Records of 17 Million Adult NHS Patients.” Preprint paper.

World Bank. 2012. World Development Report 2013: Jobs, 256–91.

World Bank. 2020. “Global Action Urgently Needed to Halt Historic Threats to Poverty Reduction.”

Embedded in U.S. employment law is the presumption that the employer and the employee have equal power, that either can as readily walk away from an employment relationship as the other, and that the employee can as easily find an equivalent job as the employer can find an equivalent worker. In other words, each is free to contract—to agree to an employment arrangement—on equal terms, without constraint or coercion. But for workers to be on a level playing field with employers requires an economy at full employment, where everyone who wants a job or even just a decent employment situation can find one. Only in this circumstance is an employee free to contract without constraint or coercion.

Yet the economy is rarely at full employment, and even on the rare occasions that it is large segments of the workforce still face substantial unemployment and difficulty finding quality jobs. Blacks, Hispanics, and those without college degrees endure a permanent recession.

This paper uses measures of unemployment from the Bureau of Labor Statistics and reviews important recent work to examine trends in overall labor market slack and the unemployment experiences of specific demographic groups, as well as to illustrate the impact of unemployment on worker behavior (i.e., quitting, switching jobs), employer behavior (i.e., recruiting intensity), and labor market outcomes such as wage growth.

The length of time workers remain unemployed during spells of unemployment and the likelihood that they will quit a job or switch jobs illustrate their diminished options when unemployment is high:

• In recent downturns the duration of unemployment and the share of the unemployed who were long-term unemployed increased as the economy moved into recession. Obviously, the prospect of quitting and becoming unemployed becomes a much more costly prospect when the economy is not at full employment.
• Also in recent downturns, the share of workers quitting declined substantially as unemployment rose, while the share climbed as the economy recovered. Thus, a worker’s ability to quit and an employer’s ability to fill a vacancy are not independent of the unemployment situation.
• Most workers find a new job by directly switching employers, rather than finding a new job after becoming unemployed or leaving the labor force, and data show that a higher unemployment environment adversely affects workers’ ability and willingness to switch employers. Compounding this constraint is the fact that switching jobs is an essential component of receiving higher wages.

Unemployment also changes employer behavior. When unemployment is higher employers are able to fill job vacancies more quickly and with less effort; their recruitment activities—choice of methods, expenditures on help-wanted ads, screening of job applicants—differ according to the level of unemployment. Moreover, when unemployment is elevated and employers can more readily find qualified workers, employers opportunistically raise the expected credentials for new hires. In the 2007–2010 downturn, for instance, employers increasingly required a bachelor’s degree for physician assistant jobs but retreated from this requirement as unemployment fell.

Finally, there is substantial evidence that higher unemployment lowers wage growth, especially for low-and moderate-wage workers and for Black and Hispanic workers. One indicator of the overall shift in power against workers is that in the last recovery from 2009 to 2019 boosting wages required lower levels of unemployment than before.

Workers in the aggregate experience full employment only occasionally, and millions of workers never experience it at all, ever. The presence of excessive unemployment—beyond full employment—tilts the power balance toward employers. Just acknowledging high unemployment leads one to recognize that in many, if not most, circumstances employers can far more readily replace a worker than a worker can find a comparable job. To believe otherwise is to live in a world without access to windows or newspapers, and it is curious and unsettling that claims of freedom of contract have been made when there was, or had recently been, very high unemployment. Simply acknowledging the persistent absence of full employment for many workers renders the freedom-of-contract framework a flawed basis for assessing employment relationships and arrangements.

Introduction

Embedded in U.S. employment law is the presumption that the employer and the employee have equal power, that either can as readily walk away from an employment relationship as the other, and that the employee can as easily find an equivalent job as the employer can find an equivalent worker. In other words, each is free to contract—to agree to an employment arrangement—on equal terms, without constraint or coercion. This assumption of equal power is pervasive but also insidious, and it undercuts our ability to have adequate statutory and common law workplace protections (Bagenstos 2020).

This is an assumption similar to the one economists make when they assume “full employment.” In the economics case, however, one can relax an assumption and model how outcomes change. In the legal case the law applies at all times. The freedom-of-contract formulation in the law is oddly applied to employer unilateral changes to contract terms for already-employed workers, as if there is a new contract each day.

Not only do common sense and observation tell us that employees can rarely walk away from their jobs as readily as employers can find replacements, but documenting this fact is easily accomplished by looking at employer and employee behavior. As business cycles proceed and as economies boom and bust, unemployment rises and falls. The economy is rarely at full employment, and even on the rare occasions that it is large segments of the workforce still face substantial unemployment and difficulty finding quality jobs. Blacks, Hispanics, and those without college degrees endure a permanent recession. Less-than-full employment manifests itself in employee behavior through lower quit rates, fewer transitions to new jobs, and longer spells of unemployment between jobs.

Unemployment also changes employer behavior. When unemployment is higher employers are able to fill job vacancies more quickly and with less effort; their recruitment activities—choice of methods, expenditures on help-wanted ads, screening of job applicants—differ according to the level of unemployment, and employers raise hiring standards when unemployment is higher. Also, when unemployment is elevated and employers can more readily find qualified workers, employers opportunistically raise the expected credentials for new hires. In the 2007–2010 downturn, for instance, employers increasingly required a bachelor’s degree for physician assistant jobs but retreated from this requirement as unemployment fell.

Last, there is substantial evidence that higher unemployment lowers wage growth, especially for low-and moderate-wage workers and for Black and Hispanic workers. One indicator of the overall shift in power against workers is that in the last recovery from 2009 to 2019 boosting wages required lower levels of unemployment than before.

There are many other reasons besides the business cycle and higher unemployment that workers, individually, have less power than their employers. For example, individual workers as a rule have little or no wealth to fall back on, they may be locked into an employer for their health coverage, they may have limited transportation options, and child care responsibilities may constrict their scheduling options (Edwards 2022). There is also evidence from the emerging literature on monopsony that quitting is insufficiently powerful to restrain employer exploitation, as evidenced by the fact that even when employers reduce wages only a small share of employees actually quit (Naidu and Carr 2022). Nevertheless, the simple acknowledgment of the difficulty in finding a job when unemployment exceeds full employment is sufficient to show how otherworldly is the assumption of equal power and the claim that freedom of contract produces optimal outcomes that regulations or institutions should not disturb.

What is the freedom-of-contract framework?

In the freedom-of-contract framework, employers and employees have equal power, and their negotiated arrangements are optimal and should not be altered or regulated by external forces such as government-set labor standards or unions. But this presumption of equal power willfully ignores the fact that workers rarely enjoy full employment, meaning that employers enjoy plentiful access to willing new workers while employees face more difficulties and costs in finding alternative comparable employment. By itself, the absence of full employment creates a power asymmetry between employers and employees.

Bagenstos (2020) explained the freedom-of-contract and employer-employee equality assumptions underlying the at-will employment doctrine:

[The at-will] doctrine authorizes both employers and employees to terminate the relationship at any time. The Supreme Court expressly relied on this supposed equality when it gave constitutional significance to at-will employment in its Lochner-era decisions. In Adair, Justice Harlan wrote that “the right of the employe to quit the service of the employer, for whatever reason, is the same as the right of the employer, for whatever reason, to dispense with the services of such employe.”1 He went on to say that “the employer and the employe have equality of right, and any legislation that disturbs that equality is an arbitrary interference with the liberty of contract which no government can legally justify in a free land.”2 And he declared, “it cannot be…that an employer is under any legal obligation, against his will, to retain an employe in his personal service any more than an employe can be compelled, against his will, to remain in the personal service of another.”3

Richard Epstein (1984), in his iconic defense of the at-will doctrine, also relied heavily on the freedom-of-contract framework and the presumed equality between employer and employee:

One manifestation of that position was the prominent place that the common law, especially as it developed in the nineteenth century, gave to the contract at will. The basic position was well set out in an oft-quoted passage from Payne v. Western & Atlantic Railroad:

[M]en must be left, without interference to buy and sell where they please, and to discharge or retain employees at will for good cause or for no cause, or even for bad cause without thereby being guilty of an unlawful act per se. It is a right which an employe may exercise in the same way, to the same extent, for the same cause or want of cause as the employer. (pp. 947–8)

Further, Epstein wrote, under an at-will arrangement:

The employer is free to demand whatever he wants of the employee, who in turn is free to withdraw for good reason, bad reason, or no reason at all. (p. 966)

The freedom-of-contract framework was central to the Supreme Court’s majority opinion in 2018 in Epic Systems,4 which determined that the Federal Arbitration Act providing for individualized arbitration proceedings must be enforced, and that neither the Arbitration Act’s saving clause nor the National Labor Relations Act suggests otherwise.5

Justice Gorsuch opened the majority opinion with these questions:

Should employees and employers be allowed to agree that any disputes between them will be resolved through one-on-one arbitration? Or should employees always be permitted to bring their claims in class or collective ac­tions, no matter what they agreed with their employers?

Justice Gorsuch presumes there is no basis for interfering with any seemingly voluntary employer-employee agreement and that the Federal Arbitration Act predominates over the National Labor Relations Act’s (NLRA) guarantee of employee collective actions (such as class actions in private arbitration proceedings), a guarantee that was affirmed by the National Labor Relations Board in 2012.

In her dissent, Justice Ginsburg noted:

To explain why the Court’s decision is egregiously wrong, I first refer to the extreme imbalance once preva­lent in our Nation’s workplaces, and Congress’ aim in the NLGA [Norris-LaGuardia Act] and the NLRA to place employers and employees on a more equal footing [emphasis added].

The Epic Systems case highlighted the disagreement within the court about the freedom-of-contract framework and the assumed equality of power between employers and employees. The majority accepted the freedom-of-contract framework even though employers imposed the forced individual arbitration agreements on employees long after they had negotiated for, and accepted, employment.6

Notably absent from any of these arguments and cases relying on the freedom-of-contract framework is any consideration of the business cycle or the fact that workers, because they typically face unemployment higher than that prevailing at full employment, are rarely on equal footing with their employers.

Indeed, unemployment was or had been excessively high at the time of these key court cases and writings. In the three years leading up to the Adair decision (1906–1908), unemployment in manufacturing, transportation, building trades, and mining (the only historic data available) rose from 5.9% to 6.9% to 16.4%.7 In 1984, the year of publication of Epstein’s famous analysis, the economy was emerging from the greatest economic downturn (at that time) since the Great Depression, with unemployment falling from 9.7% in 1982 to 9.6% in 1983 and to 7.5% in 1984 (still far above full employment).8 The 2018 opinion in Epic Systems followed the recovery from the financial crisis of 2007-2009 and the lowering of unemployment from a peak of 10% in October 2009 to just 3.9% in the summer of 2018.9

This willful ignoring of unemployment trends is essential to the presumption, perhaps the pretense, that employee-employer employment agreements are optimal, characterized as occurring between equal parties, one as willing as the other to depart the arrangement. Yet, as documented below, excess unemployment severely weakens the relative bargaining position of workers.

What is full employment?

This paper argues that the economy is rarely at full employment and sometimes never so for large segments of the workforce. To measure whether the economy is above, below, or at full employment requires an operational, measurable definition of the concept. This is a challenge, as there is no consensus on the definition of full employment, and the unemployment rate considered full employment shifts over time. The benchmark used in our analysis is setting full employment at 5%, slightly below

One approach is to categorize unemployment by reason, separately identifying those unemployed due to cyclical, structural, or frictional unemployment. As defined by the Congressional Research Service (CRS 2020), cyclical unemployment is the extra unemployment resulting from the ups and downs of the business cycle; structural unemployment is “unemployment resulting from a mismatch of skills or interest between workers and the jobs available,” due to trade, technological change, or shifts in consumer preferences; frictional unemployment is “short-term unemployment due to job searching or transition.” Full employment, in this scheme, occurs when “the economy is operating at its full potential, cyclical unemployment is zero and the unemployment rate is roughly equal to the sum of structural and frictional unemployment” (CRS 2020). However, estimates of the amount of cyclical, structural, and frictional unemployment are not readily available to operationalize this definition.

Similarly, the Full Employment Action Council, the advocacy group that campaigned to pass the Humphrey-Hawkins Full Employment Act in the 1970s, describes full employment as the absence of involuntary unemployment—everyone who wants a job can get one.

The freedom-of-contract framework presumes an equality between employers’ concerns over vacancies and workers’ fears of becoming unemployed. In this light, a balanced labor market is one where vacancies equal the number of unemployed and where outside options are comparable: Workers and firms, respectively, have equal and ready access to a replacement job or worker. This framework is comparable to an analysis of the standard macroeconomic model of unemployment.10 Certain measures of vacancies (i.e., job openings) and unemployment allow us to gauge when the two are in balance. Davis, Faberman, and Haltiwanger (2013), who provide such data (including data made available on their website) from 2001 through June 2017, find that in no month in that period did job openings exceed unemployment; on average, there were 2.7 unemployed for every opening. Even in the periods of the lowest unemployment (the first halves of 2001, 2007, and 2017), when unemployment was 4.5% or less, there were 24-49% more unemployed than job openings. This evidence suggests that the full employment rate needed to balance openings and unemployment is certainly less than 4.5%.

Much economic analysis over the last few decades has chosen as a definition of full employment the non-accelerating inflation rate of unemployment (NAIRU), or the rate below which inflation will begin to accelerate.11 There have been many critiques of the NAIRU (Galbraith 1997; Staiger, Stock, and Watson 1997; Baker 2000; Bernstein and Baker 2013a), and inflation has not accelerated when the unemployment rate has fallen (sometimes far) below it (Crump et al. 2019). For instance, unemployment averaged 3.7% in the last half of 2019, a rate below the Congressional Budget Office (CBO)’s estimate of NAIRU of 4.5% for that period, without any sign of accelerating inflation. Crump et al. (2019) “estimate that the natural rate of unemployment was about 4.0 percent toward the end of 2018.” Bernstein and Baker (2013b) detail the lack of inflation acceleration in the late 1990s boom despite unemployment falling to 4.0% in 2000, far below the 5.2% NAIRU estimated by CBO for that year. CBO, which provides an estimate of NAIRU back to 1949 and uses it to estimate potential output and the corresponding fiscal consequence of departures from full capacity, estimates that the NAIRU averaged about 5.3% over the entire 1979–2019 period and just 4.9% from 2000 to 2019.

For the 1979–2019 period the analysis below employs a somewhat arbitrary 5.0% as the full employment benchmark, a bit below the 5.3% estimated by CBO though close to the 4.9% of 2000–2019. There are three reasons for choosing 5.0% rather than 5.3%. One is simplicity. Second, experience shows that the NAIRU overstates the unemployment level corresponding to actual acceleration of inflation. Third, as discussed above, the unemployment rate that equates vacancies and unemployment is substantially below 5.0%. Choosing a 5% full employment benchmark is a conservative choice. Regardless, the choice of 5.0% rather than 5.3% does not materially affect any of our conclusions: The economy is rarely at full employment, and large segments of the workforce never experience full employment.

It should be acknowledged that the unemployment rate does not fully capture the many dimensions and negative consequences of underemployment: workers working part time but wanting full-time work; workers who have stopped looking for work and left the labor force; and workers employed at jobs for which they are overqualified. At any given rate of unemployment, there are many more workers suffering from various types of underemployment than there are unemployed workers. The selection of a particular unemployment rate as the full employment benchmark may be appropriate for designating periods of full employment and slack, though it does understate the number of workers adversely affected by slack conditions.

The employee side of higher unemployment

The share of the labor force that is unemployed is a good indicator of the labor market climate or labor market slack. Unemployment, as officially measured by the Bureau of Labor Statistics (BLS), reflects “people who meet all of the following criteria: were not employed during the survey reference week; were available for work (except for temporary illness); had made a specific, active effort to find employment sometime during the 4-week period ending with the survey reference week” (BLS 2018, 3-4). This is considered an activity measure, since unemployment is equated with not having a job and actively looking for work. The unemployment rate is measured as, “The number of unemployed people as a percentage of the labor force,” and the labor force is the sum of the employed and the unemployed (BLS 2018, 4).

This paper relies on the BLS measures of unemployment to examine trends in overall labor market slack and the unemployment experiences of specific demographic groups, as well as to illustrate the impact of unemployment on worker behavior (i.e., quitting, switching jobs), employer behavior (i.e., recruiting intensity), and labor market outcomes such as wage growth. But, as noted above, unemployment does not capture the full extent of underemployment in the labor market and therefore substantially underrepresents the extent of labor market slack at any point in time and for the persons affected.

Aggregate unemployment

The basic trends in the unemployment rate are presented in Figure A, which shows the ups and downs of the quarterly unemployment rate from 1973 through 2019. The line at a presumed full employment rate of 5% illustrates how frequently unemployment remained above full employment. The average unemployment rate over this period was 6.25%, meaning that the economy averaged 1.25 percentage points of unemployment above (a presumed) 5% full employment rate. Table 1 shows that, in the 188 quarters comprising the 1973-2019 period, in only 50 of them was unemployment at 5% or less, or 26.6% of the time (equivalent to 12.5 of the 47 years). Unemployment exceeded 6% for 85 quarters, or 45.2% of the time, equivalent to 21.5 of the 47 years from 1973 to 2019. In other words, the economy of the last five decades was infrequently at or below an unemployment rate of 5%,12 and full employment is far from the norm.

Figure A

Table 1

Unemployment for specific demographic groups

Not only is the unemployment rate frequently greater than that associated with full employment, but the unemployment rate experienced by many demographic groups never achieves full employment, ever. Table 2 presents the distribution of unemployment over the months from 1979 through 2019 for demographic groups delineated by education and race/ethnicity.13 These tabulations illustrate some regularities (reflecting our institutions and systems of discrimination) in the unemployment realm: Workers who are Black or Hispanic have higher unemployment at every level of education, and workers with less educational credentials (e.g., high school graduates) have higher unemployment than those with more credentials (e.g., college graduates). This can be seen in the average unemployment rates presented in the last row of Table 2. Blacks and Hispanics experienced unemployment rates of 11.9% and 8.6% on average, respectively, over the 1979–2019 period, far greater than whites, whose unemployment was 5.1%. This means that the norm for whites was a roughly full employment economy, while Blacks had unemployment twice as high and Hispanics 70% higher than whites.

Table 2

Of course, the averages for particular race/ethnic groups obscure the much higher unemployment for those in the working class of each race/ethnic group. This can be seen by examining the average unemployment of high school graduates in each group: Black, 12.8%; Hispanic, 8.2%; and white, 5.7%. The entire group with less than a four-year college degree (those with some college, a high school degree, or less than a high school education) experiences high unemployment, and this group comprises 81% of the Black workforce over the 1979–2019 period. Blacks without a high school degree averaged 21.4% unemployment, and those with some college (including those with a two-year degree) averaged 9.7%.

Table 2 also illustrates how rarely certain demographic groups enjoy full employment by showing the share of the months over the 1979–2019 period at which specific ranges of unemployment rates prevailed.14 For instance, there was no time over the 1979–2019 period when the Black unemployment rate was 5% or less, or even 6% or less. Blacks, on average, never experienced anything near full employment. Meanwhile, Black high school graduates faced unemployment rates of more than 10.1% for roughly 81% of the 1979–2019 period. Hispanic high school graduates experienced unemployment rates of 5% or less in only 18 (3.7%) of the 492 months in the 1979–2019 period; they faced unemployment exceeding 6.0% in 82.5% of the months. In contrast, college graduates enjoyed long periods of full employment (of 5% or less), though it was more common for white college graduates (99.6% of the time) than for Black (61.4% of the time) or Hispanic (79.7% of the time) college graduates.

In sum, full employment is a rare experience, and even when the aggregate economy has full employment large groups (primarily Blacks and Hispanics and those lacking a four-year college credential) still face excessive unemployment. Employers persistently enjoy an uneven playing field tilted to their advantage simply because workers face excessive unemployment and other types of underemployment.

How high unemployment shapes workers’ options

High unemployment carries consequences for workers and changes their behaviors and outlook. When unemployment is higher workers face greater difficulties switching jobs, have longer spells of unemployment if they become unemployed, and believe that it is harder to find a job. Accordingly, workers are far less likely to quit when unemployment is high. The idea that workers can as readily walk away from a job as an employer can replace a worker does not hold if workers do not consistently enjoy a full employment environment.

Length of unemployment spells. Unemployment rises in recessions as more people are laid off and then stay unemployed for longer spells because jobs are difficult to find (Elsby, Sahin, and Hobijn 2010; Davis, Faberman, and Haltiwanger 2011). The BLS unemployment duration data can be used to demonstrate the general pattern of the lengthening of unemployment spells in recessions, as unemployment escalates.15 Higher unemployment, in fact, arises from more workers becoming unemployed after losing jobs rather than unemployment spells lasting longer.

Table 3 uses the 1979–1983 and 2007–2009 downturns to show how the duration of unemployment (average and median) and the share of the unemployed who are long-term unemployed (greater than 26 weeks) increases as the economy moves from the cyclical peak (low unemployment) into a recession. It shows, for instance, that the duration of unemployment, both average and median, nearly doubled between 1979 and 1983, and both also increased remarkably during the 2007–2009 downturn. The share of workers experiencing long spells of unemployment (exceeding 26 weeks) also spiked. Obviously then, the prospect of quitting and becoming unemployed becomes a much more costly prospect for workers when the economy is not at full employment.

Table 3

Workers find “jobs hard to get” when unemployment rises. Not surprisingly, the share of workers who say in the Conference Board’s Consumer Confidence Survey that they find “jobs hard to get” closely follows the rise and fall of unemployment (Figure B).16 In the periods of high unemployment in 1975, 1983, 1991, and 2009, about half the respondents said it was hard to find a job, up from around 11% in 2000, when unemployment was low, and from about 20% in 2007, before the financial crisis.

Figure B

Quits. It has long been established that the scale of workers quitting their jobs is tightly related to the level of and changes in unemployment. In fact, 30 years ago Akerlof, Rose, and Yellen (1988), using BLS data from manufacturing over the 1948–1981 period to document the procyclical nature of quits, opened their paper by writing: “One indisputable [macroeconomic] regularity is the highly procyclical nature of quits: many more people voluntarily leave their jobs when unemployment is low than when it is high.”

Davis, Faberman, and Haltiwanger (2011) combined data from two BLS data sets—Business Employment Dynamics (BED) and Job Openings and Labor Turnover (JOLTS)—to trace quits from 1990 to 2010; they extended the series to 2019 using more recent JOLTS data (Figure C).17 These data clearly show the substantial decline in quits in the downturns of the early 1990s and early 2000s and from 2007 to 2009, as well as the increase in quits in the recoveries of the late 1990s, 2003–2007, and 2009–2019. The willingness and ability to quit are tightly linked to the level of unemployment. Therefore, a worker’s ability to quit work and the ability of an employer to fill job vacancies (more on this below) are not independent of the unemployment situation, a situation that generates a substantial power asymmetry between employers and employees, contrary to the assumptions of the freedom-of-contract framework. Whatever power the ability of workers to quit has on restraining employer exploitation is diminished when unemployment exceeds the levels prevailing in the relatively rare moments of full employment.

Figure C

The relationship between quits and unemployment in downturns and recoveries is illustrated in Table 4 using Davis, Faberman, and Haltiwanger quits data since 2001.18 One conclusion from Table 4 is that there are a large number of quits each year: For instance, the 45.0 million quits in 2019 represented 29.8% of total jobs held. Second, quits fluctuate a great deal and much more so than unemployment. For instance, unemployment rose 1.9 million in the 2001–2003 downturn but quits diminished far more, by 8.2 million. Likewise, the downturn of 2007–2009 caused unemployment to rise by 7.2 million but quits declined by 15.7 million. Unemployment declines in recoveries, but quits increase much more: In the 2003–2007 and 2009–2019 recoveries the change in quits were, respectively, 3.3 and 2.6 times the fall in unemployment.

Table 4

Quit rates in years of relatively low unemployment (2001, 2007, and 2019) reveal that there is no strong secular trend in quitting. The quit rate was higher in 2001 than in 2007 (31.6% versus 28.5%) but increased only slightly from 2007 to 2019 (from 28.5% to 29.8%).

Much recent research has also identified the procyclical nature of quits, i.e., falling in downturns and rising in recoveries. Elsby, Sahin, and Hobijn (2010) note that “the quit rate moves procyclically.” Davis, Faberman, and Haltiwanger (2012a) find “strongly pro-cyclical movements in quit rates even after conditioning on the employer’s growth rate,” and “the main story for quits appears to involve worker responses to outside labor market conditions [i.e., unemployment]” rather than a cross-sectional relationship to establishment growth rates.

The changes in quits affects far more workers than those who actually quit. Increases (and decreases) of quits affect the motivation of employers to retain their staff. That is, a fall in quits will affect the employment conditions of those who stay. This is an important mechanism by which higher unemployment affects a large segment of the workforce.

The logic of how unemployment affects quits and wages was ably described by Faberman and Justiniano (2015):

The fact that quits are procyclical makes intuitive economic sense. Quits reflect job switching. People are more likely to switch jobs during economic expansions. During these times, there are more jobs available and labor markets are tighter. A tighter labor market implies that em­ployers are more willing to offer higher wages to attract new workers. These higher wages provide a greater incentive for workers to leave their current posi­tion and move up what is often referred to as the job ladder. During recessions, labor markets are more slack. There are fewer available jobs and unemployment is higher, so workers have less bargain­ing power to obtain a better wage offer.

Elsby, Michaels, and Ratner (2020) emphasize that quits generate “replacement hiring” by employers needing to fill vacancies, and this need in turns lures workers from other firms, thereby generating even more quits in a “vacancy chain”; this replacement hiring can account “for a large fraction of aggregate hiring in the U.S. economy.” As Akerlof, Rose, and Yellen (1988) noted, “Quits increase as opportunities expand; the opportunities for job switching are significantly greater when unemployment is low than when it is high.” This process enables workers to obtain better jobs and compensation, as shown by Faberman and Justiniano (2015).

Switching employers. Quits reflect employees leaving their employment voluntarily (with the exception of retirements or transfers to other locations). Researchers have focused on one component of quits that is strongly linked to wage growth—employment-to-employment transitions involving switching employers. Quits, in fact, are dominated by those switching employers rather than those entering unemployment (Elsby, Sahin, and Hobijn 2010).19 This section examines the relationship between the rate of job switching and changes in unemployment.

Fallick and Fleischman (2004) pioneered the measurement of month-to-month labor flows between unemployment, employment, and “not in the labor force” using the BLS Current Population Survey (CPS). However, Fujita, Moscarini, and Postel-Vinay (2021) identify changes in the CPS in 2007 that led to a sizable understatement in employer-to-employer switching significant enough to force an evaluation of previously identified trends (specifically, CPS data suggested there was no secular decline in employer switching over the last 15 years). Their research has focused on an increased (nonrandom) incidence of missing answers to a key survey question on whether the respondent had the same employer. They correct the data with imputations to develop an alternative series, which is what is used in this section.20

Figure D and Table 5, drawing on the Fujita, Moscarini, and Postel-Vinay data, show the changes in levels and rates at which workers switch employers and experience unemployment.21 Figure D shows the rate of employment-to-employment switching rising as unemployment falls and declining as unemployment spikes in a downturn. Table 5 elaborates these trends by examining the rise and fall of job switching over recoveries and downturns. The table shows first that there is a substantial amount of job switching each year. In years of low unemployment, such as 2000 or 2007, those switching employers amounted to 30% or more of employment (there were 43.2 million job switches in 2007). Second, employment switching, like quitting, falls in downturns and rises in recoveries. For instance, employer switching fell from a 33.0% rate in 2000 and to 27.4% in 2003, a 5.6 percentage point decline (17% of the 2000 switching rate). In the 2000 to 2003 downturn an additional 3.4 million workers became unemployed but 7.6 million fewer workers switched jobs. Similarly, the switching rate fell 4.5 percentage points during the financial crisis in 2007–2009. The number of workers added to the unemployment rolls (up 8.4 million) equaled the decline in job switchers (8.5 million). In recoveries there is a larger increase in employment-to-employment switching than there is a decline in unemployment. Looking over the longer term at years of low unemployment reveals a decline in job switching between 2000 and 2007 (from 33.0% to 29.6%) but relative stability between 2007 and the end of the recovery in 2019.

Figure D

Table 5

These data show that the level and changes in unemployment greatly affect the rate and amount of employment-to-employment switching. Most workers find a new job by directly switching employers, rather than finding a new job after becoming unemployed or leaving the labor force, and a higher unemployment environment adversely affects workers’ ability and willingness to switch employers.

Researchers have established that switching jobs is an essential component of workers receiving higher wages. Fujita, Moscarini, and Postel-Vinay (2021) recently wrote that “on-the-job search by, and competition between firms for, employed workers are a natural source of worker bargaining power.” Direct moves from one employer to another have also been shown to be a major source of earnings growth (Topel and Ward 1992), and being thrown off the ”job ladder” can drastically reduce lifetime earnings (Davis and Von Wachter 2011).22

Moscarini and Postel-Vinay (2017) have identified changes in employer switching as more important to wage growth than changes in unemployment:

We thus find no empirical evidence to support the view that workers, when negotiating their wages, have a credible threat to quit to unemployment, whose continuation value naturally depends on how easy it would be to then find alternative employment. Our evidence is instead consistent with a credible threat to quit, hence an ability to extract a wage raise, only once an alternative offer has arrived, or is likely to arrive soon.

Another interpretation is simply that a key way that higher unemployment affects wage growth is by eroding opportunities, which are reflected in reducing both quits and employer switching.

The employer side of higher unemployment

In contrast to employees, the situation of employers becomes more favorable as unemployment rises: Employers recruit less intensively, fill vacancies more quickly, and generally find qualified workers more easily. Employers also use periods of high unemployment to elevate their demands for skills, requiring workers to offer more credentials for similar rates of pay. One can summarize this pattern of evidence as employers increasing their power relative to workers, especially low- and middle-wage workers, in the common instances when unemployment exceeds its full employment level.

We rely heavily on the innovative research by Davis, Faberman, and Haltiwanger (2012a, 2012b, 2013) and Faberman and Justiniano (2015), as well as the BLS JOLTS data, to illustrate key indicators reflecting the employer side of the hiring process. Davis, Faberman, and Haltiwanger build on and improve the JOLTS data on job openings, quits, etc. and extend various data series back to the early 1990s (JOLTS data started in late 2000) using the BED microdata.23

Recruitment intensity

Davis, Faberman, and Haltiwanger (2013) provide a recruiting intensity index that “summarizes, in a quantitative manner, the intensity of employer efforts to recruit for, and fill, their open job positions,” and describe24 what their metric attempts to capture:

Employers with open job positions take several actions and decisions that affect how quickly those positions are filled. Examples include the choice of recruiting methods, expenditures on help-wanted ads, how rapidly employers screen job applicants, their hiring standards, and the attractiveness of compensation packages they offer to prospective new hires.

Recruiting intensity is shorthand for the instruments employers use to influence the pace of new hires—e.g., advertising expenditures, screening methods, hiring standards, and the attractiveness of compensation packages. These instruments affect the number and quality of applicants per vacancy, the speed of applicant processing, and the acceptance rate of job offers.” The authors note that their metric is an indirect one due to data limitations.

The trends in recruiting intensity per vacancy are presented in Figure E for the years 2001–2017. The metric fell about 13% from the low-unemployment first quarter of 2001 to the unemployment high point of the second quarter of 2003; it grew modestly during the ensuing recovery (up just 3.9% by the second quarter of 2007) but then fell sharply, by 17.8%, in the financial crisis downturn through 2009.25 By the second quarter of 2017 (the latest data available) recruiting intensity per vacancy was still slightly below its 2007 peak. Clearly, recruiting requires and receives less effort by employers as unemployment rises.

Figure E

Employer efficacy in filling jobs and the duration of vacancies

Employers may exert less effort recruiting workers as unemployment rises, and they are definitely more successful in filling vacancies when unemployment is higher. This can be seen in the vacancy duration measure, which “quantifies the average number of working days taken to fill vacant job positions,” developed by Davis, Faberman, and Haltiwanger (2013)26 and presented in Figure F.

Figure F

The average days to fill a vacancy, or job opening, in early 2001 was 22.5 days, but it fell to just 18.1 days at the recession’s unemployment high point in 2003. As the economy recovered from 2003 to 2007, the days needed to fill a vacancy grew back to 23.3, a bit above the early 2001 level. Not surprisingly, as the economy descended during the financial crisis, the days required to fill a vacancy fell by 7.2 days (a 30.7% drop) to just 16.2 days. In the second quarter of 2017, when the unemployment rate had fallen to 4.4%, it was taking much longer, 29 days, to fill a vacancy, many more than observed in the series’ starting date in 2001.

Another way to observe the ease with which employers fill jobs is by examining the “job-filling rate,” the number of new hires compared with the number of vacancies, or job openings, in the prior month, as shown in Figure G using JOLTS data.27 These data draw on the same data as Figure F, though scaled to the number of days in a month available to fill a vacancy; therefore, Figure G’s measure of the job-filling rate is another way to illustrate the vacancy duration. Employers hired 1.07 workers for every job opening in early 2001 but were able to hire 1.45 workers per prior job opening at the unemployment trough of the early 2000s recession in the second quarter of 2003. Hiring per job opening slowed down by the time the recovery ended in 2007, to 1.13, but escalated to 1.56 by the summer of 2009, when unemployment was high due to the financial crisis. At the end of the recent recovery, in late 2019, with unemployment down to 3.6%, employers were only roughly half as efficient in filling job openings—a rate of 0.82—as in the very high unemployment year of 2009.

Figure G

Employers clearly are more able to recruit workers, and do so more quickly, when unemployment is higher than when it approaches full employment.

Employers know that it is easier to recruit at higher unemployment

The National Federation of Independent Business publishes a survey, Small Business Economic Trends (SBET), that tracks small businesses’ assessments of the hiring process and outcomes.28 The SBET data demonstrate the cyclicality of business assessments of the quality of labor and the difficulty in obtaining qualified workers and filling job openings. Figures H, I, and J present the NFIB quarterly data as far back as they go (to 1973 for unfilled job openings and “labor quality as the single most important problem,” and to 1993 for lack of qualified job applicants) through 2019. Though the SBET samples are relatively small (514 respondents in the March 2021 survey, but 1600-1700 in January, April, July, and October of each year), the data do provide insights on time trends over business cycles.

One can readily see in Figure H that the share of firms with unfilled job openings is greatest in years of low unemployment (1973, 1979, 1989, 2000, 2007, 2019), and there are fewer job openings when unemployment is high (1975, 1982–1983, 1992–1993, 2003–2004, 2009–2010). In fact, the share of firms with an unfilled job opening fell from 24% in the fourth quarter of 1979 to just 8% in 1982:4, and it fell from 33.3% in 2000:4 to just half as many, 16.3%, in 2003:2. It is certainly easier for firms to fill openings when unemployment is greatest, according to the small businesses themselves (who are generally the last in line to obtain new hires).

Figure H

The ability to find qualified job applicants also seems to be far easier when unemployment is high. The trend in whether a firm is seeing “few or no qualified applicants” (aggregating two series) is also clearly related to the level of unemployment (Figure I). These data show that nearly half (48.9%) of small firms reported trouble finding qualified applicants in 2000 but only a third did so in 2003:2 after unemployment peaked in the recession. Similarly, the 43.3% of small firms challenged to find qualified applicants in 2007 was reduced to just 24.7% in 2009:4 when unemployment had risen to 9.9%. Thus, higher unemployment allows firms to fill openings and do so with qualified applicants relative to times of low unemployment.

Figure I

Finally, small firms assessing “labor quality” as their single most important problem seems be at its highest when unemployment is low but is greatly minimized when unemployment is very high, as in 1982–1983, 1992–1993, and 2009–2010 (Figure J). High unemployment seems to be associated with small businesses being readily able to fill openings with qualified applicants and satisfy their needs for “labor quality.”

Figure J

Employers opportunistically ask for more credentials when unemployment rises

Evidence from the Great Recession shows that employers take advantage of their easier access to new workers when unemployment is high to require greater credentials for low- and middle-wage jobs (Modestino, Shoag, and Balance 2020). This research confirms what a CareerBuilder (2014) survey in 2013 found, that “almost one-third of employers said that their educational requirements for employment had recently increased, and specifically that they were hiring more college-educated workers for positions previously held by high school graduates” (Modestino 2019). So, employers not only fill openings more readily and do so with qualified candidates when unemployment is high, but they are also able to opportunistically require greater credentials (without increasing pay) than they previously did when unemployment was lower.

Modestino, Shoag, and Balance (2020) use the near-universe of online job postings (roughly 159 million total) aggregated by Burning Glass Technologies to document that the share of job postings requiring greater credentials—both a college degree (or more) and four or more years of experience—spiked between 2007 and 2011–2012 and then declined as unemployment declined in the recovery. Moreover, the upskilling was largely temporary for occupations in the middle- and low-skill sectors, prevailing when unemployment remained high but mostly reversing once the labor market tightened (by 2017, the latest data) (Burke et al. 2020). This opportunistic upskilling occurs within occupations and in occupations in the same firm and does not “simply reflect a shift in the composition of employers or the positions that they seek to fill” (Modestino, Shoag, and Balance 2020). Researchers found that “the increase in employer skill requirements was greater in areas where the unemployment rate rose more dramatically and the decrease was larger in areas where the unemployment rate fell more swiftly during the recovery. These effects are very robust, showing up within specific occupations and even job titles. For example, only 15% of physician assistant jobs required a Bachelor’s degree or higher in 2007. That share jumped to 35% in 2010 and has since fallen to just 12% as of 2017” (Modestino and Shoag 2018).

This pattern of evidence confirms that this opportunistic credential upskilling reflected employers’ increased power relative to low- and middle-wage workers when unemployment escalated in the Great Recession. As unemployment receded, employers were forced to normalize the credentials they required, retreating to what they asked for before the recession.

The bottom line: Higher unemployment leads to lower wage growth, especially for low- and middle-wage workers

It has long been established that higher unemployment leads to lower wage growth (Phillips 1958) and does so particularly for those with the least power in the labor market. This uncontested fact alone validates the importance of recognizing the persistent divergence of actual unemployment from full employment as it pertains to the supposed equal power of workers and their employers.

Mishel and Bivens (2021) review the impact of excessive unemployment—the degree to which unemployment exceeds full employment—on the wages of low- and middle-wage workers. They first note the degree to which unemployment departed from full employment over the last few decades:

These contractionary policies caused unemployment to remain 0.8 percentage points above even a conservative estimate of full employment (the NAIRU)—5.5%—between 1979 and 2017, a sharp contrast from the 0.51 percentage points that unemployment remained below the NAIRU in the prior 30 years.

They also estimate the corresponding wage impact, drawing on the lower bound of estimates from Bivens and Zipperer (2018)29:

The impact of excessive unemployment…reduced wages for the median worker by 10.0% between 1979 and 2017. Adjusting for the “flattening of the Phillips curve since 2008, as we do here, lessens the impact of higher unemployment on wage growth; without this adjustment the impact would have been 12.2%. If the unemployment rate had been held lower, say to 5% on average, then median wages would have been about 18.3% higher by 2017. Of course, a 5.5% target for full employment is a modest goal, and if policymakers had achieved a reasonable target of 4.5% the impact of excessive unemployment would be double the 10.0%” adverse wage impact on the median worker.

Excessive unemployment had a somewhat larger impact on low-wage than middle-wage workers. Had unemployment averaged 5.5% rather than the 6.3% that prevailed over the 1979–2017 period, the wages of the 10th percentile [low-wage] worker would have been 11.6% higher….[T]hese estimates take into account the “flattening” of the Phillips curve post-2008. We would note that the impact of higher unemployment would be twice as large if full employment was assumed to be 5.0%.

Mishel and Bivens note that these estimated wage impacts are far below those of Katz and Krueger (1999, Table 8), whose Phillips curve estimates using a 1973–1998 time series were double those of Bivens and Zipperer (2018) at the median and three times those at the 10th percentile.

In sum, higher unemployment has consequential adverse wage impacts for middle-wage workers and even more so for lower-wage workers.

Conclusion

The freedom-of-contract view of the world, and the assumption of equal power between employers and employees, ignores the obvious and basic truth about labor markets: The economy is rarely at full employment, and many workers never experience full employment. The presence of excessive unemployment—beyond full employment—tilts the power balance toward employers. Just acknowledging high unemployment leads one to recognize that in many, if not most, circumstances employers can far more readily replace a worker than a worker can find a comparable job. To believe otherwise is to live in a world without access to windows or newspapers, and it is curious and unsettling that claims of freedom of contract have been made when there was, or had recently been, very high unemployment. Simply acknowledging the persistent absence of full employment for many workers renders the freedom-of-contract framework a flawed basis for assessing employment relationships and arrangements.

About the author

Lawrence Mishel is a distinguished fellow and former president of the Economic Policy Institute. He is the co-author of all 12 editions of The State of Working America. His articles have appeared in a variety of academic and nonacademic journals. His areas of research include labor economics, wage and income distribution, industrial relations, productivity growth, and the economics of education. He holds a Ph.D. in economics from the University of Wisconsin at Madison.

Acknowledgments

The comments and data provided by the following are greatly appreciated: Josh Bivens, Jason Faberman, Shigeru Fujitay, Andrew Heritage, Wilma Liebman, and David Ratner. Melat Kassa and Jori Kandra provided excellent and needed research assistance.

Notes

1. Adair v. United States, 208 U.S. 161, 174–75 (1908) [citation in original].

2. Id. at 175 [citation in original].

3. Id. at 175–176 [citation in original].

4. Epic Systems Corp. v. Lewis, 138 S. Ct. 1612 (2018).

5. See the SCOTUS blog, “Epic Systems Corp. v, Lewis.”

6. Justice Ginsburg points out in the second footnote: “The Court’s opinion opens with the question: ‘Should employees and employers be allowed to agree that any disputes between them will be resolved through one-on-one arbitration?’ Ante, at 1. Were the ‘agreements’ genuinely bilateral? Petitioner Epic Systems Corporation e-mailed its employees an arbitration agreement requiring resolution of wage and hours claims by individual arbitration. The agreement provided that if the employees ‘continue[d] to work at Epic,’ they would ‘be deemed to have accepted th[e] Agreement.’ App. to Pet. for Cert. in No. 16–285, p. 30a. Ernst & Young similarly e-mailed its employees an arbitration agreement, which stated that the employees’ continued employment would indicate their assent to the agreement’s terms. See App. in No. 16–300, p. 37. Epic’s and Ernst & Young’s employees thus faced a Hobson’s choice: accept arbitration on their employer’s terms or give up their jobs.”

7. See Table 4, “Unemployment in Manufacturing, Transportation, Building Trades, and Mining, 1897-1926,” as estimated by Paul H. Douglas, from Real Wages in the United States, 1890-1926, in the Committee on Economic Security report.

8. BLS data on unemployment from series LNS14000000.

9. BLS data on unemployment from series LNS14000000.

10. In particular, an economic analysis of the Beveridge curve (Beveridge 1942), which focuses on vacancies being equal to openings at full employment. David Ratner pointed this out to me.

11. According to Crump et al. (2019): “This natural rate of unemployment, ut *, is broadly defined as the unemployment rate such that, controlling for supply shocks, inflation remains stable.”

12. If the analysis had used 5.5% as the benchmark for full employment, then an additional 11.5% of the quarters would have had full employment (36.8% overall), but there were still no quarters in which Blacks experienced full employment, and there were just 3.5% more quarters of full employment for Hispanics.

13. The data are tabulations of the basic BLS monthly Current Population Survey microdata available in the EPI State of Working America Data Library.

14. Bernstein and Jones (2020) present a similar analysis for all workers, Blacks, and whites by quarter.

15. The BLS measure of unemployment duration is flawed, however, because it is a point-in-time measure of not-yet-completed spells of unemployment (Horrigan 1987; Valletta 2002). Valletta (2002) describes the biases: “The upward bias occurs because longer spells, purely by virtue of their length, are more likely to be in the monthly unemployment sample than are shorter spells. The downward bias arises because the use of in-progress spells precludes measurement of completed spell durations.”

16. This is one of the eight indicators included in the Conference Board Employment Trends Index. I am greatly appreciative of the Conference Board’s sharing of these data.

17. The analysis relies on the Davis, Faberman, and Haltiwanger data because it improves on the available JOLTS data. The authors’ 2013 paper showed that the JOLTS survey understated worker churn at very high/very low growth establishments. Thus, their series differs from JOLTS at the start of the 2000s. The differences disappear by 2017, however. Correspondence form Faberman explains: “the shrinking difference comes from a combination of using the OLS regression as a quick prediction rather than the micro data, [and] the fact that JOLTS people at the BLS responded to our paper (and a couple others on JOLTS measurement) by getting better at capturing the worker churn at very high/low growth establishments over time (for example, they introduced a birth/death adjustment in there at some point).”

18. The quit rates are those developed by Davis, Faberman, and Haltiwanger (2012a) and updated in June 2021. Jason Faberman graciously provided the data and answered numerous questions. These data correct for some understatement of worker churn in the JOLTS data in the earlier years. JOLTS and the Davis, Faberman, and Haltiwanger data are similar in 2019. Unemployment from BLS. The quit level uses the quit rate and the employment level implicit in the JOLTS data (annual quit level divided by the annual quit rate).

19. Another possibility is quitting to leave the labor force.

20. “We uncover a drastic increase in the incidence of missing answers to the pertinent survey question (SAMEMP) starting in January 2007, predating by about a year the full introduction of new interviewing policy, the Respondent Identification Policy (RIP). We provide evidence that these answers are not missing at random, and these interviewing changes caused a serious permanent downward bias in the standard measure of employer-to-employer transitions. We propose a model of selection by observable and unobservable worker characteristics, and build on it to impute the missing answers to recover the true aggregate employer-to-employer monthly transition probability. We show that its decline observed during the Great Recession started about a year later and was much less dramatic than the raw, biased series indicates, and had fully recovered by 2016, if not earlier” (Fujita, Moscarini, and Postel-Vinay 2021, 42).

21. Shigeru Fujita kindly provided the updated data and answered questions.

22. Fujita, Moscarini, and Postel-Vinay (2021) elaborate on the earlier research.

23. The Davis, Faberman, and Haltiwanger indicators can be found at https://www.dice.com/indicators/.

24. These are from the Q&A offered at the website presenting their indicators.

25. Davis, Faberman, and Haltiwanger (2012b) note an even stronger decline, 21%, from December 2007 to the trough of the Great Recession.

26. Rothstein (2012) offers some useful caveats regarding these data. “A more important concern is that measured job openings data and the openings-to-unemployment ratio are only loosely related to the efficiency of the economic matching process, particularly in an unprecedentedly long period of labor market weakness.” For instance, firms “might hold out for better-qualified workers, extending its search, or might be less choosy in order to hire more quickly (Diamond 2013). Either decision affects the number of measured job openings and the job filling rate, but neither reflects changes in labor market matching efficiency.”

27. The data series developed by Davis, Faberman, and Haltiwanger (2013) and provided at their website or by Jason Faberman directly does not include job openings. We, therefore, use the BLS JOLTS data.

28. The SBET data were kindly provided by Andrew Heritage. Findings and methodological details are in Wade and Heritage (2020).

29. Bivens and Zipperer (2018) find that a 1 percentage point drop in unemployment results in annual wage growth 0.5–1.5 percentage points faster for workers at the 10th percentile. For example, if annual real wage growth is 1%, then a 1 percentage point fall in unemployment would result in annual real wage growth rising to 1.5% to 2.5%. For workers near the median of the wage distribution, wage growth is faster by 0.4–0.9 percentage points, and for workers at the 90th percentile it is 0.3–0.5 percentage points faster. These estimates indicate that excessive unemployment generates increases in the wage gaps between low- and middle-wage workers and between middle-wage and higher earners.

References

Akerlof, George A., Andrew Rose, and Janet L. Yellen. 1988. “Job Switching and Job Satisfaction in the U.S. Labor Market.” Brookings Papers on Economic Activity, 1988, no. 2: 495–582.

Bagenstos, Samuel. 2020. “Lochner Lives On: Lochner Presumption of Equal Power Lives in Labor Law and Undermines Constitutional, Statutory, and Common Law Workplace Protections.” Unequal Power series. Economic Policy Institute.

Baker, Dean. 2000. NAIRU: Dangerous Dogma at the Fed Financial Markets Center. Center for Economic Policy and Research.

Bernstein, Jared, and Dean Baker. 2013a. “Getting Back to Full Employment: A Better Bargain for Working People.” Center for Economic Policy and Research.

Bernstein, Jared, and Dean Baker. 2013b. “The Unemployment Rate at Full Employment: How Low Can You Go?” Economix (New York Times), November 12.

Bernstein, Jared, and Janelle Jones. 2020. The Impact of the COVID19 Recession on the Jobs and Incomes of Persons of Color. Center on Budget and Policy Priorities.

Beveridge, William. 1942. Social Insurance and Allied Services. London: H.M. Stationery Office.

Bivens, Josh, and Ben Zipperer. 2018. The Importance of Locking in Full Employment for the Long Haul. Economic Policy Institute.

Burdett, Kenneth, and Dale T. Mortensen. 1998. “Wage Differential, Employer Size, and Unemployment.” International Economic Review 39, no. 2: 257–73.

Bureau of Labor Statistics (BLS). 2018. Current Population Survey: Handbook of Methods. Last updated April 2018.

Burke, Mary A., Alicia Sasser Modestino, Shahriar Sadighi, Rachel B. Sederberg, and Bledi Taska. 2020. “No Longer Qualified? Changes in the Supply and Demand for Skills Within Occupations.” Federal Reserve Bank of Boston, Working Paper No. 20-3.

CareerBuilder. 2014. “Education Requirements for Employment on the Rise, According to CareerBuilder.” Press release, March 20.

Congressional Research Service (CRS). 2020. Introduction to U.S. Economy: Unemployment.

Crump, Richard K., Stefano Eusepi, Marc Giannoni, and Aysegul Sahin. 2019. “A Unified Approach to Measuring u*.” Brookings Papers on Economic Activity Conference draft, Spring, 143–238.

Davis, Steven J., Jason Faberman, and John C. Haltiwanger. 2011. “Labor Market Flows in the Cross Section and over Time.” National Bureau of Economic Research, Working Paper No. 1729.

Davis, Steven J., Jason Faberman, and John C. Haltiwanger. 2012a. “Labor Market Flows in the Cross Section and Over Time.” Journal of Monetary Economics 59, no. 1: 1–18.

Davis, Steven J., Jason Faberman, and John C. Haltiwanger. 2012b. “Recruiting Intensity During and After the Great Recession: National and Industry Evidence.” American Economic Review 102, no. 3: 584–88.

Davis, Steven J., Jason Faberman, and John C. Haltiwanger. 2013. “The Establishment-Level Behavior of Vacancies and Hiring.” Quarterly Journal of Economics 128, no. 2: 581–622.

Davis, Steven J., and Till Von Wachter. 2011. “Recessions and the Costs of Job Loss.” Brookings Papers on Economic Activity, 2011 (Fall): 1–72.

Diamond, Peter. 2013. “Cyclical Unemployment, Structural Unemployment.” IMF Economic Review 61: 410–55.

Edwards, Kathryn. 2022 (forthcoming). “The Economic Vulnerability of Workers.” Unequal Power series. Economic Policy Institute.

Elsby, Michael W.L., Ryan Michaels, and David Ratner. 2020. “Vacancy Chains.” Federal Reserve Bank of Pennsylvania, Working Paper 20-28.

Elsby, Michael W.L., Aysegul Sahin, and Bart Hobijn. 2010. “The Labor Market in the Great Recession.” Brookings Papers on Economic Activity, 2010 (Spring): 1–48.

Epstein, Richard A. 1984. “In Defense of the Contract at Will.” University of Chicago Law Review 51, no. 982: 947–63.

Faberman, R. Jason, and Alejandro Justiniano. 2015. “Job Switching and Wage Growth.” Chicago Fed Letter (Federal Reserve Bank of Chicago), no. 337.

Fallick, Bruce, and Charles A. Fleischman. 2004. “Employer-to-Employer Flows in the U.S. Labor Market: The Complete Picture of Gross Worker Flows.” Board of Governors of the Federal Reserve System Finance and Economics Discussion Series, no. 2004-34.

Fujita, Shigeru, Giuseppe Moscarini, and Fabien Postel-Vinay. 2021. “Measuring Employer-to-Employer Reallocation.” National Bureau of Economic Research, Working Paper No. 27525.

Galbraith, James K. 1997. “Time to Ditch the NAIRU.” Journal of Economic Perspectives 11, no. 1: 93-108.

Horrigan, Michael W. 1987. “Time Spent Unemployed: A New Look at Data from the CPS.” Monthly Labor Review (July): 3–15.

Katz, Lawrence, and Alan Krueger. 1999. ”The High-Pressure Labor Market of the 1990s.” Brookings Papers on Economic Activity 1999, no. 1.

Mishel, Lawrence, and Josh Bivens. 2021. “Identifying the Policy Levers Generating Wage Suppression and Wage Inequality.” Unequal Power series. Economic Policy Institute.

Modestino, Alicia Sasser. 2019. Is the Skills Gap Real? Changes in the Employer Skill Requirements During the Great Recession. EconoFact, March.

Modestino, Alicia Sasser, and Daniel Shoag. 2018. “When the Economy Is Good, Employers Demand Fewer Credentials.” Harvard Business Review: 1–8.

Modestino, Alicia Sasser, Daniel Shoag, and Joshua Balance. 2020. “Upskilling: Do Employers Demand Greater Skill When Workers Are Plentiful?” Review of Economics and Statistics 102, no. 4: 793–805.

Moscarini, Giuseppe, and Fabien Postel-Vinay. 2016. “Wage Posting and Business Cycles.” American Economic Review 106, no. 5: 208–13.

Moscarini, Giuseppe, and Fabien Postel-Vinay. 2017. “The Relative Power of Employment-to-Employment Reallocation and Unemployment Exits in Predicting Wage Growth.” American Economic Review, Papers and Proceedings 107, no. 5: 364-68.

Naidu, Suresh, and Michael Carr. 2022 (forthcoming). “If You Don’t Like This Job, You Can Always Quit?” Unequal Power series. Economic Policy Institute.

National Federation of Independent Business (NFIB). 2021. “Small Business Job Openings Hit Record Level in March.” Press release, April 1.

Phillips, A.W. 1958. “The Relationship Between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom, 1861–1957.” Economica 25, no. 100: 283–99.

Rothstein, Jesse. 2012. “The Labor Market Four Years into the Crisis: Assessing Structural Explanations.” Industrial Labor Relations Review 65, no. 3: 467–500.

Staiger, Douglas, James Stock, and Mark Watson. 1997. “The NAIRU, Unemployment and Monetary Policy.” Journal of Economic Perspectives 11 (Winter): 35–50.

Topel, Robert, and Michael Ward. 1992. “Job Mobility and the Careers of Young Men.” Quarterly Journal of Economics 107, no. 2: 439–79.

Valletta, Rob. 2002. “Recent Trends in Unemployment Duration.” FRBSF Economic Letter (Federal Reserve Bank of San Francisco), no. 35.

Wade, Holly, and Andrew Heritage. 2020. Small Business Problems and Priorities. National Federation of Independent Business.

Inequalities abound in the U.S. economy, and a central driver in recent decades is the widening gap between the hourly compensation of a typical (median) worker and productivity—the income generated per hour of work. This growing divergence has been driven by two other widening gaps, that between the compensation received by the vast majority of workers and those at the top, and that between labor’s share of income and capital’s. This paper presents evidence that the divorce between the growth of median compensation and productivity, the inequality of compensation, and the erosion of labor’s share of income has been generated primarily through intentional policy decisions designed to suppress typical workers’ wage growth, the failure to improve and update existing policies, and the failure to thwart new corporate practices and structures aimed at wage suppression. Inequality will stop rising, and paychecks for typical workers will start rising robustly in line with productivity, only when we enforce labor standards and embrace policies that reestablish individual and collective bargaining power for workers.

Between 1979 and 2017, the compensation of median workers trailed economywide (net) productivity growth by roughly 43%, leading to rising inequality. The effects have been felt broadly: During this time 90% of U.S. workers experienced wage growth slower than the economywide average, while workers at the top (mostly highly credentialed professionals and corporate managers) and owners of capital reaped large rewards made possible only by this anemic wage growth for the bottom 90%. Because the historical legacy of racism has concentrated Black and Latinx workers in the lower half of the wage scale more so than white workers, widespread wage suppression based on class position has inflicted disproportionate harm on them. Further, while women’s wages have grown faster than those for men in recent decades, women’s wage growth still has lagged the economy’s potential. In the fight for a piece of the ever-shrinking share of economic growth available to the bottom 90%, any one group’s gain can feel like another’s loss, leading to political divisions and hindering the formation of cross-racial coalitions based on common interests as workers. In other words, the disappointing wage growth of recent decades is an important economic and political issue.

Yet sluggish wage growth is not a political secret; it has been widely recognized across the political spectrum, even cited by both the Republican and Democratic Party platforms in 2016.1 The root causes of the trend have frequently been misidentified, however. One prominent interpretation is that disappointing wage growth is an unfortunate result of apolitical market forces that one neither can nor would want to alter. Since labor markets are generally competitive and workers and employers have roughly balanced degrees of market power, this argument naively assumes, fundamental apolitical forces like technological change and automation, as well as globalization, have mechanically shifted demand away from non-college-educated and middle-wage workers. But, as this paper will show, the premier research cited in support of an automation/technological theory has itself actually offered empirical metrics that demonstrate that automation/technological change fails to explain wage trends and wage inequality, especially in the period since 1995. Since the automation/technological change explanation is the preeminent explanation drawn from competitive labor market analyses based on equal bargaining power between employers and employees, the failure of automation/technological change to explain wage suppression and wage inequality represents the inability of competitive labor market analyses to adequately explain one of the most salient features of the economy over the last four decades.

Thus, we need to look further for more convincing empirical explanations of why, during a period of rising productivity, hourly compensation for the bottom 90% of all workers has risen so slowly in spite of overall income growth. Doing so requires explaining the key dynamics. The growing wedge between rising productivity and compensation growth for the typical worker financed the increased share of compensation going to top earners, especially those in the top 1% and 0.1%, along with a declining share of income going to labor. In addition, over the last four decades there has been a persistent disparity in the growth of earnings between those in the 90–99% range and those in the middle. Further, wage disparities by gender, race, and ethnicity from the late 1970s, reflecting systemic sexism and racism, remain with us and have sometimes even worsened. Any accounting of where we are and what policies we need must address these issues.

This paper offers a narrative and supporting evidence on the mechanisms that have suppressed wage growth since the late 1970s. We refer in this analysis to wage suppression rather than wage stagnation because it was an actively sought outcome—engineered by policymakers who invited and enabled capital owners and business managers to assault the leverage and bargaining power of typical workers, with the inevitable result that those at the top claim a larger share of income. These policy changes and the change in business practices they enabled have systematically undercut individual workers’ market (exit and voice) options and the ability of workers to obtain higher pay, job security, and better-quality jobs. These corporate and policy decisions had the most adverse consequences for low- and middle-wage workers, who are disproportionately women and minorities, the groups whose legacy of being discriminated against in labor markets means that they especially need low unemployment, unions, strong labor standards, and policy supports for leverage when bargaining with employers.

Neither slow productivity growth nor inevitable economic forces can explain U.S. wage problems. Rather, wage suppression reflects the failure of economic growth to reach the vast majority. It was a “failure by design” (Bivens 2010), engineered by those with the most wealth and power. The dynamics are primarily located in the labor market and the strengthening of employers’ power relative to their rank-and-file workforce (which increasingly includes those workers with a four-year college degree). In other words, the dynamics that have challenged the growth of living standards for the vast majority are based on workers not sharing in economic gains, not, as some have argued, on consumers suffering from monopolistic prices. Changes in product market monopoly and corporate structures have had an impact, but primarily by squeezing supply chain profits and wages rather than by spurring higher consumer prices through much wider profit margins.

As we will discuss, six factors can collectively explain most of the growth of wage inequality and the erosion of labor’s share that resulted in wage suppression over the last four decades (specifically 1979–2017):

1. Austerity macroeconomics, including facilitating unemployment higher than it needed to be to keep inflation in check, and responding to recessions with insufficient force;
2. Corporate-driven globalization, resulting from policy choices, largely at the behest of multinational corporations, that undercut wages and job security of non-college-educated workers while protecting profits and the pay of business managers and professionals;
3. Purposely eroded collective bargaining, resulting from judicial decisions, and policy choices that invited ever more aggressive anti-union business practices;
4. Weaker labor standards, including a declining minimum wage, eroded overtime protections, nonenforcement against instances of “wage theft,” or discrimination based on gender, race, and/or ethnicity;
5. New employer-imposed contract terms, such as agreements not to compete after leaving employment and to submit to forced private and individualized arbitration of grievances; and
6. Shifts in corporate structures, resulting from fissuring (or domestic outsourcing), industry deregulation, privatization, buyer dominance affecting entire supply chains, and increases in the concentration of employers.

Concretely, our analysis attempts to account for the 43 percentage point divergence between the growth of productivity (net of depreciation) and median hourly compensation (wage and benefit) growth between 1979 and 2017. This 43 percentage point wedge excludes any impact of the differing measures of prices used to inflation-adjust productivity and compensation growth. Had median hourly compensation grown with net productivity it would have increased from $20.48 in 1979 to $33.10 in 2017 ($2019). In fact, median hourly compensation was $23.15 in 2017, a $9.95 shortfall from the net productivity benchmark.

We estimate that the first three factors—the impacts that are largest and best measured, i.e., excessive unemployment, eroded collective bargaining, and corporate-driven globalization—explain 55% of the divergence between growth in productivity and median hourly compensation, and specific other factors included above—a diminished overtime salary threshold, employee misclassification, employer-imposed noncompete agreements, and corporate fissuring-subcontracting and major-buyer dominance—explain another 20%. Together, the factors for which we have been able to assess their impact on the median wage can account for three-fourths of the divergence between productivity and median hourly compensation growth from 1979 to 2017. Other factors that we have not been able to empirically assess—increased wage theft and weak enforcement, anti-poaching agreements, increased discrimination, forced arbitration agreements, guestworker programs, and increased prevalence of employer-created “lawless zones” in the labor market where workers are deprived of effective labor protections because of their immigration status—have also contributed to wage suppression.

Our analysis also seeks to account for the falling wages at the 10th percentile and the growth of the wage gap between the 10th percentile and the 50th. We find that these are readily explained by excessive unemployment and the failure to maintain the real value of the minimum wage, factors that have lowered the earnings of the bottom third. Other factors for which we do not yet have good measures of their impact (increased wage theft, the increased share of workers without effective legally protected rights due to their immigration status, and employee misclassification) likely play a role as well. In contrast, our analysis of data (in a section below and in Appendix A) related to automation and skill-biased technological change finds that these factors have had no impact on the suppression of median wages for at least the last 25 years.

Centering power and policy, not apolitical ‘market forces,’ in debates over U.S. wages

The large increase in inequality in the U.S. economy coincided with a pronounced political movement that called for increasing the reach and influence of markets in American life. This movement, known as market fundamentalism and often shorthanded as neoliberalism, has a long history in both U.S. and international policy debates. By the late 1970s, it achieved great prominence, even within the Democratic Party. Neoliberalism’s belief that markets are more efficient and effective than alternative instruments for distributing resources and organizing economic life led it straight to many of the policy recommendations that drove the rise in inequality. Neoliberals, for example, see minimum wages as an inefficient friction in otherwise competitive and efficient labor markets. So, minimum wages have been allowed to be battered by inflation.

The neoliberal policy agenda has often had its most enthusiastic proponents among economists. Under the influence of neoliberalism the economics profession, encompassing both the liberal and conservative wings, shed many older but supremely valuable insights about the importance of institutional checks on markets (especially labor markets) and began analyzing all labor market developments through the lens of textbook competitive models. In these models, the great wage deceleration for the vast majority of workers after 1979 and the resulting rise in inequality could only have happened if impersonal market forces shifted relative demand or supply curves for different sorts of labor.

The mistake of assuming that markets are always well characterized by the simplest competitive models of textbooks proliferated well beyond the labor market. With regard to product markets, the assumption led economists to argue for a retreat from robust anti-trust enforcement and for the deregulation of industries in sectors such as trucking, airlines, interstate busing, and utilities. The assumption that financial markets were competitive led economists to argue for financial deregulation. The promotion of shareholder primacy, that corporations and executives should only advance the profit needs of shareholders, was an important component of this market fundamentalism.

But the greatest damage occurred in labor markets. The focus on labor market competition led to the promotion of “labor flexibility” to achieve growth; this policy agenda sought to weaken collective bargaining, worker protections, and the social safety net (e.g., unemployment insurance). If your model says that only apolitical market-driven shifts of demand and supply curves can explain wage and employment trends, then what might these forces be? As wage suppression took hold, the consensus of elite economists, both liberals and conservatives, excused it as the result of computer-driven automation, a factor we would neither want to nor could restrain. In this scenario, the sole answer is to provide more skills and college education for the workers who have skill deficits—or essentially telling workers that they themselves are to blame for their loss of quality jobs.

But these explanations just don’t fit the data, and when data and model conflict, the wise move is to follow the data. So, if the neoliberal analysis fails to explain wage trends, then the resulting neoliberal policy recommendations should be jettisoned. If we manage to do this, our analysis of just what has happened to U.S. wages and inequality will be stronger and provide a better basis for fixing these issues going forward.

A good example of how the policies we have adopted rather than markets explain wage trends can be found by looking at the wage gaps between workers of different races, genders, and ethnicities. Understanding wage suppression as the result of the exercise of power in labor markets widens our understanding of the potential sources of these long-standing race and gender disparities, gaps that widened during the coronavirus pandemic. The systemic racism that slots minority workers into lesser-paid jobs has made these workers the primary victims of the systematic weakening of worker power. Consequently, one of the key mechanisms to lessen racial and gender inequities is to restore worker power generally as well as to shape policy to ensure all workers have access to good jobs.

This paper’s analysis complements and points in the same direction as other recent research that has focused attention on worker power. For instance, Stansbury and Summers (2020) also argue that reduced worker power explains sluggish wage growth and a declining labor share of income. New empirical examinations of employer monopsony power have identified a growing (at least since the late 1990s) and pervasive employer ability to mark down wages from 20% to 50% and to exert more power over low-wage workers than others. This new monopsony literature provides a top-down analysis, estimating the aggregate potential employer power to suppress wages and then examining the contributing role of countervailing forces like unionization, high-pressure labor markets, and high values of minimum wages in explaining an aggregate net metric of employer power. In contrast, we provide a bottom-up analysis examining the impact of many specific factors and gauging their contribution to the overall divergence between productivity and median compensation growth.

Joseph Stiglitz (2012, 2021) has long focused on power in markets, emphasizing both product market monopoly power and the weakening of employee power relative to employers. He recently provided an analysis similar to the framework of this paper:

The commonsense statement that employers have power over their employees has long been heretical in the economics profession…. More and more, firms have demonstrated high and increasing levels of market power. At the same time, the bargaining power of workers has weakened…. [T]his imbalance of market power has consequences…. It enables firms to suppress wages of workers below what they would be in a competitive marketplace—contributing to the inequality crisis facing the country…. Employers and employees need to be able to bargain on more-equal footing. (Stiglitz 2021)

Our research and these other recent findings demonstrate that employer power is ubiquitous in labor markets, and that wages will be lower and wage growth suppressed absent institutions and policies that provide countervailing power. In other words, employer power is a constant of modern labor markets, but what has changed over the past generation or two is the erosion of institutions and policies—high-pressure labor markets, robust enforcement, unions, and meaningful minimum wages—that once provided that countervailing power.

The paper proceeds as follows. The first step is examining the wage and profit trends that any theory of wage suppression needs to explain. The second section assesses the conventional explanation of “skill-biased technological change”—namely, that in the face of rapid technological change or automation workers lack the skills necessary for more modern production systems. The third section identifies the six factors, from excessive unemployment and eroded collective bargaining to shifts in corporate structures, that we believe much better explain wage suppression. The final section reviews how this paper fits into the overall literature on wage inequality and draws on the estimated impact of the various factors to establish how much they explain the overall divergence between productivity and median hourly compensation as well as the growth of the 50/10 wage gap and changes of the 10th percentile wage.

Wage trends and patterns to be explained

There are three remarkable disparities in growth of wages by workers’ wage rankings that policymakers need to understand and economists need to explain: the one between the highest earners (the top 1% and top 0.1%) and other high-wage earners; the one between high-wage and middle-wage earners (the 95/50 or the 90/50 wage gaps); and the one between middle- and low-wage earners (the 50/10 wage gap). In addition, a theory about wage trends will need to explain the decline in the share of overall income accruing to labor, since this drop saps wage growth; the differing growth rates by educational credential, especially four-year college degrees; the growing divergence between typical workers’ pay (including both wages and benefits) and economywide productivity; and the widening wage gaps that appear between workers of different races, genders, and ethnicities.

Wage growth trends by percentile

Below we delve into some detail in these trends. The rough summary of inflation-adjusted wage growth, detailed below, is as follows. Between 1979 and 2019 (the end of the last business cycle), inflation-adjusted annual wages at the very top have grown tremendously. Those in the top 1% enjoyed 160% growth, and those at the very top—the top 0.1%—experienced growth of 345%. Growth was much slower at the 95th percentile—63% (using hourly wage data), slower still at the 50th (15%), and a snail’s pace at the 10th (3%)—though it is worth noting that growth rates at the middle and the bottom were not remarkably different since the late 1980s. Two key wage gaps have grown since the late 1980s: the one between the top and very top on the one hand and all other earners, including even those at the 95th percentile, on the other, and the gap between high earners and middle earners, illustrated by the ratio of wages at the 95th (or 90th) percentile and the median wage.

Wages at the top—the upper 1% and 0.1%

One of the trends that went unrecognized and unexamined in the mainstream labor economics literature for decades was the remarkable growth for those in the upper 1% and the even faster growth among those in the top 0.1% of wage earners. This growth in wages for the highest-earning households is a key cause of the better-known phenomenon of growth in the household incomes of the top 1%, a finding popularized by Occupy Wall Street and based on the well-known data developed by Thomas Piketty and Emmanuel Saez (2003). The headline Piketty and Saez data are for households (“tax units” to be precise, but importantly they are not for individual workers) and include both wages earned by household members but also other income such as dividends, capital gains, and business income. However, one can analyze data to focus solely on wage and salary income of individual earners. Such an examination also shows superlative growth at the very top and demonstrates the importance of the upward redistribution of wage and salary income to the overall growth of top 1% household incomes.

Figure A, based on Social Security Administration data (see Mishel and Kandra 2020 for details), shows wage growth by the differing wage groups, including those at the top and very top. The Social Security data include all W-2 earnings as reported by employers (including the value of realized stock options and vested stock awards received by executives). Between 1979 and 2019, annual wages for the top 1% grew 160%. The wage growth of the top 0.1%—345%—is listed but not represented on the graph because it is on such a different scale.

Figure A

In contrast, the wages of the bottom 90% grew by just 26% over those 40 years, with most of the growth occurring in a couple of brief and discrete periods of sustained low unemployment. The total wage growth in the 1995–2000 and 2013–2019 periods was 20%, roughly three-quarters of the cumulative wage growth for the bottom 90% over the entire 1979–2019 period. High earners in the 90th to 95th percentiles had wage growth faster than the vast majority (up 52%) but nowhere near that of those at the very top.

The share of all wages earned by those in the top 1% nearly doubled from 1979 to 2019, from 7.3% to 13.2%. Correspondingly, the share of wages earned by the bottom 90% eroded throughout this time, from 69.8% in 1979 to 60.9% in 2019 (Mishel and Kandra 2020). Had this redistribution not taken place, wages for those in the bottom 90% could have grown by 44.6%, 18.5 percentage points more than was the case.2 In short, the redistribution of wages from low and middle to high percentiles has greatly mattered for what the vast majority took home in their paychecks.

Wage gaps for most everyone else: The 95/50 and 50/10 gaps

Outside the top 1%, the clearest growth in wage gaps between percentiles in the wage distribution has been between high earners and middle-wage earners. This is often illustrated by the wage gap between the earners at the 95th percentile and the median (50th percentile) worker. Perhaps surprisingly to some, wage gaps in the bottom half of the distribution of wages (those between middle-wage and low-wage workers) have not grown for more than 30 years (basically since 1987–1988).

Figures B and C show the trends in the 95/50 and 50/10 wage gaps by gender. One striking dimension of the growth of wage inequality is the persistent, continuous growth of the wage gap between those higher up, the 95th percentile, and middle-wage workers since 1979, among both men and women (Figure B). The growth of the 95/50 wage gap has accelerated over time, growing faster in the more recent period (either 1995–2019 or 2000–2019) than it did earlier.3 The growth of the gap between those at the top and the middle of the wage distribution, and its acceleration after the late 1990s, is one of the key wage patterns that need to be explained.

Figure B

In contrast, there has been little change (or an actual decline) in the 50/10 wage gap since the late 1980s (Figure C). The log 50/10 wage gap (using the same data as in the figure) grew from 0.57 in 1979 to 0.75 in 1987 and then declined to 0.68 in 2000 and to 0.65 in 2019. There was a jump in this gap in the early and mid-1980s, especially among women. The fact that there was a much more modest increase in the 50/10 wage gap among men provides an important clue as to the source of this rising gap: the erosion of the real value of the federal minimum wage. Federal inaction on the minimum wage in the 1980s had a much larger impact on low-wage women than low-wage men because women are far more likely to earn wages low enough to be affected by the minimum wage. The importance of institutions (the federal minimum wage) in this case will become a pattern as we look at the data on wage gaps and their causes more generally. The decline in the 50/10 wage gap since 1987–1988 presents a severe challenge to explanations of wage inequality focused on skill or education premiums, since it appears that those with the least education fared somewhat better than those with middling educations. The fact that the 50/10 wage gaps were relatively constant since the late 1980s indicates that a similar, or related, set of forces were at work suppressing wages at both the bottom and the middle.

Figure C

Wage differences by education

The academic and policy debate about wage inequality has often focused on wage gaps across education groups, especially between those with a (four-year) college degree versus those without. Growing education wage differentials have been the basis of the premier conventional explanation of growing wage inequality, which attributes it to “skill-biased technological change.” This theory holds that labor markets are well characterized by competitive models and that an exogenous technology shock—often identified as automation driven by information technologies—has reduced the relative demand for workers without a college degree relative to their college peers, leading to widening inequality between wage earners. The policy corollary to this view has been a call for workers to get college degrees.

The automation-driven skills-gap narrative is critically examined in a later section and in Appendix A. In this section the basic contours of education wage gaps over time are examined. The key metric for capturing education wage differentials is the college–high school premium, or simply the college wage premium, the difference in hourly wages between a college graduate and a high school graduate, controlling for demographic characteristics such as gender, race, and experience. The trend in the college wage premium since 1979 is shown in Figure D.4

Figure D

The college wage premium grew sharply in the 1980s and early 1990s, the era when research arguing for a skills-based explanation of wage inequality grew in prestige. However, growth in the college wage premium has been quite modest since the mid-1990s: It rose by 17.5 log points over the 16 years between 1979 and 1995 but by only 6.4 log points in the 24 years from 1995 to 2019. In annual terms the premium grew 1.1% in the 1979–1995 period but just 0.3% between 1995 and 2019—and just 0.1% per year in the 2000s. The deceleration did not occur because there was an upsurge in the availability of college graduates; in fact, Autor, Goldin, and Katz (2020) show that the relative supply of college graduates grew more slowly in the 1999-2017 period than in the earlier 1979–1999 period.5

It is worth noting that other education wage differentials, such as that between high school graduates and those who left high school and between associate-degree holders and high school graduates, have been fairly stable since 1995. For example, the wage gap, regression-adjusted, between those with a high school degree and those not completing high school (or a GED) rose only slightly between 1979 and 1987—up 2.6 log points—hardly sufficient to explain the sharp rise in the 50/10 wage gap over that period. Between 1987 and 2018 this wage gap shrank by 3.0 log points, reversing the entire 1979–1987 rise and leaving the wage gap where it was in 1979. Since there was hardly any growth in education wage gaps among the bottom three education groups, any explanation for wage gaps in the bottom half that relies on education gaps has little explanatory power.

The gap between productivity and median hourly compensation

The last four decades have seen a systematic divergence between the growth of economywide productivity (the amount of income generated in an average hour of work) and the growth of hourly compensation (wages and benefits) for typical workers. We proxy the wages of “typical” workers as either wages for nonsupervisory workers (roughly 80% of the private-sector workforce) or wages for the worker earning the median wage. Figure E, which shows the growth of productivity and the typical worker’s hourly compensation since 1948, uses the hourly compensation of production-nonsupervisory workers because that is the only series available for the entire period since 1948.6 While productivity and a typical workers’ compensation grew in tandem over the 1948–1973 period, they diverged thereafter, splitting entirely after 1979. In the latter period productivity decelerated significantly, but much more rapid deceleration (or even stagnation) occurred in a typical worker’s compensation. Productivity grew 108.1% from 1948 to 1979, accompanied by 93.2% growth in a worker’s compensation. Between 1979 and 2018 productivity grew 69.6% (1.2% annually) further, but a typical worker’s compensation (wages and benefits) grew only by 11.6% (0.24% annually).

Figure E

This divergence was first pointed out in the early 1990s (Mishel and Bernstein 1994) to demonstrate that stagnant wages for the typical worker over the previous decade or so could not be explained solely by the slowdown of productivity growth. Bivens and Mishel (2015) examined the wedge between typical workers’ pay and productivity and decomposed the main factors generating it. Once technical influences like differing price deflators were accounted for, the remaining wedge between pay and productivity growth was driven by two facets of inequality: the decline in labor’s share of income (particularly in the 2000s) and the growth of inequality of compensation, such that compensation grew far faster at the very top, as shown above.7 In other words, the factors driving inequalities in the labor market are responsible for workers’ inability to make gains commensurate with productivity growth. Whether workers make wage gains commensurate with future productivity growth will depend on whether we prevent this ongoing, and eminently preventable, growth in wage inequality. These dynamics will, in turn, dictate the extent of inequalities across households.

What has this divergence meant for a typical worker’s compensation? Building on the Bivens and Mishel (2015) analysis, we calculate that between 1979 and 2017 economywide productivity rose 68.1% while median hourly compensation rose 13.0% (the median hourly wage rose 12.2%), yielding a 55.2 percentage point divergence. Excluding the impact of differing price deflators (using the same index to deflate both productivity and median hourly compensation) yields a 43 percentage point divergence. Thus, the pay for typical workers would be more than 40% higher today if inequality had not risen over the 1979–2017 period. In the final section we examine the impact of particular factors to gauge whether their cumulative impact can explain the productivity–median compensation divergence. We use the 1979–2017 period as the benchmark, since that period corresponds to some key estimates of the impact of particular factors on median wages.

The decline in labor’s share of income

One of the trends that alerted analysts to the erosion of worker bargaining power and the corresponding strengthening of employer bargaining power has been the erosion of labor’s share of income in the 2000s. The distributional conflict between workers and employers (or capital and labor shares) is best examined in the corporate sector, where all income is divided between compensation going to workers and income accruing to owners of capital. Focusing on the corporate sector hence avoids issues of having to decide whether some other form of income—”proprietor’s income,” or income of noncorporate businesses—is labor or capital (see Bivens 2019 for measurement details).

The trend in the labor share of corporate-sector income is presented in Figure F.

Figure F

The data clearly show a lowering of labor’s share in the 2000s, even when cyclical ups and downs are factored in. In business cycles before the 2000s, labor’s share generally rose during recessions, as profits fell, but then dropped sharply in early recoveries, as high unemployment weighed on wage growth even as the economy and profits recovered. By the time of each business cycle peak, labor’s share had generally recovered its previous peak. After 2000, though, labor’s share has often not recovered its previous peak following recessions. For example, it fell from 82.4% in 2000 to 77.9% in 2007, the last year before the Great Recession, and by 2016, when unemployment had reached levels comparable to what had prevailed in 2006 and 2007, it remained roughly 2.5 percentage points below its 2007 level. The fall in labor’s share from 82.4% in 2000 to 75.5% in 2016 is the equivalent of an 8.4% across-the-board cut in compensation for every employee; equivalently, it would require an across-the-board compensation boost of 9.1% to restore labor’s share to its 2000 level.8 This computation may exaggerate the impact of labor’s falling share, since 2000 was a near high point for it historically, driven in part by unemployment falling to its lowest level in decades (4.0%). However, the unemployment rate in 2018 and 2019 also averaged below 4.0%, and labor’s share ended 2019 at 77.3%, well below 2000’s level. This shift toward greater capital income and returns is even more impressive given that real interest rates have fallen sharply in recent years, a development that should (all else equal) be accompanied by a lower return to capital (Farhi and Gourio 2018).

Wage disparities by race and gender

Race and gender wage disparities have been persistent and insidious and reflect key dynamics that need to be understood in any analysis of the labor market.

Most measures of overall gender wage inequality have fallen over the last four decades. Figure G traces the gender wage gap over 1979–2019 in two ways, the percentage gap between the male and female median wage and the unexplained (log) percent of the gender wage gap (on average) after controlling for race and ethnicity, education, age, and geographic division. By both measures the wage disparities between men and women have substantially declined. The gap at the median has been cut in half, from 36.7% in 1979 to 15.8% in 2019, and the gap that is unexplained fell from 37.7 to 22.6 log points.

Figure G

Conversely, most measures of Black–white wage disparities have grown in the last four decades (Figure H) and have exacerbated the growth of overall wage inequality (Wilson and Rodgers 2016; Wilson 2020). The increase has not occurred along a straight line, however. During the early 1980s rising unemployment, declining unionization, and policies such as the failure to raise the minimum wage and lax enforcement of anti-discrimination laws contributed to the growing Black–white wage gap. During the late 1990s the gap shrank in part due to increases in the minimum wage and in part to tighter labor markets, which made discrimination more costly to employers. Since 2000 the gap has grown again. As of 2015, relative to the average hourly wages of white men with the same education, experience, metro status, and region of residence, Black men made 22.0% less and Black women 34.2% less. Black women earn 11.7% less than their white female counterparts. The median Black worker’s hourly wage rose only 5.3% from 1979 to 2019, lagging far behind the 20.0% hourly growth enjoyed by the median white worker. The 7.6 log point growth of the “unexplained” portion of the Black–white average wage gap suggests that the economic bite of discriminatory pay practices has grown and accounts for a large share of the relative wage growth disadvantage of Black workers.9

Figure H

The causal connection between racial, gender, and ethnic wage growth differences and wage growth differences by percentiles likely runs in both directions. On the one hand, because white women and Black workers tend to be disproportionately concentrated in the bottom 90% of the wage distribution, any force that suppresses wages for the bottom 90% will make them lose ground to white men and whites overall, all else equal. In a study of racial differences in household income, for example, Manduca (2018) finds that the percentile rank of the median Black household in the overall household income distribution climbed rapidly between 1968 and 2016, from the 24th to the 36th percentile. Yet because inequality concentrated gains in the top half of the income distribution over that time, and because white households were disproportionately concentrated in the top half, the income gap (measured in dollars or as a percentage) between white and Black households actually grew. So, a redistribution of income or wages to top and very top earners can increase gender and racial inequities even when Blacks and women achieve better equity among the bottom 90%.

Conversely, racial and gender differences in pay could in theory feed through to measures of inequality by percentile. For example, if the gap in wages paid to otherwise-equivalent Black and white workers can be pocketed entirely by the employers of Black workers, then this pay-setting discrimination would translate one-for-one into lower measured wages of typical workers of all races. If, on the other hand, pay-setting discrimination reflects employer preferences for white workers over Black workers, then some of this Black/white wage gap is a de facto transfer from Black workers to white workers. Less mechanically, if the maintenance of racial and gender wage gaps fosters division and hamstrings collective cross-race and cross-gender efforts to raise wages in workplaces, then these gaps could be a powerful force for employers looking to keep overall wages suppressed.

The failure of automation and skill gaps to explain wage suppression or wage inequality

That a huge proportion of U.S. workers have “skills deficits,” i.e., lack the skills necessary to deal with technological change (including primarily automation of the tasks performed by workers without a four-year college degree), has been the predominant explanation offered by economists, pundits, policymakers, and the media to explaining sluggish wage growth and inequality in the United States, at least until recently (see Box A for why we focus on “automation” rather than a more general “technological” explanation). This is the skill-biased technological change hypothesis, which points to the increased use of computer equipment in the workplace and the onset of the information age. One version, focused on education wage gaps, argues that computerized automation has made more-educated workers—generally referring to those with at least a four-year college degree—more valuable to employers and has correspondingly reduced the value of those without a college degree (Katz and Murphy 1992; Goldin and Katz 2007, 2008). This growing wage gap between college-educated and non-college-educated workers—the college wage premium—is used to explain rising wage inequality between high earners and the majority of earners who lack a four-year college credential (62% of earners in 2019, down from 82% in 197910).

A second version of the automation story, frequently referred to as the “job polarization thesis,” argues that technological change

has increased the value of abstract reasoning, of creativity, of expertise, of judgment and devalued a lot of skilled work that people did that followed well-understood rules and procedures. So that would be many clerical jobs, phone answering jobs, calculating, accounting, bookkeeping, copying, and filing, but also many production jobs, which often involved skilled, repetitive tasks. But increasingly once we understand the rule book for that type of work, it’s feasible to encode it in software and have it executed by machines or by computers.” (David Autor, interviewed by Greenhouse 2020)

This type of automation sees the replacement of human labor by computers or other modes of automation as not just falling along the lines of greater/fewer education credentials (whether a worker has a college degree or not). Instead, occupational susceptibility to having tasks automated is said to be “polarized”—middle-wage occupations specializing in routine, automatable tasks shrink while employment expands in both higher-wage “abstract reasoning” occupations and in lower-wage non-routinized occupations.

These versions of skill-biased technological change portray the cause of wage suppression and wage inequality as due to a factor, automation, that is both inevitable (one can’t stop technology’s forward march) and desirable (after all, technological change is a key driver of rising living standards). Thus, the resulting economic adversity for some workers is the unfortunate byproduct of a dynamic that one would neither want to nor could change. Given this view, the only appropriate remedy is to adapt to automation, primarily by upgrading workers’ skills and education and perhaps by providing a more adequate safety net for workers temporarily displaced.

The skills narrative is the premier explanation of wage inequality that relies on the existence of competitive labor markets and sees a role for policy only in fostering skills development—not in boosting workers’ bargaining power or mandating changes in corporate practices. As such, the failure of the skills narrative to explain wage suppression and wage inequality also reflects the failure of analyses based on equal bargaining power between employers and employees and competitive labor markets to adequately address one of the most salient features of the economy over the last four decades.

As examined in detail in Appendix A, skill-biased technological change has always been a weak explanation for the wage trends since 1979, but is a prima facie implausible explanation for the trends since the mid-1990s or since 1999. None of the basic indicators of automation’s impact and of skill deficits used to establish these narratives has been evident over the last 25 years. Consequently, there is no basis for considering automation-driven skill-biased technological change as a significant factor in wage suppression or the growth of wage inequality since the mid-1990s—and we assign it an impact of zero since 1995 in our analysis below.

The extreme weakness of the hypothesis is why the consensus, at least of center-left economists, no longer highlights it in discussions of wage problems. Lawrence Summers went so far as to say:

I am concerned that if we allow the idea to take hold that all we need to do is there are all these jobs with skills and if we just can train people a bit then they will be able to get into them and the whole problem will go away. I think that is fundamentally an evasion of a profound social challenge. (Summers 2015, 75–76)

The following sections, drawing on Appendix A, offer some critiques of the skills narrative, focusing particularly on its inability to explain wage trends since the mid-1990s.

Omissions in the skills-based wage-gap story

One problem with the automation narratives is that they ignore, or perhaps fail to address, important dimensions of wage suppression and wage inequality. For instance, they sidestep the superlative wage growth of the top 1% (and the top 0.1%) and the corresponding upward shift of 6 percentage points of aggregate earnings to the top 1% between 1979 and 2019 (Mishel and Kandra 2020). The growth of wages for the top 1% primarily reflects the growth of executive compensation and the expansion of the financial sector (and its high earners). Similarly, the narratives accord no attention to the erosion of labor’s share of income, and the data show that the link between automation and the decline of labor’s share is at least as inconsistent with real-world data as is the link between automation and wage inequality (Stansbury and Summers 2020). These are substantial oversights.

Evidentiary problems with the skills-gap, or automation, hypothesis

Recent research by the leading proponents of the skill-biased narrative (Autor, Goldin, and Katz 2020) demonstrates the failure of this hypothesis to contribute to our understanding of wage suppression since the mid-1990s. The three biggest evidentiary failures are the following.

There has been little growth in the college wage premium

Autor, Goldin, and Katz (2020) note that “returns to a year of college rose by 6.5 log points, from 0.076 in 1980 to 0.126 in 2000 to 0.141 in 2017.” Yet note the slowdown from the 1980–2000 period to the 2000–2017 period. In the former period, the log college wage premium rose 0.0325 percentage points each year, far faster than the 0.0088 percentage point increase each year between 2000 and 2017. This represents a 70% reduction in growth. Given that the log 95/50 wage gap grew faster between 2000 and 2019 than in 1979–2000 (see Appendix Figure A), it is evident that the education wage gap has not driven wage inequality in the top half since 2000.

There has been a sharp deceleration in automation-driven relative demand for college graduates since the mid-1990s

The substantial deceleration in the college wage premium, even as the supply of college graduates slowed, implies a dramatic slowing in the growth of relative demand for college graduates. As Autor, Goldin, and Katz (2020) note:

[T]he model’s results…divulge a puzzling slowdown in the trend demand growth for college equivalents starting in the early 1990s. Rapid and disruptive technological change from computerization, robots, and artificial intelligence is not to be found though the impact of these technologies may not be well captured by this two-factor setup.

Their results (based on Autor, Goldin, and Katz, Table A2) show a deceleration in growth of relative demand for college graduates in the 1999–2017 period relative to earlier periods: a 45.8%  deceleration relative to the 1979–1999 period and a 41.8% deceleration relative to the longer 1959–1999 period. The period since 1999, therefore, has been one featuring a historically small impact of automation on (relative) demand for college graduates.

Autor (2017) provides similar findings in an update of the Katz and Murphy (1992) metric of relative demand for college graduates and dates the slowdown to the mid-1990s.

If automation’s impact has been far less in the last 25 years than in earlier decades, it cannot explain the ongoing strong, even faster, growth of wage inequality in the top half, illustrated by the growth of the 95/50 wage gap.

There has been no expansion of the wage gap in the bottom half since 1987

In the skills-gap story, the more education workers have, the more they are in demand and the higher their wages. Yet over the last three decades there have been no increases in the wage gaps between those with some college, those with a high school diploma, and those who left high school. Similarly, that the wage gap between median (50th percentile) workers and low-wage (10th percentile) workers has been stable or declining since 1987 stands as a long-standing critique of the education wage-gap hypothesis (Mishel, Bernstein, and Schmitt 1997a; Card and DiNardo 2002; Acemoglu and Autor 2012).

Evidentiary problems with claims that ‘occupational polarization’ of labor demand can explain wage patterns

As mentioned earlier, a recent twist on the traditional story of skills-driven wage trends is the idea of labor market “polarization.” In this view, labor demand has not been rising commensurately with every level of education; rather, there has been an increase in labor demand for high-skill occupations but also for nonroutine manual occupations generally associated with low wages. The losers are those with routine, nonabstract skills that have traditionally been associated with middle-wage occupations, such as in manufacturing.

There are two clear empirical problems with the polarization narrative as an explanation for the wage trends over the full 1979–2019 period.

There has been no occupational employment polarization since 1999

Most strikingly, Autor (2010) and Acemoglu and Autor (2012) document that job polarization has not been evident since 1999:

[G]rowth of high-skill, high-wage occupations (those associated with abstract work) decelerated markedly in the 2000s, with no relative growth in the top two deciles of the occupational skill distribution during 1999 through 2007, and only a modest recovery between 2007 and 2012. Stated plainly, the U-shaped growth of occupational employ­ment came increasingly to resemble a downward ramp in the 2000s. (Autor 2014)

A “downward ramp” and the absence of a “U-shaped growth of occupational employ­ment” amount to an acknowledgment, although offered in a less-than-direct fashion, that job polarization was not present between 1999 and 2012.

Occupational employment patterns do not correspond to wage patterns

The job polarization narrative relies on mapping occupational employment patterns to explain wage trends. But surprisingly, the polarization literature has never presented evidence that these occupational employment shifts directly affect wages. And indeed, they don’t. Mishel, Shierholz, and Schmitt (2013) show that in the 1980s, 1990s, and 2000s, changes in occupational employment shares (whether employment in an occupation expands or contracts relative to other occupations) were not related to changes in relative wages by occupation (whether wages rose or fell in that occupation relative to wages of other occupations). It is also worth noting that middle-wage occupations have shrunk and higher-wage occupations have expanded since the 1950s, but median wages and wage inequality have risen and fallen over this time with no apparent correspondence to employment polarization trends.

If occupational employment patterns do not directly correlate with occupational relative wages, then the narrative that employment polarization reflects a shift in the relative demand for specific types of skills makes little sense; shifting relative labor demand curves should cause both employment and wages to move together. Moreover, if occupational job polarization does not shape relative occupational wages, then it is certainly not much of an explanation for wage inequality.

A more convincing theory of sluggish wage growth and inequality in the U.S.: Policy-driven wage suppression

If forces unrelated to policy decisions, particularly automation, do not seem to be driving wage trends, what are the factors leading to wage suppression? Our answer is that there has been an intentional policy assault—including policy forbearance in the face of new anti-worker business practices—that diminished the institutional sources of leverage and bargaining power for typical workers in the labor market. The point was to suppress labor costs. This policy assault (acts of commission and omission, such as failing to update labor law or the value of the minimum wage) either directly undercut these institutional sources of power or accommodated employers’ efforts to undercut them. Business forces were secure knowing that policymakers (legislators, executive branch officials, and judges) would not change legislation, enforcement priorities and effectiveness, or legal interpretations to countermand this assault on a typical workers’ power in the labor market.

Why this policy and corporate assault began when it did, and why it was successful politically, are questions mostly outside the bounds of this paper. (See two relatively recent books by Hacker and Pierson (2011, 2020) for the political science explanations.) Among academic and policy economists starting in the 1970s, policies promoting wage suppression were forwarded as the solution to decelerating productivity growth, accelerating inflation, and high unemployment, where the goal was to improve aggregate efficiency and growth. But it did not work: Economic growth between 1979 and 1995 was historically slow, and the brief pickup between 1995 and 2000 was driven by a burst of business spending to connect to the internet and the decision by the Federal Reserve to shunt aside inflation fears and allow unemployment to fall to historically low levels. Slow productivity growth reasserted itself quickly in the 2000s as the burst of information-technology investment receded.

The opinions of academic economists aside, the larger political dynamics of why this wage-suppression campaign began and became a key priority for corporate and business interests in the 1970s are clearly important. Levy and Temin (2007), Bivens (2010), and Mishel, Rhinehart, and Windham (2020) provide an overview of this period and document the concerted shift of the corporate sector away from the “Treaty of Detroit” regime established in the early 1950s that tolerated (or even encouraged) collective bargaining and toward policies maximizing shareholder value and deregulation. The falling profitability of the corporate sector in the 1970s was clearly a part of this dynamic, as was the rising ease of moving production offshore.

While the root causes and the political support behind the policy assault on bargaining power are not addressed here, a growing body of evidence shows that the specific policies launched in this attack can explain the overwhelming majority of wage suppression experienced in recent decades. This section examines these policies and estimates the impact of each. When added together, as we do in the final section, the impact is large enough to explain the lion’s share of the 43% difference since the late 1970s between the growth of median hourly compensation and economywide productivity.

Austerity macroeconomic policy: Excessive unemployment

The Federal Reserve Board’s dual mandate is to pursue the maximum level of employment consistent with stable inflation. However, since 1979 the Fed’s actions suggest that it took the inflation mandate more seriously, thereby tolerating (by failing to lower) or actually generating excessive unemployment for extended periods in the name of keeping inflation tame. Whenever an economic expansion pushed unemployment down, the Fed often feared that tighter labor markets would mean that workers, endowed with more leverage since they were now in a better position to quit or strike, would demand higher nominal wages, in turn putting upward pressure on inflation.

Wage growth resulting from tight labor markets can indeed feed into price growth, and so sufficiently empowered workers may demand even higher wages, allowing wage/inflation momentum to build. The policy recourse for stopping the wage/price spiral has traditionally relied on the Fed raising interest rates to slow the expansion and stop the downward movement of unemployment.

Presumably in this policy vision there is a sweet spot where workers can experience decent wage growth without fostering unsustainable inflationary pressure. But nobody knows for sure beforehand where that level is, and efforts to empirically identify the economy’s “natural rate of unemployment” are notoriously imprecise (Staiger, Stock, and Watson 1997). Given this uncertainty, the Fed must exercise judgment in weighing the benefits of tighter labor markets against the risks of inflationary pressure. Too often in the post-1979 period, Fed policymakers have been so worried about the inflation risks and not impressed enough by the benefits of full employment that they have raised interest rates prematurely and cut expansions short before they generated decent wage growth. The result has been unemployment higher than it had to be to ensure stable inflation.

Historically, the anti-inflation orientation of the Fed was quite political and conscious of the institutional determinants of wage growth. Specifically, past Fed chairs, determined to keep wage growth “moderate,” explicitly saw the use of high unemployment as a means to restrain union-negotiated wage increases or even to seek union wage concessions.

Mitchell and Erickson (2005) characterized this policy orientation during the era of Paul Volcker’s chairmanship of the Fed (1979–1987), when there was a deep recession and a sharp reduction in unionization (40% of the erosion of unions over the 1979–2017 period occurred in 1979–1984):

Volcker viewed affecting union wage determination through monetary restraint as important for the Fed’s disinflation campaign. One commentator characterized the Fed chair’s view as founded on the idea that “inflation would not be securely defeated…until all those workers and their unions agreed to accept less. If they were not impressed by words, perhaps the liquidation of several million more jobs would convince them.” …Others at the Fed apparently had similar wage-push ideas. To Volcker, direct intervention in particular wage settlements was not desirable (and clearly not the province of the Fed). But a monetary squeeze that forced the union sector to hold down nominal wages in the hopes of preserving jobs was an appropriate policy instrument. Squeeze the unions and other wages (and prices) would fall into line.

Hooper, Mishkin, and Sufi (2019, 25) note that “since the 1980s the Fed focused much more on avoiding labor market overheating in order to stabilize inflation.” The discussion in Appendix B provides a further analysis of the Federal Reserve policy that maintained excessive unemployment.

Bivens and Zipperer (2018), analyzing the links between excess unemployment and wage growth, note that full employment (at least by the too conservative measure of matching actual unemployment to preexisting estimates of the “natural rate”) was the norm after World War II but became the exception after 1979. Between 1949 and 1979, the cumulative difference between the actual unemployment rate and estimates of the unemployment rate consistent with stable inflation—the “natural rate” or the NAIRU, the nonaccelerating inflation rate of unemployment—was negative 15.3 percentage points, meaning that on average actual unemployment was 0.52 percentage points below the estimated NAIRU each year. In contrast, between 1979 and 2017 the cumulative difference was positive 35.7 percentage points, meaning that actual unemployment was persistently above the estimated natural rate. This trend was not driven only by the Great Recession: Between 1979 and 2007 the cumulative difference was a positive 15.5 percentage points. Put another way, unemployment was 1 percentage point higher each year in 1979–2007 than in 1949–1979.

This consistent excess unemployment was deeply damaging to wage growth. Research indicates that a 1 percentage point drop in unemployment results in annual wage growth 0.5–1.5 percentage points faster for workers at the 10th percentile. For example, if annual real wage growth is 1%, then a 1 percentage point fall in unemployment would result in annual real wage growth rising to 1.5% to 2.5%. For workers near the median of the wage distribution, wage growth is faster by 0.4–0.9 percentage points, and for workers at the 90th percentile it is 0.3–0.5 percentage points faster. These estimates indicate that excessive unemployment generates increases in both the 50/10 and 90/50 wage gaps.11

Using the lower bound of the Bivens and Zipperer estimates to assess the impact of excessive unemployment on median and 10th percentile wages in the 1979–2017 period, we find excessive unemployment had lowered the median hourly wage by 12.2%. These estimated impacts of unemployment are far below those of Katz and Krueger (1999, Table 8), whose Phillips curve estimates using a 1973–1998 time series were double those of Bivens and Zipperer at the median and three times those at the 10th percentile.

However, to err on the side of caution we make an adjustment to our estimates of the wage impact of higher unemployment to account for the “flattening” of the Phillips curve in recent years (a lessening of the relationship between unemployment and wage growth): We apply one impact for the 1979–2007 years and a lesser impact for the 2008–2017 years.12 The coefficient on unemployment levels in regressions explaining wage growth is notably smaller in the time period after 2008.13 It should be noted that, as Hooper, Mishkin, and Sufi (2019) found, the “wage-Phillips curve…flattened significantly less and retained greater nonlinearity” than the price Phillips curve. However, one should be clear what is driving the flattening of the wage Phillips curves in the post-2008 period: the failure of high unemployment to force actual reductions in nominal wages in just a small number of years (2008-2014, mostly), a finding highlighted in Bivens (2019). If one removes this subset of years from the sample, the regression coefficient for the overall post-1979 sample is far closer to what one obtains from running the regression over the period from 1979 to 2007. It seems clear that this is the issue of “downward nominal wage rigidity bending the Phillips curve,” as expressed by Daly and Hobijin (2014). This implies that we are overcorrecting in our markdown of the impact of unemployment on wages post-2008; that is, we are understating the negative impact of high unemployment.

Taking this flattening of the Phillips curve into account we find that, if unemployment over 1979–2017 had averaged just the “natural rate” of 5.5% rather than 6.3%, median wages would have been 10.0% higher in 2017. If the unemployment rate had been held even lower, say 5.0%, median wages would have been 18.3% higher by 2017.

About a fourth of this impact is the result of the corrosive effects of the extended period of high unemployment following the Great Recession. In those years, the real culprit for keeping labor markets too soft to generate decent wage growth was not the Federal Reserve Board (the Fed actually tried hard to boost economic growth in those years). Instead, fiscal policymakers—both the Republican-led Congress and state and local governments after 2010—imposed historically contractionary degrees of spending austerity.

As analyzed in the final section, excessive unemployment’s impact on the median wage of 10.0% can explain nearly a fourth of the productivity–median compensation divergence of 43.0% and is a sizable factor expanding the wage gaps in both the bottom and top halves of the wage structure. For instance, excessive unemployment had a larger impact on low-wage workers, lowering the 10th percentile wage by 11.6% by 2017 and raising the 50/10 wage gap by 2.7 percentage points. If our analysis uses 5.0% rather than 5.5% as the full employment target, then the 10th percentile wage would have been 21.2% higher in 2017 absent excessive unemployment.

Because Black workers are disproportionately concentrated in the bottom half of the wage distribution, the corrosive effect on wage growth of excess unemployment falls more heavily on them. Wilson (2015) has found that the wage-depressing effect of each additional percentage point of unemployment is roughly twice as large for the median Black wage as for the median white wage. Wilson and Rodgers (2016) find that the wage penalty for the median Black worker relative to the median white worker increased by nearly 9 percentage points between 1979 and 2014. This growth in the Black–white wage gap could be entirely explained by excessive unemployment over this period, given the differing responsiveness of Black and white wages to unemployment.

Erosion of workers’ rights to form unions and bargain collectively

The erosion of collective bargaining has been a major factor that has depressed wage growth in the middle and drove the growth of wage inequality over the last four decades. In fact, the only factor having a larger impact is the excessive unemployment perpetrated by policymakers. The impact has been especially adverse for men because they were far more likely to be unionized in 1979 than women (31.5% versus 18.8%), so men had more to lose from the subsequent attack on unions and collective bargaining.14

That collective bargaining leads to more equal wage outcomes was firmly established by research by Richard Freeman and James Medoff in the late 1970s and popularized in their important book, What Do Unions Do?, published in 1984 (Jake Rosenfeld’s 2014 book, What Unions No Longer Do, provides an update of the issues). Consider first the ways that collective bargaining leads to more equal wage outcomes among unionized workers and in unionized industries and occupations.15 First, unions make wage differences between occupations more equal because they give a larger wage boost to low- and middle-wage occupations than to high-wage occupations. Second, unions have boosted wages for low-wage workers the most, and increases have been larger at the middle than at the highest wage levels, larger for Black and Hispanic workers than for white workers, and larger for those with lower levels of education. This pattern of wage increases narrows wage inequalities. Third, unions make wages of workers with similar characteristics more equal in union settings because wages are “standardized,” meaning that wages are set for particular types of work and do not vary across people doing the same work to the same degree as exists in nonunion settings. Fourth, unions have historically been more likely to organize middle-wage than high-wage workers, which lowers inequality by closing gaps between, say, blue-collar and white-collar workers.16 The union impact on inequality is even greater with compensation than with wages alone (Pierce 1999).

Research from the early 1990s documented that the erosion of collective bargaining was responsible for around a fifth of the rise in wage inequality among men in the 1980s (Card 1991; DiNardo, Fortin, and Lemieux 1996; Freeman 1991) but had a more modest impact on women’s wage inequality.

More recent research has incorporated an assessment of the impact of unions on nonunion workers’ wages—sometimes referred to as “spillover effects”—and finds a much larger impact. When the share of workers who are union members is relatively high, as it was in 1979, wages of nonunion workers are higher. For example, had union density remained at its 1979 level, weekly wages of nonunion men in the private sector would have been 5% higher in 2013, equivalent to an additional $2,704 in earnings for year-round workers; among those same workers but without a college education wages would be 8% higher, or $3,016 more per year (Rosenfeld, Denice, and Laird 2016; Denice and Rosenfeld 2018).17 Consequently, estimates of the impact of eroded collective bargaining on wage inequality that incorporate union spillover impacts find a larger role of the impact of unions on wage inequality. For instance, Western and Rosenfeld (2011, Table 2 and analyzed in Mishel et al. 2012, Table 4.38) find that the weakening of collective bargaining explains a third of the increase in male wage inequality and a fifth of the rise of wage inequality among women over the 1973–2007 period. Such research demonstrates that the erosion of collective bargaining has been the largest single factor driving a wedge between middle- and high-wage male workers, which, as established above, is the main dimension of wage inequality that grew among men other than the soaring of earnings for the top 1% (keep in mind that the wage gap between middle- and low-wage men has not grown).

The most recent research provides a more up-to-date analysis that incorporates a spillover impact and provides additional insight because the results directly report on the impact of eroded collective bargaining on the wage gap between high-wage (90th percentile) and middle-wage (50th percentile) workers by gender. These results from Fortin, Lemieux, and Lloyd (2021) are shown in Table 1.

Table 1

The erosion of unions (column 3) can explain from 28.8% to 36.7% of the growth of male wage inequality as measured, respectively, by the standard deviation of log wages and the Gini coefficient. The most salient finding is that eroded unionization explains 37.3% of the growth of the 90/50 wage gap over the 1979–2017 period. As discussed earlier, this is the only source of growing wage inequality among men in the bottom 90% of earners. Fortin, Lemieux, and Lloyd (2021) show a smaller impact of eroded unions on women’s wage inequality; the erosion explains 6.7% to 8.8% of the growth of women’s wage inequality as measured by the standard deviation of log wages and the Gini coefficient and only 13.0% of the growth of the 90/50 wage gap.

Using an (unpublished) analysis by Thomas Lemieux, who relied on the model in Fortin, Lemieux, and Lloyd (2021), we can examine the impact of deunionization on the median and the 90th percentile wages for men and women combined over the 1979–2017 period. Deunionization raised the log 90/50 wage gap by 7.7 log points, almost entirely by reducing the median hourly wage by 7.6 log points, or by 7.9% (0.2% annually). The impact on men alone is larger, with deunionization lowering the male median wage by 10.9 log points, or 11.6% (0.29% annually). Deunionization therefore explains 33.1% of the 23.2 log point growth of the 90/50 wage gap over the 1979–2017 period.

The Fortin, Lemieux, and Lloyd (2021) results are very comparable to those of Stansbury and Summers (2020). In particular, we have benchmarked the Fortin, Lemieux, and Lloyd estimates of the impact of deunionization on the median wage with the estimate of Stansbury and Summers of the impact on the non-college-educated wage. Once one aligns the time periods, one finds that the Stansbury and Summers estimate is even larger.18 The Fortin, Lemieux, and Lloyd results for how much deunionization explains the growth of wage inequality also align with those by Western and Rosenfeld (2011).19

Newly developed historical data from the early postwar period affirm that collective bargaining was a strong force for greater equality of wages. For instance, Callaway and Collins (2017), using data from a survey of men living in Philadelphia; New Haven, Conn.; Chicago; St. Paul, Minn.; San Francisco; and Los Angeles in 1951, found “the [union] wage premium was larger at the bottom of the income distribution than at the middle or higher, larger for African Americans than for whites, and larger for those with low levels of education,” findings that are “consistent with the view that unions substantially narrowed urban wage inequality at mid-century.” It follows, of course, that the consequent erosion of collective bargaining would increase wage inequality and have the most adverse impact on nonwhite workers, those with the least education, and low- and moderate-wage workers.

Likewise, Farber et al. (2021, 33), who developed data on union households from Gallup surveys going back to 1936, find that “mid-century unions [were] a powerful force for equalizing the income distribution.” This happened because unions disproportionately represented “disadvantaged” workers (nonwhite or less educated), raised the wages of low- and moderate-wage workers the most, and had a large, stable impact of raising wages for union workers by roughly 15–20 log points over the last 80 years.20

Some pundits and analysts, skeptical about the impact of weaker unions on wages or wage inequality, claim that the decline of unions reflects a decline in worker interest in unions or is due to globalization and automation, i.e., endogenous factors. Neither objection is well founded.

Kochan et al. (2018) examined the level of interest in joining a union among unorganized workers and found that the “demand for unions” has risen substantially since the late 1970s. Kochan and Kimball (2019) described these results as

differences in the percentage of non-union workers who indicated a preference for union representation in nationally representative surveys in 1977, 1995, and 2017. Note that the 1977 and 1995 results were nearly identical: approximately one third of the non-union workforce indicated they would vote to have union representation if given an opportunity to do so on their current job. In 2017 that number increased to 48 percent. This number translates into an under-representation of unions of approximately 58 million workers.

Mishel, Rhinehart, and Windham (2020) assess the endogeneity of union decline and find that manufacturing employment decline can account for only a small part of it, perhaps 15–20%. Ongoing automation of manufacturing and globalization surely contributed to the shrinkage of manufacturing employment and the loss of union jobs in manufacturing, but the authors point out that the share of workers covered by collective bargaining declined strongly across the private sector in sectors not heavily affected by globalization, including construction, transportation, communications, utilities, supermarkets, hotels, and mining. So, any focus limited to manufacturing will not capture the full picture. Moreover, the erosion of unions in manufacturing is not due only to automation and globalization.

An Organisation for Economic Co-operation and Development (OECD 2019b) analysis of the cross-country decline in collective bargaining across advanced nations found:

Contrary to a commonly held belief, the combined contributions of demographic changes and structural shifts, such as the shrinking of the manufacturing sector, are small and leave most of this declining trend [in collective bargaining] unexplained.

This confirmed an earlier analysis by Schmitt and Mitukiewicz (2012) that: “The observed patterns suggest that national politics are a more important determinant of recent trends in unionization than globalization or technological change.”

The primary reason that collective bargaining eroded was a concerted corporate attack on unions starting in the 1970s that exploited weaknesses in our labor laws to suppress the ability of workers to choose collective bargaining and organize (Windham 2017). The scale of union organizing collapsed, dramatically, in the 1970s as the share of nonagricultural workers in private-sector National Labor Relations Board elections fell from 1.0% to 1.2% each year in the 1950s and 1960s to just 0.3% each year in the 1980s and to 0.1% each year in the early 2000s (Mishel, Rhinehart, and Windham 2020). As Windham (2017) documents, this collapse of organizing was due to increased employer aggressiveness and use of both legal and illegal tactics, including captive audience meetings, threats of shutdowns or relocation, firing of union organizers, use of a rapidly expanded group of anti-union consultants, and process delays.

McNicholas et al. (2019), analyzing union representation elections that took place in 2016–2017 and building on earlier work by Bronfenbrenner (2009), have recently documented the pervasive lawlessness prevailing in union organizing attempts:

Employers are charged with violating federal law in 41.5% of all union election campaigns. And one out of five union election campaigns involves a charge that a worker was illegally fired for union activity. Employers are charged with making threats, engaging in surveillance activities, or harassing workers in nearly a third of all union election campaigns.

Other developments, enabled by employer aggressiveness and changes in the rules governing collective bargaining, limited collective bargaining power by eroding coverage and weakening unions’ ability to strike (Mishel, Rhinehart, and Windham, 2020).

Managing globalization on capital’s terms

Evaluating the impact of globalization on wages has been controversial because it has inevitably been intertwined with the politics of trade agreements. Some analysts contend that globalization has a barely measurable impact, while others see it as an all-encompassing force shaping everything, limiting our ability to have higher wages or pursue regulatory or tax policies. In fact, globalization has indeed put substantial downward pressure on wages for the vast majority, but its effects are likely not just driven by the mechanical impacts it has on demand and supply curves for different types of labor in a competitive model. Instead, it is as likely that globalization has played a powerful role in disempowering workers and giving capital owners and managers a much-improved fallback position in their bargaining with workers.21 As such, the way that globalization has proceeded from U.S. workers’ perspective has been profoundly shaped by intentional policy decisions that maximized its wage-suppressing effects.

Bivens (2017a) presents a summary of globalization’s wage impacts, based on his own calculations and on the wider economics literature. He finds:

The challenge of globalization for American workers is often painted as a problem of industrial workers losing their jobs to imports. Because manufacturing employment is now a small share of overall employment, these trade-induced losses are often described as “small and concentrated”.… But this assessment of the situation is wrong—growing trade (particularly with poorer nations) actually inflicts losses on the majority of the American workforce.… [T]he big damage is the permanent wage loss resulting from America’s new pattern of specialization that requires less labor and more capital. Further, this wage loss is not just suffered by workers in tradeable goods sectors who are displaced by imports; it’s suffered by all workers who resemble these workers in terms of credentials and labor market characteristics. A simple way to say this is that while landscapers may not be displaced by imports, their wages suffer from having to compete with apparel (and auto, and steel) workers who have been displaced by imports…. The wage-suppressing effects of globalization hit all workers without college degrees, across the country. Workers of all races and ethnicities are affected, and communities of color are disproportionately harmed. The harm of globalization is absolutely not a niche issue affecting only white working-class workers in the upper Midwest.

Globalization has contributed significantly to wage suppression. The first round of the academic debate about trade and wages was spurred by debate over approval in 1993 of the North American Free Trade Agreement, and accordingly the debate relied largely on data from the late 1980s or early 1990s. The majority view of economists coming out of this debate was that trade was not putting significant downward pressure on wages. But this view persisted even as the underlying reality changed significantly. For example, Bivens (2013) found that the implied wage effects of trade expanded rapidly after 1995, as trade with lower-wage nations (particularly Mexico and China) picked up significantly. He also found that, by 2013, trade flows with low-wage nations were likely reducing wages for workers without a four-year college degree by roughly 5.6%. For a non-college-degreed worker making the median hourly wage and working full time, full year, the earnings reduction translated into just under $2,000 annually.

This estimate is nearly identical to what Autor, Dorn, and Hanson (2013) found in a regression-based investigation of the wage impacts of imports from low-wage countries. Their results indicate that each $1,000 in imports per worker from low-wage countries lower American wages by 0.7%. Imports from all low-wage countries in 2016 stood at roughly $8,000 per worker, implying a wage reduction of roughly 5.6%, or about $2,000 annually, for a full-time worker earning the median wage.

Policy decisions have amplified globalization’s downward wage pressure. Much of the conventional wisdom among Washington policymakers has assumed that the trade agreements signed in the past 20-odd years have been exercises in good-faith liberalization of trade and have greatly expanded access by the world’s poor to U.S. markets. But this is not the right way to think about these agreements. Instead, these agreements have been the result of corporate capture creating selective and regressive protectionism that has severely restricted the policy space of our trading partners.

Even globalization that was driven solely by changes in communications and transportation technology and political change among our trading partners would have been likely to depress wage growth for the majority of American workers. But U.S. policy failures significantly amplified these damaging effects, turning globalization from a manageable challenge into a deep economic wound for workers—and a political disaster for the country. These policy failures include failing to secure any reasonable compensation or countervailing domestic boost to bargaining power and leverage in labor markets for those on the losing end of globalization; failing to address currency misalignments that have led to large trade deficits and hemorrhaging employment in manufacturing; and passing trade agreements that have consistently aimed to undercut workers’ economic leverage while carving out ample protections for corporate profits. This unbalanced feature of trade agreements has encouraged U.S.-based employers to substitute imports for the production formerly made by U.S. workers, and thus their jobs.

Weakened labor standards

Recent decades have seen the steady weakening of a number of key labor standards that once provided leverage and bargaining power for workers to improve job quality. The rapid erosion of the federal minimum wage’s purchasing power is the most dramatic and most consequential; others are the erosion of overtime protection for salaried workers, weaker labor-standards enforcement and rising wage theft, the increased share of the workforce with no effective labor protections because of its immigration status, and more extensive misclassification of workers as independent contractors.

Erosion of the federal minimum wage

The failure to update the value of the minimum wage in line with wage or productivity growth is a premier illustration of policy choices, made on behalf of capital owners and corporate managers, that have had a huge impact on wage growth for low-wage workers and is the primary explanation for any growth in the wage gap between low- and middle-wage workers over the last four decades. Specifically, the failure to raise the federal minimum wage to an adequate level (defined for our purposes as $15 an hour by 2025) has lowered the wages of at least the bottom 22.2% of earners and a full 31.0% of earners if one includes those benefitting from state and local minimum wage increases since 2017.22

In 2019, the federal minimum wage of $7.25 was worth 16.3% less, after adjusting for inflation, than when it was last raised in 2009—the longest period of history without a minimum wage increase—and 25% below its peak inflation-adjusted value in 1968. In contrast, productivity doubled (up 103.3%) and the average wage grew by 65% from 1968 to 2019.23 Thus, the value of the minimum wage fell while the labor market became far more capable of paying higher wages overall. This lag in the minimum wage occurred despite low-wage workers being older (and therefore likelier to have greater work experience) and significantly more educated than their counterparts in 1968 (Mishel 2014).

Another way of illustrating how low the minimum wage has fallen is by examining the impact of raising the federal minimum wage to $15 in 2025, which represents a 79.0% real increase24 over its current value and a 25.6% increase in purchasing power from the 1968 peak. This is a substantial, bold increase from the current policy, yet still falls short of updating the minimum wage to correspond to the growth of average wages or productivity (projected in 2025 to be 119% greater than in 1968), and would barely make up for its shortfall relative to median wages. Yet, increasing the minimum wage to $15 in 2025 (Cooper 2019) would have a tremendous impact.

The growth of the minimum wage shapes the entire wage distribution of the bottom half, essentially setting the scale of the gap between the lowest-wage workers at the 10th percentile and the wages at the median. In 1968, the federal minimum wage was equal to roughly half (52.8%) of the wage of the typical (median) U.S. full-time worker. In 2019, that share was less than a third (31.7%). An increase of the minimum wage to $15 in 2025 would reset it to 55.1% of the typical or median worker’s wage, restoring the wage standard achieved in 1968.25 This outcome implies that the declining value of the minimum wage more than explains the increase in the 50/10 wage gap over the 1968–2019 (or 1979–2019) time frame, and the rise to $15 in 2025 would totally reverse the increase in wage inequality in the bottom half, despite the modesty of the $15 proposal.

The work by Fortin, Lemieux, and Lloyd (2021), presented earlier in Table 1 (see the “changes in wage measure” column), provides further evidence. The 50/10 wage gap among women grew by 21.2 (log) percentage points from 1979 to 2017, and three-fourths (73.6%) of that increase can be explained by the erosion of the inflation-adjusted value of the minimum wage. Among men there was essentially no increase in the 50/10 wage gap, which is not surprising since men’s wages at the bottom have been less affected by a falling or rising minimum wage.

Using (unpublished) analyses provided by Thomas Lemieux based on the Fortin, Lemieux, and Lloyd (2021) model, we can pinpoint the impact of the decline in the minimum wage on the 10th percentile wage and on the 50/10 wage gap for men and women combined (Table 2). Over the 1979–2017 period the fall in the real value of the minimum wage meant that the 10th percentile wage fell 0.022 log percentage points instead of rising by 0.133 log percentage points, a 0.155 log point impact of the failure to maintain the value of the minimum wage. The impact of the declining minimum wage threshold was comparably as large among both men and women at the 10th percentile. Breaking out the 1979–1988 period shows that it was the dramatic fall in the minimum wage in the 1980s (a 0.179 log point decline in the 10th percentile wage) that explains the impact over the longer-term 1979–2017 trend.

Table 2

These results are echoed in the 50/10 wage gap. Among men and women combined the 50/10 gap grew 0.091 log percentage points from 1979 to 2017 but would have fallen by 0.047 log points had the minimum wage been maintained. The 50/10 wage gap’s increase of 0.214 log points among women would have been just 0.060 log points absent the fall in the minimum wage. Again, the results over the entire 1979–2017 period were driven by the developments in the 1979–1988 period.

The timing and pattern of changes in the wage gap in the bottom half (the 50/10 wage gap) seems to be entirely explained by changes in minimum wage policy—the choice to not maintain the real value let alone restore it to its 1968 peak. This can also be seen in Figure C above, where the growth of the 50/10 wage gap was limited to the 1979–1987 period. The 50/10 wage gap has been stable or falling since the late 1980s as the minimum wage was raised moderately26 (in 1990–1991, 1996–1997, and 2007–2009) and periods of persistent low unemployment boosted the lowest wages.

Other research confirms the major impact of the fall in the minimum wage on the expansion of the wage gap in the bottom half. For instance, the analysis of Autor, Manning, and Smith (2010) in Mishel et al. (2012, Table 4.41) shows that the fall in the minimum wage explains 65.5% of the 25 (log) percentage point increase in the 50/10 wage gap between 1979 and 2009. The wage gap at the bottom half grew very little among men (as they are far less likely to be affected by the minimum wage) over the 1979–2009 period, rising just 5.3 (log) percentage points.

That analysis, however, understates the degree to which policy changes have adversely affected the earnings in the bottom half if the standard applied is restoring the real value of the minimum wage to its 1979 value. Other reasonable counterfactuals would yield substantially higher estimated effects of a lower minimum wage. For instance, one could hold the ratio of the minimum to the median at its 1979 level or even benchmark to the 1968 peak rate. Further, using the standard of the ratio of the minimum wage to the median wage understates the impact of policy since the median itself was suppressed. That is, the minimum wage could be raised substantially more if median wage growth had not been suppressed. In this light, the fall of the median due to rising inequality of wages and the falling labor share of income—what we estimate to be a 43% decline—also adversely affected low-wage earners. If, for example, the minimum wage had kept up with economywide productivity growth after its high-water mark of 1968 (as it roughly had for the first 30 years of its existence), the federal minimum wage would be roughly $20 today.

The erosion of overtime protection among salaried workers

To be exempt from the minimum wage and overtime protections of the Fair Labor Standards Act under the “white collar” rule, a worker must be paid a salary (i.e., not be paid by the hour), must have bona fide “executive, administrative, or professional” duties (i.e., be an executive or a highly credentialed professional, or have supervisory duties), and earn above a specific salary threshold. Without a strong salary threshold, salaried workers who spend only a small share of their time actually doing exempt/”professional” work can be required to do hourly-worker-type duties (e.g., a store “manager” stocking shelves, unloading trucks, doing checkout at the cash register) for most of their worktime, including those beyond 40 hours in a week. Those hours beyond 40 are essentially unpaid.

The eroded share of the salaried workforce eligible for this overtime protection (i.e., receiving 150% of regular hourly wages when working more than 40 hours a week) is another example of a labor standard that was substantially weakened in the last four decades. The share of the salaried workforce automatically eligible for overtime based on its pay was whittled down from roughly half (49.6%) in 1975 to just 9.9% in 2014 (Kimball and Mishel 2016).

This issue gained prominence in the latter years of the Obama administration when the Department of Labor issued a rule that raised the annual salary threshold from $23,700 to $47,476 (an increase that still fell short of returning it to its 1975 level) (Department of Labor 2016). The benefit to workers of receiving overtime protection is a combination of higher hourly and annual wages and more leisure (as overtime hours are scaled back by employers when they are required to pay for them). The Department of Labor’s analysis (2016) of the 2016 rule showed that raising the salary threshold increased hourly wages by 1.1%, 2.8%, and 1.4% for salaried workers directly affected by the rule who, respectively, occasionally worked overtime, who regularly worked overtime and were newly covered by overtime protections, and who regularly worked overtime and remained exempt.27 However, because these groups comprised only about 40% of the 4.2 million workers directly affected by the higher threshold, the overall impact would be to raise hourly wages of directly affected workers less than 1.0%.

A broader group, however, is affected by the salary threshold. The Department of Labor estimated that another 8.9 million workers would have had their “overtime protection strengthened in Year 1 because their status as overtime-eligible will be clear based on the salary test alone without the need to examine their duties.” Eisenbrey and Mishel in two analyses (2015, 2016) presented evidence that much of this broader group had actually already lost overtime protections due to court cases and changes in rules introduced in the early 2000s and therefore would be directly affected by the new, higher salary threshold.

The changes in overtime eligibility affected middle-wage workers but not low-wage or higher-wage (90th percentile) workers. How much has the erosion since 1979 affected median wages? This depends on the impact on hourly wages of those affected and the share of middle-wage workers, say the middle fifth, affected by these overtime rules. The impact on the hourly wages of those affected by eroded protections would likely be about 1%.28 But not all mid-level earners have been affected by changes in overtime protections. In 2017 there were 160 million wage and salary earners, so the middle fifth comprised 32 million earners. As noted, the changes in the 2016 rule might have affected 13.5 million workers and, even if they all fell within the middle fifth, they would comprise only about 42% of middle-wage earners. In fact, the workers affected by the lowered overtime salary threshold fall between the 20th percentile and somewhat above the 60th percentile of the overall wage distribution.29 One should add in those who would have been covered if the 2016 rule had restored the 1979 salary threshold: That measure would increase the total affected by 22%.30 If one-third of middle-wage earners lost 1% of wages due to lost overtime protections, then the overall impact would be a 0.3% reduction of hourly wages for the middle fifth.

Wage theft and weaker enforcement of labor standards

Many workers, particularly low-wage workers and the women and men of color who are disproportionately in this category, frequently fail to receive the wages they are owed.31 This is referred to as “wage theft” and reflects workers being paid below the minimum wage, not being paid for all hours worked, not being paid time-and-a-half though legally eligible for overtime, experiencing illegal deductions from pay, and having their tips stolen by employers or supervisors. A recent example involved Amazon, which agreed to pay the Federal Trade Commission $61.7 million because it shorted its drivers tips over a two-and-a-half-year period starting in late 2016 (Greene 2021).

How extensive is wage theft? The available information shows the problem is pervasive. The best analysis is from a 2008 study of 4,387 workers in low-wage industries in Chicago, Los Angeles, and New York; it found that two-thirds of workers surveyed experienced at least one pay-related violation in any given week. The average violation amounted to 15% of earnings (Bernhardt, Milkman, and Theodore 2009).

Another study (Cooper and Kroeger 2017) examined the wage theft affecting minimum wage workers in the 10 most populous states (accounting for more than half the national workforce); it found that total wages stolen from workers due to minimum wage violations alone exceeds $15 billion annually on a national basis. The average minimum wage violation amounted to $64 per week or $3,300 per year for a year-round worker—nearly a quarter of the worker’s earned wages—and affected 17% of low-wage workers. The authors detail who was affected:

Young workers, women, people of color, and immigrant workers are more likely than other workers to report being paid less than the minimum wage, but this is primarily because they are also more likely than other workers to be in low-wage jobs. In general, low-wage workers experience minimum wage violations at high rates across demographic categories. In fact, the majority of workers with reported wages below the minimum wage are over 25 and are native-born U.S. citizens, nearly half are white, more than a quarter have children, and just over half work full time.

A back-of-the-envelope estimate, extrapolating from the Bernhardt, Milkman, and Theodore (2009) study, suggests that by 2016 aggregate wage theft was on the order of $52 billion annually (McNicholas, Mokhiber, and Chaikof 2017). This estimate does not include the stealing of tips where the wage received is nevertheless the minimum or violations of prevailing wage laws. This understated estimate of aggregate wage theft is, nonetheless, four times greater than the FBI’s $13 billion estimate of the total annual value of all robberies, burglaries, larceny, and motor vehicle theft.

Other insights can be gleamed by identifying the amounts recovered in wage theft suits and enforcement actions by the U.S. Department of Labor, by states, and through class action settlements. McNicholas, Mokhiber, and Chaikof (2017) did so and found $880.3 million in 2015 and $1.1 billion in 2016, for a total of $2 billion over both years. This estimate, according to the researchers, likely dramatically underrepresents the problem of wage theft, since probably only a fraction of victims actually file government complaints or are involved in class action settlements.

Shierholz (2021) details the reasons for the prevalence of wage theft:

One reason is workers’ diminished bargaining power relative to their employers. The fact that this unlawful employer behavior is not being “competed away” underscores that, for a variety of reasons, workers do not have the ability to quit these jobs as a de facto form of enforcement.

Another reason is that government resources devoted to combating violations of workplace protections are insufficient and have diminished. For instance, at the federal level, Shierholz writes: “In 1978, there were 69,000 workers per wage and hour investigator on average, but today that ratio is 175,000 to one.” The increased use of forced arbitration agreements, as detailed in a later section, has prevented workers from holding their employers accountable both in court and in the private arbitrations they are forced to use (frequently without the ability to do so as a class). As Shierholz (2021) notes:

To underscore the importance of class-action lawsuits in our enforcement framework: in 2015 and 2016, the top 10 private wage and hour class-action settlements alone exceeded the combined total wages recovered by all state and federal enforcement agencies. Forced arbitration with class and collective action waivers make it virtually impossible for low-wage workers to get any meaningful type of remedy.

Last, declining unionization enables more wage theft to go unchallenged. In effect, nearly every option workers have of holding employers accountable for stealing their wages has eroded in recent years: There is less protection through government enforcement, weaker access to legal recourse, and far less union advocacy in workplaces.

How much does wage theft affect wages in the middle and at the bottom? We do not have an estimate of aggregate wage theft across the wage spectrum or for middle-wage workers, so it is not possible to assess the impact of wage theft on the median wage. Among low-wage workers, Bernhardt, Milkman, and Theodore (2009) found that 68% experience wage theft violations averaging 14.95% of earnings. This translates into an average loss across all low-wage workers of 10.2%.32 Even this estimate omits certain types of wage theft, and we do not know how representative the study’s sample is of the national low-wage workforce. Nevertheless, the Bernhardt, Milkman, and Theodore study indicates that wage theft drains a substantial amount of low-wage workers’ earnings. We also do not know the extent of wage theft’s growth over the 1979–2017 period, though researchers judge that there has been substantial growth. A speculative estimate is that if wage theft has doubled to the 10.2% level implied by the Bernhardt, Milkman, and Theodore study, then it caused low-wage workers’ earnings to fall 5% over the 1979–2017 period due to weaker wage standards enforcement, less access to legal recourse, and eroded unionization. For mid-level wages, theft of overtime pay, unpaid worktime, and the undercutting of prevailing wages likely also had an adverse impact.

Misclassifying employees as contractors

Employers in an array of industries have increasingly (and illegally) misclassified employees as independent contractors or are paying workers “off-the-books.” This practice cheats workers of fringe benefits, social insurance protection (Social Security, unemployment insurance, workers’ compensation), labor protections (regarding safety/health and race, age, and gender discrimination), and union rights. The point of this misclassification is to lower labor costs, and it undercuts labor standards and “undermines other, more responsible employers who face costs disadvantages arising from compliance with labor standards and responsibilities” (Weil 2017).

It is difficult to quantify the extent of misclassification, since it is an illegal activity, and the extent to which it lowers wage and benefit costs. The fact that venture capitalists force this model on gig economy upstarts provides practical confirmation that the business strategy lowers labor costs and shifts risks to workers.33 Uber, a prominent example of a firm whose business strategy is built on misclassifying rideshare drivers, acknowledged in its registration for an initial public offering (Uber 2019) that misclassification provides substantial cost savings:

If, as a result of legislation or judicial decisions, we are required to classify Drivers as employees (or as workers or quasi-employees where those statuses exist), we would incur significant additional expenses for compensating Drivers, potentially including expenses associated with the application of wage and hour laws (including minimum wage, overtime, and meal and rest period requirements), employee benefits, social security contributions, taxes, and penalties.

David Weil, who has studied misclassification and has served as the administrator of the Wage and Hour Division of the Department of Labor (which oversees this issue), has noted:

Week after week, it seemed, I was witness to an investigation from our district offices involving the incorrect classification of all types of workers: janitors, home health aides, drywall workers, cable installers, cooks, port truck drivers, and loading dock workers in distribution centers. In one telling case, construction workers went home at the end of the week as employees only to be informed on the following Monday that, perhaps by the magic of some unknown force, they had become “member/owners” of hundreds of limited liability companies, effectively stripping them of federal and state job protections…. At the Department of Labor…we saw long-standing practices of employment undermined as misclassification spread quickly across sectors like restaurants, residential construction, and trucking and logistics. (Weil 2017)

Indications are that the practice has greatly increased. The last comprehensive federal estimate of independent contractor misclassification, a General Accounting Office (GAO) examination of tax year 1984, “found that 15% of employers nationwide and across industrial sectors engaged in misclassification of a total of 3.4 million workers” (Carré 2015). There is no national estimate for recent years, but Carré’s review (2015) of state audits of unemployment compensation data across 21 states completed as of 2012 showed that “the prevalence of misclassification ranges from 11% to 30% depending on the method used by state unemployment insurance agencies to select companies for audits…. [T]here is remarkable consistency in the prevalence of misclassification found across states using similar audit methods.” Weil’s (2017) interpretation is that “anywhere from 10% to 20% of employers misclassify at least one employee.”

Industry analyses provide information, too, on the numbers of workers affected. Ormiston, Belman, and Erlich (2020) estimate that in construction “between 1.30 and 2.16 million workers were misclassified or working in cash-only arrangements in an average month of 2017.” The major rideshare companies, whose business model incorporates misclassification, have between 1 million and 2 million drivers. Other online demand firms also rely on misclassification. A newspaper investigation by Locke and Ordonez (2014) analyzing payroll records for government-backed construction housing projects across 28 states found that “companies using stimulus money routinely snubbed labor law and the Internal Revenue Service by treating workers as independent contractors in a clear violation of what’s allowed.” These companies “listed workers as contractors instead of employees in order to beat competitors and cut costs….Scofflaws can save 20% or more in labor costs by treating employees as independent contractors.”

Misclassification is common in trucking. Following deregulation, large transport companies “sold their trucks to the drivers, then contracted with them on a per-load basis” (Bensman 2009, cited in Carré 2015). Appel and Zabin (2019) summarize the history:

The trucking firm practice of contracting with drivers for their services became a standard strategy across many parts of the commercial trucking industry by the mid-1990s. Contracting allows companies in many instances to shift responsibility for equipment to truck drivers, reduce payroll expenses such as employment taxes and employee fringe benefits, and retain the same effective control over the transporting of loads. Some trucking firms transformed their business model after deregulation entirely, becoming brokers by selling their trucks to former employee drivers and leasing those drivers’ services on an exclusive basis….For twice as much measurable output today, long-haul truckers now make 40% less in wages than they did in the late 1970s, when trucking was considered highly desirable blue-collar work.

A series of articles in USA Today in 2017 focusing on port truck drivers, who haul goods from ports to warehouses, found that

a good chunk of the port trucking industry relies heavily on a modern-day form of indentured servitude. The abuse starts when a trucking company pressures its drivers to sign lease-to-own contracts on their tractor-trailer rigs. Often, these drivers speak little English and do not understand what they are getting into. Once under contract and in debt, these short-haul drivers are at the mercy of the companies they’ve signed on with. The truckers can work days on end without making enough to cover expenses the companies charge them. If they complain, they are fired or given less profitable routes.

Precise estimates of the impact of rising misclassification are not possible with available data, but one can speculate about a range of possible impacts. To gauge the impact we assume that the 3.4 million misclassified workers found by GAO in 1984 (4.4% of nonagricultural wage and salary employment) have risen to 9.0 million (a 7.4% share), and that misclassification lowers wages by either 15% or 30%. Further, we will assume that misclassification is either spread throughout the private nonagricultural wage and salary workforce or, more likely, targeted at the bottom two-thirds; in the former case the share of misclassified workers in total employment rises by 3.0 percentage points, while in the latter case it rises by 4.5 percentage points. The impact is likely to have been on both low-wage and middle-wage workers. These parameters provide a range of impacts: Misclassification lowered wages by between 0.5% and 0.9% if applied across the whole workforce and between 0.7% and 1.4% if affecting and applied to only the bottom two-thirds.34 If one included all workers, including those in the public sector, then the estimated impacts would be proportionally less. We take a 1% decline in the median wage as our ballpark estimate.

Immigration policy that creates ‘labor standard free zones’

Employers have increasingly hijacked immigration policy to create zones in the labor market where workers’ ability to obtain enforceable basic labor standards is compromised by their immigration status. Note that the issue is not just the presence, or supply, of immigrants, but the legal situation that makes undocumented workers exploitable. In our economy, if you can be exploited, you will be. This exploitation of a sizable share of the workforce lowers the wages of migrants as well as those of the workers in their occupations and industries.

Costa (2019) has examined this issue and argues:

[I]n certain local labor markets and industries where a significant share of workers are migrants who do not have access to worker protections and basic labor rights—either because they are unauthorized immigrant workers or because they are migrants employed through nonimmigrant, temporary work visa programs (i.e., “guestworkers”)—the migrant workers’ lack of rights makes it difficult for them to bargain effectively for decent wages, and their weak leverage spills over to undercut the leverage of U.S. workers—i.e., citizens and immigrants who are lawful permanent residents.

Unauthorized immigrants comprise nearly 5% of the U.S. labor force, and these nearly 8 million workers (Krogstad, Passel, and Cohn 2019) are not fully protected by U.S. labor laws because of their precarious immigration status.35

As of 2013 approximately 1.4 million guestworkers were employed through temporary work visa programs, and they accounted for roughly 1% of the labor force (Costa and Rosenbaum 2017); the Organisation for Economic Co-operation and Development recently estimated that 1.6 million full-time equivalent jobs in the U.S. were filled by guestworkers in 2017 (OECD 2019a).36 Although they are legally authorized to work, guestworkers are among the most exploited laborers in the U.S. workforce because the employment relationship created by the visa programs leaves workers powerless to defend and uphold their rights.

Combining the estimates of unauthorized immigrants and guestworkers means that 6% of the workforce is vulnerable to exploitation due to its legal status. The presence of these exploitable migrants is greatest in various service industries—hospitality, restaurants—as well as in construction and agriculture. But there is also evidence that skilled, college-educated guestworkers are underpaid and have been subjected to financial bondage and even human trafficking (Costa and Hira 2020; Smith, Gollan, and Sambamurthy 2014; Stockman 2013). Costa and Rosenbaum (2017) estimate that there were roughly 800,000 skilled guestworkers employed in the United States in 2013.

Wage theft is a common form of exploitation of migrants, and estimates are that it occurs to an extreme degree. As Costa (2019) notes:

The exploitation described here is not theoretical. A landmark study and survey of 4,300 workers in three major cities (Bernhardt et al. 2009) found that 37.1% of unauthorized immigrant workers were victims of minimum wage violations, as compared with 15.6% of U.S.-born citizens. Further, an astounding 84.9% of unauthorized immigrants were not paid the overtime wages they worked for and were legally entitled to.

Research by Apgar (2015) comparing the wages of comparable migrant Mexican workers who were undocumented, had legal permanent resident status, or were temporary guestworkers in the H-2A and H-2B visa programs,37 found that unauthorized workers earned about 13% less than legal permanent residents; temporary foreign workers (i.e., guestworkers) earned about 11% less than legal permanent residents, and their wages did not significantly differ from unauthorized workers’ wages.

Apgar concluded: “Wage gaps suggest that employers use fear of deportation to pay lower wages—not just to unauthorized immigrant workers but to temporary foreign workers as well.”

The presence of exploitable migrant workers therefore undercuts labor standards in immigrant-intensive occupations and industries and thereby depresses wages and benefits of nonmigrants. This exploitation is likely to affect wages in the bottom two deciles more than those at the mid-level, although there are probably some impacts in the middle and further up the scale, where technology-sector guestworkers depress labor standards.

While there is not a deep economic literature that quantifies this type of spillover effect, much qualitative and historical research highlights that it is exactly the intent of many employers in expanding the pool of workers in the U.S. who lack basic worker rights because of their immigration status. Nevertheless, the conscious policy decisions to allow these circumstances clearly contribute to wage suppression.

Failures to police or check new forms of employment ‘contracts’

Employers are increasingly requiring employees to relinquish various rights when they accept employment or even after they are already employed. Noncompete and forced arbitration agreements are chief among these restrictions, and employers within various franchise chains also collude against employees through anti-poaching agreements. All of these agreements limit workers’ options by limiting access to courts and the ability to readily find another job or even to know the basic terms of their employment arrangement. This works to suppress wages.

Noncompete agreements

Employers have increasingly required employees to sign noncompete agreements, which limit options for future employment and are now widespread. The practice suppresses worker mobility and suppresses wages, and it depresses firm entry and dynamism because employees are prohibited from starting their own firms.

This section draws heavily on the Federal Trade Commission testimony of Evan Starr (2020) and Starr’s recent review (2019b) of the evidence on noncompete agreements. John Lettieri’s recent American Enterprise Institute study provides an additional useful analysis.

An example of a noncompete agreement is the one Amazon requires its employees to sign (Woodman 2015):

During employment and for 18 months after the Separation Date, Employee will not, directly or indirectly, whether on Employee’s own behalf or on behalf of any other entity (for example, as an employee, agent, partner, or consultant), engage in or support the development, manufacture, marketing, or sale of any product or service that competes or is intended to compete with any product or service sold, offered, or otherwise provided by Amazon (or intended to be sold, offered, or otherwise provided by Amazon in the future) that Employee worked on or supported, or about which Employee obtained or received Confidential Information.

This is a remarkable agreement for the breadth of its scope (Amazon sells a wide array of goods) and for the fact that it applies to low-wage warehouse workers, even seasonal ones, and that Amazon is one of America’s largest private-sector employers.

How widespread are noncompete agreements? According to a national survey of private-sector American business establishments with 50 or more employees in 2017 (Colvin and Shierholz 2019), “roughly half, 49.4%, of responding establishments indicated that at least some employees in their establishment were required to enter into a noncompete agreement. Nearly a third, 31.8%, of responding establishments indicated that all employees in their establishment were required to enter into a noncompete agreement, regardless of pay or job duties.” The researchers estimate that “somewhere between 27.8% and 46.5% of private-sector workers are subject to noncompetes.”

These data indicate that noncompetes have grown in their use since a survey (Starr, Prescott, and Bishara 2020) of employees in 2014 showed just 18% of the U.S. workforce covered by them, though 38% were subject to one at some point in their careers. The precise extent of the increased incidence of noncompetes is uncertain, however: The Colvin and Shierholz employer-based survey probably captures more noncompete use than the earlier employee-based survey, since many employees are unaware of having signed a noncompete agreement.

Noncompetes are intensively used in establishments with high pay or high levels of education, but they are also common in those with low pay and low education credentials. The 2014 employee survey found that 53% of those covered by noncompetes were paid by the hour and had median earnings around $14 an hour; 33% of the surveyed workers earning under $40,000 reported signing a noncompete agreement during their careers.

Job-to-job mobility is critical for earnings growth—employers don’t have to pay as much if they know workers don’t have good outside options—and so noncompetes, especially enforceable ones (not every state allows their enforcement) can be expected to lower wages. Some workers will be pushed to industries in which they are paid less. There is evidence that employers require employees to sign noncompete agreements even in states where they are not enforceable and that even in those states the agreements hamper wage growth and mobility.38

The contrary view is that noncompetes must be advantageous to employees or they would not have signed them. This “freedom of contract” paradigm presumes employees and employers have equal power and options in a competitive labor market. Lettieri (2020) rejects the freedom of contract interpretation:

The vast majority of noncompete agreements are not subject to any negotiation between the employer and employee, suggesting that the employee is unlikely to receive any benefits in return for their signature. A large share of these agreements are pre­sented for signature only after the employee has already accepted the job offer—often on the first day of work. Employers frequently exploit work­ers’ lack of knowledge and resources when crafting noncompetes. For example, employers commonly request that workers sign noncompetes even in states where they are completely unenforceable—and work­ers nevertheless sign the agreements assuming they are valid. Likewise, employers often craft extremely broad provisions knowing that employees generally lack both an understanding of what is enforceable and the wherewithal to challenge the terms in court.

In fact, Starr (2019b) reports that “less than 10% of workers reported negotiating over non-competes”; 83% of affected workers simply read and signed the contract. “When you ask workers, what did you—what were you promised in exchange for signing a non-compete, 86% of them say nothing, and roughly a third of non-competes were delayed until after the worker accepted the job without any change in responsibilities or a promotion.”

Starr (2019b), assessing studies examining the relationship between noncompete enforceability and wages, found:

[W]orkers in states that enforce non-competes earn less than equivalent workers in states that do not enforce non-competes. One recent study finds that the Hawaii ban on noncompetes for technology workers increased new-hire wages by 4%. The same study also documents that technology workers who start jobs in an average enforceability state have 5% lower wages even eight years later relative to equivalent workers in non-enforcing states. Another two studies looking at broader segments of the labor market document that the negative wage effects of non-compete enforceability are generally borne by those with less education.

Simply put, noncompetes lower wages and mobility for both technical and low-wage workers, whether they reside in states where the contracts are enforceable or in those, such as California, where they are not. Moreover, “where non-competes are really common and highly enforced, the whole labor market suffers” (Starr 2019b), as wages, job mobility, and job satisfaction decline even among those not directly affected.

In his testimony to the Federal Trade Commission, Starr (2020) noted that noncompetes are frequently accompanied by other restrictive agreements, such as nondisclosure, nonsolicitation of clients, nonsolicitation of co-workers, intellectual property assignment (giving the firm ownership of any intellectual property created on the job), and forced arbitration. That noncompetes are included in this cluster of agreements provides further evidence that their purpose is to restrict employee options rather than to protect beneficial investments in and information held by employees. Colvin and Shierholz (2019) also find a strong correlation between the presence of noncompete and forced arbitration agreements.

What is the impact of increased use of noncompete agreements on median wages, low wages, and various wage gaps? The best evidence regarding noncompetes and wage levels is the Lipsitz and Starr (2020) examination of the relationship between Oregon’s 2008 ban on noncompetes for low-wage workers and the average hourly wages of hourly paid workers. The finding that the ban raised wages for hourly workers by 2.2% to 3.1% reflects the impact on those directly affected (about 14% of hourly workers are subject to noncompetes) and the spillover effects on other comparable workers. It is important to note that these results are for hourly, not all, workers, and hourly workers comprised 67% of Oregon earners. Two indications in the Lipsitz and Starr research provide clues to the impact on the median worker. One is that the ban’s impact was comparable across the 20th to 80th wage percentiles of hourly workers, suggesting that the impact on the median would be comparable to the 2–3% average effect (if hourly workers comprise the bottom 67% of earners, then the overall median is the 75th percentile of hourly workers). On the other hand, Lipsitz and Starr report that the impact of the noncompete ban was higher for two occupation groups with wage levels close to the overall median: In construction occupations and installation, maintenance, and repair occupations wages rose by 4.9% and 4.3%, respectively.39 The ban’s impact on wages in construction occupations occurred even though the incidence of noncompetes in construction was below average.40 The ban had basically no effect on a low-wage occupation, food service preparation. These differences across occupations reflect both the incidence and direct impact of noncompetes. In sum, these results suggest that the impact of noncompetes on the median is in the 4.3% to 4.9% range, there is little if any impact for the lowest-paid workers, and noncompetes actually narrow the wage gap (50/10) in the bottom half by depressing the median wage but not affecting the lowest-wage workers.

Another study, Johnson, Lavetti, and Lipsitz (2020), analyzed how the legal enforceability of noncompetes affects wages and labor market mobility. Using a newly constructed state–year panel of noncompete enforceability spanning 1991 to 2014, the authors found that enforceability of noncompetes leads to a decline in workers’ earnings and mobility and increases racial and gender wage gaps (“the earnings effect among women and black workers is twice as large as the effect among white men”). They estimate that average earnings would increase by 8.5% nationally if noncompetes were made unenforceable. We do not draw on these results for our assessment of noncompetes on the median wage, but the 4.3% to 4.9% range we do rely on is far below that of these results. Johnson, Lavetti, and Lipsitz also show that noncompetes lower the wages of college-educated workers more than non-college-educated workers, suggesting that noncompetes narrow the 90/50 wage gap, especially since there is a greater incidence of noncompetes among college-educated workers (25% compared to the average incidence of 18%, according to Starr, Prescott, and Bishara 2020, Figure 3). Elimination of noncompetes would also drain the business profits that accrue to upper-income households, so the overall impact of eliminating them or scaling back their use and enforceability on overall income inequality is uncertain.

Assessing the impact of the increased use of noncompetes on median wage growth since 1979 requires quantifying that increased use. Unfortunately, there is no historical series on noncompete incidence. The agreements have been used for many years, especially among higher-wage professionals and executives, and use has increased as evidenced by the increased public and policymaker attention to the agreements, particularly for middle-wage or lower-wage (e.g., Jimmy John’s sandwich shop workers) workers. If, say, the incidence among hourly workers has doubled since 1979 and the wage impact is roughly 4.5% in recent years, then noncompetes have lowered the median wage by about 2.25%. The impact might be lower if we base the assessment on the 2.2% to 3.1% overall wage effect identified by Lipsitz and Starr (2020) or if the rise in incidence is less than double. It seems equally plausible to us, however, that “doubling” is an underestimate, since we know the incidence of forced arbitration agreements has enormously increased since the early 1990s (from 2% in 1992 to more than 50% in 2017), and firms insisting on forced arbitration also tend to insist on noncompetition agreements. So, we take the 2.25% impact on median wage growth as a rough estimate.

Forced arbitration and class action waivers

Employers increasingly requiring their workers to sign arbitration agreements is another clear example of policy decisions, this time from court decisions starting in 1991, limiting workers’ options to resist workplace exploitation and contributing to wage suppression and inequalities. As noted above, forced arbitration is among a suite of agreements being forced on workers as a condition of employment (Colvin and Shierholz 2019). This trend accompanies the weaker enforcement of labor standards and workplace civil rights by public authorities, as discussed in the wage theft section. Forced arbitration suppresses claims, makes them less likely to succeed, and reduces awards. The consequence is the undermining of the enforcement of employment rights ranging from minimum wage and overtime pay to rights to equal pay and freedom from discrimination or harassment based on race, gender, or religion (see Stone and Colvin 2015, Colvin 2018, and Deutsch et al. 2019).

The incidence of forced arbitration agreements took off after key Supreme Court decisions in 1991 and 2001 made clear “that an American employer may, with near total impunity, require an employee, as a condition of hiring and continued employment, to use private arbitration as the means of resolving public claims against the employer that involve a statutorily protected right” (Lipsky 2007). In 1992, just over 2% of the workforce was covered by forced arbitration agreements, but that share rose to almost a quarter by the early 2000s. By 2017 the share was 56.2% (Colvin 2018).

Of the employers who mandate arbitration, some 30.1% also mandate class action waivers, meaning that besides losing the right to file a lawsuit on their own behalf employees also cannot pursue even their private arbitrations through collective legal action. Colvin (2018) found that “23.1 percent of private-sector nonunion employees, or 24.7 million American workers, no longer have the right to bring a class action claim if their employment rights have been violated.” Individual lawsuits are often unrealistic for low-wage and even middle-income workers because the cost of legal representation may exceed their lost wages. In contrast, class and collective actions allow workers to aggregate claims, making litigation cost-effective (Deutsch et al. 2019). The Supreme Court’s 2018 Epic Systems v. Lewis decision held that employers can require employees to give up their right to sue either as an individual or on a collective basis, a development expected to lead to a further surge in the incidence of mandatory individual and class forced arbitration agreements.

Mandatory arbitration agreements are required across the wage spectrum but have the highest incidence among low-wage workers: Workplaces paying less than $13 per hour mandated arbitration 64.5% of the time, while those with higher wages did so 48–54% of the time (Colvin 2018, Table 4). The largest employers mandate arbitration most often: Those with at least 5,000 employees mandate it 67.7% of the time, while employers with fewer than 500 employees do so 49% of the time (Colvin 2018, Table 1).

There’s a catch-22 to the mandatory arbitration scheme, as illustrated by this example (Deutsch et al. 2019):

Low-wage employers like Chipotle have demonstrated that they never intended arbitration to be a viable alternative to class-action litigation. After being sued for systemic wage theft, Chipotle forced workers to sign arbitration clauses to prevent them from joining the suit. Then when individual workers filed for arbitration, Chipotle blocked the arbitrations from proceeding by refusing to pay its share of the fees.

It is not possible to assess the wage impact of the spread of forced arbitration agreements. The practice is intended to and does undermine the enforcement of employment and civil rights workplace protections, further limiting employee options to resist employer exploitation.

Has any harm done by forced arbitration agreements been offset by the gains made by workers in agreeing to such terms? The majority opinion in Epic Systems asked, “Should employees and employers be allowed to agree that any disputes between them will be resolved through one-on-one arbitration?” and assumed that when employers and employees have a “freedom” to contract they will both obtain optimal outcomes. Justice Ginsburg in her dissent noted the incongruity of such a claim given that these forced arbitrations were imposed on already-employed workers given “a Hobson’s choice: accept arbitration on their employer’s terms or give up their jobs.” Justice Ginsburg asked:

Were the “agreements” genuinely bilateral? Petitioner Epic Systems Corporation e-mailed its employees an arbitration agreement requiring resolution of wage and hours claims by individual arbitration. The agreement provided that if the employees “continue[d] to work at Epic,” they would “be deemed to have accepted th[e] Agreement.” Ernst & Young similarly e-mailed its employees an arbitration agreement, which stated that the employees’ continued employment would indicate their assent to the agreement’s terms.

Employer collusion and anti-poaching agreements

Do employers collude to suppress pay and restrict competition in the labor market? Adam Smith is often cited as recognizing extensive collusion, writing that employers “are always and everywhere in a sort of tacit, but constant and uniform combination, not to raise the wages of labour above their actual rate.” This behavior cannot be readily studied since, as Smith also said, “We seldom, indeed, hear of this combination, because it is the usual, and one may say, the natural state of things, which nobody ever hears of” (quoted in Krueger and Ashenfelter 2018). A remarkable case of collusion did come to light in recent years when a class action civil suit on behalf of more than 64,000 software engineers and other employees of Apple, Google, Adobe, Intel, Intuit, Pixar, and Lucasfilm was settled in 2015 for half a billion dollars.

We do not know the extent of collusion among employers and how it has changed over time. It is, after all, illegal. There is research on explicit collusion in franchising, however, because this is a gray area in the law.41 Krueger and Ashenfelter (2018) report on anti-poaching agreements in franchises, agreements that limit a particular franchise of McDonald’s, for instance, from hiring a worker from another McDonald’s franchise. To be clear, this type of agreement differs from a noncompete, where employers limit a worker’s options. An anti-poaching agreement is a direct collusion among employers not to recruit workers from each other.

In discussing their findings, Krueger (2018) noted:

58 percent of franchise companies have a no-poaching clause that prevents or restricts the ability of one franchisee in a chain from hiring workers employed by other franchisees. This is up from 36 percent in 1996. The practice is particularly common in fast food chains. We find that 80 percent of the 40 largest Quick Service Restaurant franchise chains have a no-poaching requirement. Since the human capital would remain within the chain, there is little business justification for such a clause other than to restrict worker mobility and opportunities.

No-poaching agreements are more common in low-wage and high-turnover industries (Krueger and Ashenfelter 2018).

Krueger and Ashenfelter (2018) provide a quantitative example to illustrate the impact of no-poaching agreements to facilitate employer market power over workers within franchise chains. In Rhode Island, for instance, instead of 261 employers competing for Quick Service Restaurant workers, anti-poaching agreements effectively limit the competition to just six.

Of course, employers do not have to pay as much when workers have fewer options. Unfortunately, there is no systematic evidence of no-poaching agreements’ impact on workers’ pay and within-franchise job mobility. We do know that these agreements grew substantially over the 1996–2016 period, however, and disproportionately affect workers in low-wage industries and “potentially affect a large number of workers” (Krueger and Ashenfelter 2018).

Tolerating new business structures that disempower workers

In recent decades, employers have increasingly tried to build up concentrated power in product markets (as well as labor markets directly) and to leverage this increased product market power to augment their profitability and the pay of executives by lowering costs and suppressing wages. One mechanism has been to match market concentration with efforts to outsource key parts of their production or workforce to keep those costs from making a claim on the firm’s income. In past years, policymakers might have used industry regulation such as in airlines and trucking and antitrust enforcement to keep these changes in check. But in recent decades, the pushback against these changes in business structure has been rare and muted.

Fissuring: contracting out/outsourcing, temping, and franchising

Perhaps the most pronounced way that employers have attempted to shape labor market outcomes to their advantage through changes in business structure is the “fissuring” of workplaces. David Weil’s book, The Fissured Workplace (2014), as well as analyses by Appelbaum and Batt (2014), provide the details about what fissuring is and how it works to the advantage of employers.

In fissuring, Weil explains, companies seek to focus on the tasks that provide the greatest return while shedding those that are harder to reap extra-normal returns from. The “lead business” maintains tight control of the outcomes of the tasks it sheds by careful monitoring of the spun-off or contracted supplier firms. A classic example is universities: Many once hired employees directly to run student services like cafeterias, but they now hire outside firms instead. In retailing, detailed mechanisms of monitoring and control exist between the lead firm and contracted firms, including franchising agreements, supply chain monitoring mechanisms, and increasingly sophisticated software algorithms. Essentially, lead firms insist on a high degree of control over much of the contracted firms’ output, but abdicate any responsibility for the contracted firms’ employment conditions.

Weil points out that fissuring should be distinguished from contingent work or alternative work arrangements. “Fissured workplace arrangements can exist even though employment itself might be traditional (that is, ongoing and full time) when the worker is employed by a subcontractor, franchisee, or other business organization undertaking the work of a lead business.” In fact, as discussed below, traditional employment arrangements dominate the fissured workplace, and the increase in fissuring, as best as we can surmise, has primarily occurred through subcontracting to firms still using traditional employment arrangements. In short, most of the “fissured economy” does not reflect independent contractors, the gig economy, or other forms of contingent work.

Fissuring is a corporate strategy that emerged from the focus on shareholder value. It raises profits in part by squeezing the costs of subcontractors, who in turn cut wages, and shifting risks onto other firms and workers. Further, fissuring allows workers who may have once benefited from firm-specific economic rents to be cast out of the firm and forced to do the same work at a contracted firm that does not have access to these rents. In a real sense, part of the rationale for fissuring is to hire contracting firms to do the dirty work of cutting pay for workers who were once in the lead firm but who happened to not be performing tasks where rents could be extracted. Fissuring also redistributes profits from contractors to the lead firms. It does not reflect a new way of producing goods and services but rather a rearranging of the business hierarchy in which particular workplaces fit (Appelbaum and Batt 2014). As such, fissuring probably has no impact on aggregate productivity, or on making the production of goods and services more efficient. Instead, its effects are overwhelmingly distributional. Fissuring also does not reflect technological change in how goods and services are produced. Technology does, though, play a role in that communications and other connectivity better enable firms to monitor dispersed economic activity and the fulfillment of standards.42

Appelbaum and Batt (2014) argue that political and institutional factors have also enabled fissuring:

[F]inancial market deregulation gave investors and stockholders more power to pressure firms to maximize shareholder value, and the lax enforcement of labor laws and the decline of union power freed them from prior constraints to do so.

So, fissuring and domestic outsourcing should be interpreted is a business strategy choice and not a beneficial “market force” that satisfies consumer preferences or adapts to newfound abilities to improve efficiency.

Weil’s (2019) estimates show the fissured economy to be substantial:

[C]lose to 19 percent of the private-sector workforce were in industries where fissured arrangements predominate…prevalence could easily double, the presence of fissuring in one workplace spills over to the wage-setting decisions of other businesses and to the labor markets in which they compete for workers.

But estimating the size of the fissured economy is a major challenge researchers are only now undertaking. However, it seems clear that somewhere between a fifth and a third of the economy is characterized by fissuring as a dominant force. Moreover, where fissuring is present it can be extensive, as witnessed by the fact that “about half of Google’s workers are contractors who don’t receive the same benefits as direct employees” (Bergen and Eidelson 2018).

For fissuring to contribute to wage suppression in recent decades it would have to have grown in size and/or provided a more intense push downward on wages and job quality in this time. In terms of growth there is much we don’t know and some things we do. Bernhardt et al. (2016, 40) concluded that “available information points to rapid growth in domestic outsourcing in a wide range of industries since the 1980s.” Fissuring’s growth seems to have primarily occurred in its least-well-measured component, business-to-business subcontracting (frequently relying on traditional employment arrangements), and not through any substantial growth in other components that are better measured, such as use of independent contractors,43 staffing agencies,44 or franchising.45

In any case, there are likely to be downward wage pressures from fissuring. As Weil (2019) notes: “The fissured workplace has been accompanied by the erosion of wage and other workplace norms in many industries and occupations” (p. 160). This is because “[b]y shedding their employees in a variety of ways and making those workers the employees of other organizations, a wage-setting problem becomes a pricing problem. The janitor, maintenance person—or even lawyer—who no longer is a member of the company also no longer need be bounded by the pay considerations of that company’s wage structure” (p. 154).

Various studies confirm that workers in contractor firms earn less, including security guards and janitors (Berlinski 2008; Dube and Kaplan 2010); call center workers (Batt, Doellgast, and Kwon 2006; Batt, Holman, and Holtgrewe 2009; Batt and Nohara 2009); airline workers (Callaci 2019); and logistics (i.e., truck drivers, warehouse workers), cleaning, security, and food services workers (Goldschmidt and Schmieder 2017). Reviews of this literature all indicate the adverse impact of outsourcing on wages (Weil 2019; Appelbaum 2017; Bernhardt et al. 2016; and Rosenfeld 2021).

Moreover, outsourcing not only lowers wages; it also leads to greater wage theft and more dangerous working conditions (Weil 2014, 2019).

Fissuring, and particularly the outsourcing of particular tasks, is probably responsible for the fact that workers in the largest firms no longer receive higher pay than those in medium-sized firms. Where large firms once likely had a core competency that generated economic rents for workers but also directly employed many in occupations outside this core competency, fissuring has allowed lead firms to outsource these noncore occupations and restrict employment to only those jobs performing the tasks that generate or share rents. At the same time, large subcontracted firms dominated by the cast-offs of lead firms now provide relatively low-wage work concentrated in occupations where rents are harder to generate and protect.

As Bloom et al. (2018) show, those in firms with more than 2,500 employees were not paid more in the 2007–2013 period than those in firms that had 1,000 to 2,500 employees, a sharp drop from what prevailed in the 1980s. For instance, a firm with 10,000–15,000 employees paid 15 (log) percentage point more than a firm with 1,000–2,500 employees in the 1980s, but that differential was absent in the 2007–2013 period. The impact is felt beyond wages, since large firms also provide better-quality jobs overall, including higher benefits and better working conditions (Maestas et al. 2017). This erosion of the quality of jobs in large firms affected a large swath of the workforce, as employment in firms exceeding 2,500 employees comprised 39.0% of all jobs in 2014 compared to 37.0% in 1999 and 35.3% in 1979.46

Weil (2019) drives home the point in showing the wage disparities in delivery services: a traditional union employee at UPS earns $23.10 an hour, an “independent contractor” for FedEx earns $14.40, and an Amazon contracted worker (driving the “last mile” of a delivery scheme) earns $5.30, less than the minimum wage.47 In all three case the basic tasks are the same but the corporate structure differs.

A speculative gauge of the impact of fissuring might find that a shift of 15 percentage points of employment into fissured workplaces earning 15.0% less (Goldschmidt and Schmieder 2017) would yield an overall decline of wages of 2.25% overall but even more among non-college-educated and more vulnerable segments of the workforce.

Product and labor market concentration

There has been increasing interest in two key changes in corporate structure in recent decades: product and labor market concentration (sometimes referred to as monopoly and monopsony). We will not use the label “monopsony” here for labor market concentration, as modern labor economics has adopted the term “monopsony” to describe a wide range of influences—including but not limited to market concentration—that give employers power to set wages. It seems clear that there are many reasons for policymakers to be concerned about market concentration, and robust efforts (antitrust or regulation) to confront the malign effect of concentration should be part of the policy toolkit going forward.

Labor market concentration. Though labor market concentration is definitely associated with lower wages, evidence remains lacking that it has increased so as to greatly contribute to wage suppression. Key papers assessing the effect of labor market concentration include Azar, Marinescu, and Steinbaum (2017); Benmelech, Bergman, and Kim (2018); Rinz (2018); and Naidu, Posner, and Weyl (2018). Bivens, Mishel, and Schmitt (2018) surveyed the evidence presented in these papers and found it to show persuasively that labor market concentration is negatively correlated with average wages. The evidence also shows that the average labor market is highly concentrated but the average worker is not employed in a concentrated labor market. This is because concentration is prevalent in rural areas and other labor markets with lower population density, but most people work in labor markets with low to moderate degrees of concentration. The disproportionate labor market concentration faced by rural workers is a key source of the economic stress felt by such workers.

Evidence on trends in aggregate labor market concentration is harder to obtain. Labor market concentration within manufacturing shows a modest increase between 1979 and 2009. This time-series analysis also shows that the adverse wage effects of concentration within the manufacturing sector are greatly attenuated by unionization. If overall labor market concentration rose as fast as concentration within manufacturing over the 1979–2014 period, then rising concentration would have lowered average wages by roughly 1% over the period.

Rinz (2018) found that local labor markets (which he defines as industry by geography) have become substantially less concentrated over time, while national labor markets have become more concentrated. He explains this finding by noting that large players in local labor markets may have become bigger national players over time. For example, if 30 years ago regional fast-food chains dominated local labor markets but now just one or two national chains dominate all local labor markets, local labor market concentration would be unchanged (there was always only a small number of players in any local labor market) but national concentration would be greater. The effect of this pattern of concentration on earnings and wage inequality depends on what is more important—local or national labor markets.

Product market concentration—monopoly power in product markets. The growth of product market monopoly power over prices does not seem to have contributed to wage suppression, though firms may have leveraged their monopoly positions to suppress wages and profits in supplier chains. Key papers directly assessing the effect of product market concentration include Autor et al. (2017b); Barkai (2020); De Loecker, Eeckhout, and Unger (2020); Grullon, Larkin, and Michaely (2015); and an analysis by Goldman Sachs (Struyven 2018). Bivens, Mishel, and Schmitt (2018) surveyed some of the evidence presented in these papers and found that they demonstrated that product market concentration rose steadily across six sectors from 1982 to 2012 (manufacturing, retail, wholesale, services, finance, and utilities and transportation). The magnitude of this rise has varied substantially across sectors. Data after 2012 are not yet available from the Census Bureau. Other sources of data on industry concentration (for example, from Compustat, a private database of publicly owned firms) are often viewed with some suspicion by scholars of industrial organization.

The main channel through which increased product market concentration would affect wages in most textbook conceptions is through a shift in income from wages to profits. The increased concentration gives firms pricing power, and the higher prices increase profit margins while reducing the purchasing power of wages. Between 1997 and 2012, during the time when concentration in a number of industries increased, the overall share of corporate-sector income claimed by labor shrank significantly. However, labor markets were very different in 1997 as compared to 2012; the unemployment rate was 4.9% in 1997 and 8.4% in 2012. As labor markets tightened after 2012, labor’s share rose in the corporate sector.

To more rigorously test the hypothesis that there has been a secular decline in product market competition that has transferred income from wages to profits through higher prices, one would want to see the labor share in equivalently tight labor markets over time. Bivens (2018) notes that economists have often been too quick to declare declines in the labor share “structural” and hence not amenable to reversal by more expansionary macroeconomic policy. That is, the fall in labor’s share could be the result of firm monopoly power but could also be due to eroded worker power due to excessive unemployment. Many of the studies of monopoly’s impact on labor’s share cannot distinguish between these two competing, or complementary, explanations.48

The rise in labor’s share as unemployment fell in the recent recovery suggests a strong role for eroded worker power, since monopoly power is not expected to diminish with lower unemployment. For instance, labor’s share (in the corporate sector) rose from 75.6% in 2016, when unemployment was 4.9%, to 77.3% in 2019, when unemployment averaged 3.7%.49 Stansbury and Summers (2020) note in support of the eroded worker power explanation that “Greenwald, Lettau, and Ludvigson (2019) find that a reallocation of income from labor to shareholders can account for a large share of the rise in equity valuations from 1989 to the present.”

Barkai (2020) finds that increases in product market concentration in recent decades have reduced both labor and capital shares across detailed industries, with a concomitant rise in the “pure profit” share, which he interprets as evidence of declining competition and greater monopoly power. His framework of parsing out profit and capital shares separately is extremely useful. However, the capital share requires estimating a return to physical capital, and because this return and the pure profit return are bundled together in any observable data series, it is a real empirical challenge to do and no professional consensus exists on the scale of his estimates.

Autor et al. (2017b) estimate that the increase in product market concentration between 1997 and 2012 could account for a third of the decline in the labor share over that period. Bivens, Mishel, and Schmitt (2018) calculate that this translates into a reduction in overall wages of roughly 2.9% over that time. It is also possible that what Autor et al. are measuring reflects fissuring rather than product market concentration.

While the rise of profit shares is consistent with evidence of rising monopoly power in product markets, Stansbury and Summers (2020) note that this is also consistent with reduced labor market power of workers. They further note that textbook models of increased monopoly power in product markets should also see faster inflation, reduced output, and an increase in the unemployment rate consistent with stable inflation, i.e., the nonaccelerating inflation rate of unemployment, or NAIRU. Decreased worker power, conversely, should see slower inflation and a reduced NAIRU—and these seem more consistent with real-world macroeconomic evidence over recent decades. In short, rising monopoly power in product markets that has harmed U.S. households through excessive price growth seems unlikely to be a major channel through which concentration may be dragging on wage growth.

Product market concentration—dominant buyers squeezing suppliers. In traditional conceptions of the harms done by product market concentration, firms’ monopoly power is leveraged against consumers of their output, with prices being pushed above what would prevail in competitive markets. However, many real-world firms with substantial market share (Walmart and Amazon, for example) charge their own customers seemingly low prices while leveraging their market power instead against the firm’s own suppliers, coercing them into providing supplies at low prices. This in turn squeezes both profits and wages for the supplier firms.

Path-breaking research by Wilmers (2018) has identified and quantified the impact of these “dominant buyers.” He documents that dominant buyers have been able to squeeze profits and lower wages in their supplier chains, and their economic heft has grown over time. Using Securities and Exchange Commission filings of publicly owned companies to document the large increase in dominant buyers, Wilmers estimates that the share of nonfinance suppliers’ revenue obtained from dominant buyers increased from 5% in 1979 to 19% in 2014 overall and from 6% to 26% in manufacturing and logistics.50 The biggest increase in the role of dominant buyers came in manufacturing, wholesale and shipping, primary resource extraction, utilities, construction, and information and professional services. Over the course of the last four decades “the composition of these dominant buyers shifted toward large retailers [e.g., Walmart] and other intermediaries“ (Wilmers 2018, 221).

Wilmers argues that there was not only an increase in the role of dominant buyers but also an intensification of their wage impact:

[N]egative wage effects have intensified since the early 2000s, and in several years in the 1980s, reliance on dominant buyers had negligible effects. During the period of wage stagnation and the restructuring of buyer-supplier relations, the wage effects of dominant buyer reliance turned increasingly negative. (Wilmers 2018, 229)

Wilmers estimates that the increase in dominant buyers lowered the growth of average annual earnings by 3.4 percentage points over the 1979 to 2014 period among publicly owned nonfinancial firms. The impact among low- and moderate-wage workers almost surely was larger than this.51

Deregulation of industries

Starting in the late 1970s, Congress deregulated various industries, including airlines, trucking, interstate busing, telecommunications, utilities, and railroads.52 In each of these industries, deregulation had a strong adverse impact on the wages and compensation of blue-collar workers. Fortin and Lemieux (1997) showed that 9% of the workforce in the 1980s was affected by industry deregulation and that in such industries there was a much larger erosion of middle-wage jobs.53 According to their estimates, deregulation explained about 7% of the rise in male wage inequality between 1979 and 1988, especially for those above a low-wage threshold. Card (1996) showed a 10% decline over 1980–1990 in the relative earnings of airline workers after deregulation. Deregulation weakened the ability of employers to pay high wages and in many sectors, most notably trucking, led to a steep erosion of unionization (Viscelli 2016). A New York Times (2018) editorial noted, for example, that “the 1.7 million heavy and tractor-trailer truck drivers in the United States earned an average of $44,500 last year…and far below the $55,500 (in 2017 dollars) earned in 1979, despite drivers working longer hours.”

The aggregate impact of the policy choices generating wage suppression

This section draws on the earlier assessments of the factors generating wage suppression to account for the divergence between the growth of net productivity and median hourly compensation over the 1979–2017 period and the suppression of the growth of the 10th percentile wage and the wage gap in the bottom half (i.e., the 50/10 wage gap). We start by looking at the relationship between our analysis and that of other contributions to understanding wage inequality and, specifically, how our analysis complements and points in the same direction as other recent research that has focused attention on worker power.

Relation to other literature

Our analysis builds on what Stansbury and Summers (2020) referred to as a “long history of progressive institutionalist work exemplified by Freeman and Medoff (1984), Levy and Temin (2007), and Bivens, Mishel, and Schmitt (2018).”

An important recent marker in this tradition was the keynote address by former chairman of the Council of Economic Advisers and Princeton economist Alan Krueger (Krueger 2017) to the Federal Reserve Board Jackson Hole conference. Krueger said that certain economic models “give employers some discretion over wage setting” and, in a footnote, said, “Notice that I don’t call these features ‘imperfections.’ They are the way the labor market works. The assumption of perfect competition is the deviation from the norm of ‘imperfection’ as far as the labor market is concerned.”

Krueger’s speech reflects a shift away from the conventional labor market view that has been accepted by both liberal and conservative economists for decades: that competitive labor markets are the same as any other market and are characterized by equally empowered employers and employees. What is emerging, however, is an appreciation that labor is a distinct type of commodity and that unequal bargaining power is an inherent characteristic of the employer-employee relationship. This change of perspective has a profound impact on the analysis of labor market dynamics and which labor policies and economic policies in general are required to rebalance power in the labor market and thus obtain robust wage growth and better-quality jobs for the vast majority.

There has also been a growing acceptance that anemic U.S. wage growth can be explained by the “rigging of the system” against the typical worker. This framework was adopted by former President Trump, who picked up on themes earlier laid out by Senators Elizabeth Warren and Bernie Sanders.

That the economic system is rigged to create upward redistribution has been the theme of economists (e.g., Dean Baker, Josh Bivens, Jared Bernstein, Joseph Stiglitz, Lawrence Mishel, Heidi Shierholz, Elise Gould, and James K. Galbraith) and policymakers for some time, though articulations of how the system is rigged vary. President Trump, for instance, focused on immigration, trade treaties, and bureaucracy. Others focus on growing product market monopoly power affecting consumers. Some analyses focus on the need for more competition (Block and Harris 2021). The approach in this paper is to pinpoint and measure the impact of the factors that have empowered employers in the labor market to be able to suppress wage growth for the vast majority; these include maintaining excessive unemployment, weakening labor standards, diminishing unions and their economic and political power, leveraging monopoly power to lower labor costs through fissuring and subcontracting, and by adopting forced arbitration, noncompetes, and other contractual mechanisms to undercut worker options.

This paper returns to a mode of analysis, dominant in the 1990s but abandoned in recent literature, that focuses on empirically identifying the impact of specific causes of wage inequality and evaluating their contribution to explaining overall wage inequality trends.

Comparison to the new monopsony literature

The new monopsony literature reinforces our narrative in important ways and further highlights the need to identify the specific factors generating employer power over wages and ways this power has changed over time. A broad interpretation of employers’ “monopsony” power does not hinge on labor market concentration (i.e., the proverbial one-company town), but instead diagnoses labor markets as being affected by employers’ exercise of power that allows them to cut wages without fear of losing a large portion of their workforces—regardless whether the source of this power is market concentration or anything else.

A number of studies—including Webber (2015 and 2020); Dube, Giuliano, and Leonard (2019); Dube et al. (2020); Bassier, Dube, and Naidu (2020); Azar, Marinescu, and Steinbaum (2019); Langella and Manning (2020); Card et al. (2018); and the meta-analysis by Sokolova and Sorensen (2020)—show that employer power is ubiquitous in the modern U.S. labor market. These studies estimate the labor supply elasticities facing firms, i.e., the responsiveness of employees to exit following changes in wages offered by employers. For instance, Webber (2020, 18) notes:

The majority of firms (median labor supply elasticity=0.85) compete for workers in labor markets where the typical employee is highly unlikely to move in response to small or even modest changes in their wage. This gives these firms considerable latitude to pay lower wages without worrying about a mass exodus of employees.

The monopsony literature has identified a substantial amount of employer power such that employers are able to, as Bassier, Dube, and Naidu (2020) put it, “mark down” wages by anywhere from 20% to 50%. There is some evidence on the time trend of employer monopsony power; two studies have shown that employer power increased since the late 1990s (Webber 2020; Langella and Manning 2020), though Bassier, Dube, and Naidu (2020, Table 6) show stability over the 2003–2012 period. One consistent finding of these studies is that employers are able to exert more power over low-wage than other workers, affirming that employer power generates wage inequalities.

What these studies, which overwhelmingly focus on recent data, convincingly demonstrate is that employers wield power over U.S. wages. These studies also point to influences that seem to blunt the effect of employer power—like unionization, high-pressure labor markets, and high values of minimum wages. For example, Bassier, Dube, and Naidu (2020), examining the 2003–2012 period, find that monopsony power was greater in the 2007–2010 period of very high unemployment than in the preceding or subsequent years. Langella and Manning (2020) find monopsony power in the U.S. declining in recoveries (1996–2000 and 2002–2007) and increasing as unemployment escalates (2000–2002 and after 2008 but continuing through 2016).

What these studies suggest is that employer power is ubiquitous in labor markets and, absent institutions and policies that provide countervailing power, wages will be lower and wage growth suppressed. One way to interpret the review of evidence in the current paper is that employer power is the constant of modern labor markets, but what has changed over the past generation in the United States to generate anemic wage growth is the erosion of institutions and policies—high-pressure labor markets, unions, and binding minimum wages—that once provided countervailing power.

Naidu and Sojourner (2020) usefully place this shift into a broader context:

It is not clear that monopsony power has increased over time, but that is also not necessary to make monopsony an important force for explaining inequality…. It is clear that the scope that employers have to exercise the market power they do hold has increased over time, as argued by Erickson and Mitchell [2008]. In an economy without unions, without strong internal labor markets, and with low-cost worker performance monitoring, the forces that may have restrained employers from exercising the latent market power they held (collective bargaining agreements, implicit seniority rules, within-firm equity norms, and efficiency wages) have diminished. It may not be that monopsony has gone up, but it is certainly true that the countervailing forces have declined.

In this light, this paper’s analysis can be considered an examination of “what has changed” in the face of ongoing employer power in labor markets.

This new monopsony literature provides a top-down analysis, which has primarily focused on estimating the aggregate scale of employer power. Some of the recent contributions have started to identify the underlying factors, examining the role of unionization, high-pressure labor markets, and high values of minimum wages, in explaining an aggregate metric of monopsony power. In contrast, our study is a bottom-up analysis examining the impact of myriad specific factors and gauging their contribution to the productivity–median compensation divergence over the past four decades.

Explaining the divergence between productivity and median hourly compensation growth

In examining the corporate and government policy levers that have suppressed wages, our conclusion is that they can account for the vast majority of wage suppression. In the next section we find that these policies can account for the entirety of the fall in wages at the 10th percentile and the corresponding growth of the wage gap between low-wage workers and the median worker.

It can be difficult to assess causality and take interactions into account. But looking at the sum of the impact of the key factors supports the narrative that intentional policy decisions (either of commission or omission) have generated wage suppression. Analysts may differ on the assessment of particular factors, but our hope is that this compilation inspires further efforts, including ones for which we do not have sufficient empirical work to even make guesses.

How much needs to be explained? Above we estimated that by 2017 rising inequality and the erosion of labor’s share of income had lowered compensation of median workers by roughly 43% relative to net productivity growth. Table 3 summarizes the estimated impact on median hourly compensation of the factors that have driven rising inequality and the decline in labor’s share of income. The first panel details the productivity and compensation trends. Between 1979 and 2017 net productivity (economywide productivity net of depreciation) grew 56.0% while median hourly compensation (wages and benefits) grew 13.0%, leaving a 43.0% divergence. If we used hourly compensation of production workers (82% of payroll employment) as the pay measure rather than the median wage, the divergence would be even greater. Benefits are included in these analyses using the ratio of compensation to wages from the National Income and Product Accounts to convert wages to compensation. By deflating both net productivity and the pay measures by the CPI-U-RS index, we have stripped out the influence of differing deflators (for productivity and compensation) from our calculation of the divergence, leaving only the changes in labor’s share of income and changes in compensation inequality as drivers of the divergence (see Bivens and Mishel 2015).

Table 3

The impact of specific factors on the growth of the median wage is detailed in the second panel and discussed next. For purposes of the analysis we equate the impacts on median wages, identified above, with that on median hourly compensation: this is not a consequential decision since the 13.0% growth of median hourly compensation over the 1979–2017 period just slightly exceeded the 12.2% growth of median hourly wages.54

The share of the various factors in explaining the overall divergence of net productivity and real hourly compensation is presented in Table 4 and illustrated in Figure I (examining growth, in percent, of factors) and Figure J (examining growth, in dollars, of factors).

Table 4

Figure I

Figure J

Austerity macroeconomic policy (excessive unemployment)

The impact of excessive unemployment caused by contractionary macroeconomic policy, promulgated to control inflation, (suppressing labor costs in the name of controlling inflation) reduced wages for the median worker by 10.0% between 1979 and 2017. Adjusting for the “flattening” of the Phillips curve since 2008, as we do here, lessens the impact of higher unemployment on wage growth; without this adjustment the impact would have been 12.2%. These contractionary policies caused unemployment to remain 0.8 percentage points above even a conservative estimate of full employment (the NAIRU)—5.5%—between 1979 and 2017, a sharp contrast from the 0.51 percentage points that unemployment remained below the NAIRU in the prior 30 years. If the unemployment rate had been held lower, say to 5% on average, then median wages would have been about 18.3% higher by 2017.

Of course, a 5.5% target for full employment is a modest goal, and if policymakers had achieved a reasonable target of 4.5% the impact of excessive unemployment would be double the 10% presented in Table 3.

Erosion of collective bargaining

The erosion of collective bargaining had an adverse impact by lowering the wages of non-college-educated workers, particularly men, and has also lowered the wages and benefits of nonunion workers in sectors where collective bargaining had previously set wage patterns. Fortin, Lemieux, and Lloyd (2021) estimated that eroded unionization (including spillovers) explains 37.3% of the growth of the 90/50 wage gap (which rose 0.33 log points) among men over the 1979–2017 period, equivalent to a 0.32 percentage point annual decline and a 12% (log point) decline in wages over the entire period for the median male earner. Union coverage fell less for women (due to less unionization among women in 1979), so union erosion explains only 13% of the growth of the 90/50 wage gap (which rose 0.28 log points) among women, equivalent to a 0.10 percentage point annual decline and a 3.6% (log point) decline in wages over the entire period for the median female earner.

We relied on an unpublished analysis of the Fortin, Lemieux, and Lloyd (2021) model to pinpoint at 7.9% the impact of deunionization on the median wage of all workers (men and women combined) for the 1979–2017 period:

Deunionization raised the log 90/50 wage gap by 7.7 log points, almost entirely by reducing the median hourly wage by 7.6 log points, or by 7.9% (0.2 percent annually). The impact on men is larger, with deunionization lowering the male median wage by 10.9 log points, or 11.6% (0.29 percent annually). Deunionization therefore explains 33.1% of the 23.2 log point growth of the 90/50 wage gap over the 1979–2017 period.

Above we noted that Mishel, Rhinehart, and Windham (2020) show that automation and globalization can explain less than a fifth of the erosion in private-sector collective bargaining; the lion’s share has been due to changes in corporate practices enabled by lax legal protections and changing rules.

Globalization on capital’s terms

Bivens (2013) found that, by 2013, trade flows with low-wage nations were likely reducing wages for workers without a four-year college degree by roughly 5.6%. For a non-college-degreed worker making the median hourly wage and working full time, full year, this translates to about $2,000 annually. This estimate is nearly identical to what Autor, Dorn, and Hanson (2013) find in a regression-based investigation of the wage impacts of imports from low-wage countries.

Policy decisions have amplified globalization’s downward wage pressure. Trade agreements, for instance, have been the result of corporate capture that has engaged in selective and regressive protectionism that has severely restricted the policy space of our trading partners.

Impact of the top three factors

As summarized in Table 4 and Figure I, together these three factors alone—excessive unemployment, eroded collective bargaining, and corporate-driven globalization—can account for a 23.5% decline in the median wage from 1979 to 2017 and for 54.7% of the divergence between net productivity and median hourly compensation. Figure J shows that excessive unemployment, eroded collective bargaining, and corporate-driven globalization lowered the growth of median hourly compensation by $5.45: absent these factors median hourly compensation would have risen to $28.59 rather than to $23.59.

Weaker labor standards

The failure to update the value of the federal minimum wage is a premier example of policy action shaping the wage structure and undermining the wages of the bottom third of earners (heavily women and minorities), or 46 million workers. The minimum wage’s impact probably does not extend to the median, so we express that as zero in Table 3. The impact on the 10th percentile wage is considered below.

The erosion of other labor standards likely had an impact throughout the wage structure. Overtime protections for salaried workers declined precipitously and nicked median workers’ wages by 0.3%, while growing misclassification of workers as independent contractors lowered the median wage by 1.0%. Other practices and policies, like lax protections against wage theft, the increased presence of undocumented workers and guestworkers, and more extensive racial discrimination have likely lowered wages, but we are not able to provide an empirical assessment.

Employer-imposed contract restrictions

Employers have increasingly required employees to relinquish various rights when they accept employment, or even after they are already employed, through agreements regarding noncompetition and forced arbitration. Employers within franchise chains have also colluded against employees through anti-poaching agreements, which limit workers’ employment options. The effort to quantify the impact of these policies is still in the beginning stages. We estimate that noncompete agreements have reduced the median wage by 2.25%, but we have not been able to derive good estimates of the impact of forced arbitration (now covering more than half of nonunion employees) nor of anti-poaching agreements among franchisers.

Changes in corporate structure

Changes in corporate structure—from deregulation to fissuring to rising market concentration—likely pushed down wages by at least 5% by 2017. A speculative gauge of the impact of a shift of 15 percentage points of employment into fissured workplaces where wages are 15% less would imply an overall decline of wages of 2.25% and probably an even larger decline at the median. Wilmers’ (2018) estimated that the increase in dominant buyers lowered annual earnings by 3.4% over the 1979–2014 period among workers in publicly owned nonfinancial firms; the impact presumably fell disproportionately on low- and moderate-wage workers. Industry deregulation in airlines, trucking, interstate busing, telecommunications, utilities, and railroads permanently lowered wages for blue-collar workers in the affected industries. Fortin and Lemieux (1997) estimated that 9% of the workforce in the 1980s was affected by industry deregulation, and this explains about 7% of the rise in male wage inequality between 1979 and 1988. Increased product market concentration’s direct impact on prices and profits may not have added to wage suppression, though the leverage of monopoly enables fissuring and the dominant buyer’s ability to squeeze wages and profits in supplier chains.

There is likely to be some double counting when aggregating the fissuring and dominant buyer factors, but it is also likely that the unassessed components of corporate structures exerted at least as much downward wage pressure to offset it.

Automation/skill-biased technological change

As detailed earlier and in Appendix A, automation and skill-biased technological change are prima facie implausible explanations of the wage suppression or wage inequality experienced at least since 1995. It is notable that productivity and investments in capital, software, and information equipment—the footprints of automation—have all decelerated, at least since 2005. Moreover, the rate of change of occupational employment patterns has been historically slow. Indirect or deduced measures of automation-driven relative demand for skills (measured in terms of a college education) have also grown remarkably slower since the mid-1990s (Autor, Goldin, and Katz 2020; Autor 2017). In addition, occupational job polarization has not been present since the late 1990s (Autor 2010; Autor 2015; Mishel, Shierholz, and Schmitt 2013; and others). Given this deceleration of the salient indicators of automation and automation’s impact on key labor market metrics (relative demand for college education, occupational polarization), we assign no impact in Table 3 to automation or skills gaps in driving the productivity–pay divergence for the 1995–2017 period. Our discussion above did not cover the 1979–1995 period, so we do not assign any impact one way or another to automation and skills gaps then. We remain skeptical that there was any impact, though, following the analysis in Mishel, Bernstein, and Schmitt (1997a) and Card and DiNardo (2002).

Suppressed wage growth at the 10th percentile and growth in the 50/10 wage gap

We can also draw on our analyses of the various factors to explain the trend in the wage gap in the bottom half—between the median and the 10th-percentile earner—and the disappointing wage growth at the bottom. Recall that wages at the bottom fared similarly to those in the middle from about 1987 onward, and for low-wage workers wages plummeted in the 1979–1987 period, especially among women. The 50/10 wage gap grew only in the 1979–1987 period. Two factors, the lowering of the inflation-adjusted value of the federal minimum wage and excessive unemployment, can readily explain these trends.

Contribution of the eroded minimum wage

The failure to maintain the value of the minimum wage since 1979 greatly shaped the wage gap in the bottom half and the level of wages at the 10th percentile; this was discussed above and presented in Table 2, based on analysis of the Fortin, Lemieux, and Lloyd (2021) model (which includes spillover effects). Table 5 draws on these results and shows that the fall in the minimum wage lowered the 10th percentile wage by 16.7 percentage points between 1979 and 2017, far more than the actual 2.1% decline of the 10th percentile wage. Almost the entire growth—15.0 percentage points—of the 50/10 wage gap can be explained by a lower minimum wage (impact of 13.8 percentage points); absent that the 50/10 wage gap would have shrunk modestly (by 4.7 log points). It was primarily the lowering of the real value of the minimum wage in the 1979–1987 period that drove the growth of the 50/10 wage gap and the lowering of 10th percentile wages, especially among women.

Table 5

These estimates of the minimum wage’s impact on low-wage earners would be even greater if the counterfactual were maintaining the threshold relative to the median either as it was in 1979 or at the historic peak value in 1968. For instance, the ratio of the minimum wage relative to the median wage of all full-time, year-round workers fell from 52.8% in 1968 to 45.7% in 1979 and to 33.3% in 2017 (Cooper 2019, Figure C). Thus, the minimum wage in 2017 would have to increase by 37% to reach the 1979 level or by 59% to reach its level of 1968. In contrast, if the standard is the real value of the minimum wage, in 2017 it would have needed to be increased by 26% to attain its 1979 level. The minimum wage’s threshold could be targeted at a much higher level if not for the dramatic shortfall of the growth in the median wage relative to net productivity—the 43% shortfall highlighted in this paper.

Contribution of excessive unemployment

Excessive unemployment had a somewhat larger impact on low-wage than middle-wage workers. Had unemployment averaged 5.5% rather than the 6.3% that prevailed over the 1979–2017 period, the wages of the 10th percentile worker would have been 11.6% higher and the 50/10 wage gap would have been 2.7 percentage points lower (Table 4). As discussed above, these estimates take into account the “flattening” of the Phillips curve post-2008. We would note that the impact of higher unemployment would be double what is presented here if we assumed a baseline for full employment of 5.0%.

Summing up the factors

In all, the policy-driven factors delineated in Tables 3, 4, and 5 explain a vast share of the divergence between productivity and median hourly compensation and changes in the wage gap in the bottom half.

The best-measured impacts, those for excessive unemployment, eroded collective bargaining, and corporate-shaped globalization, can account for 23.5 percentage points (or 55% of the total) of the 43% productivity–median compensation divergence. The harder-to-measure impacts of other factors (lowering of the overtime threshold for salaried workers, misclassification, noncompete agreements, and changes in corporate structures like fissuring) can collectively account for another 9.2% of the erosion of the median wage and explain another 21.4% of the divergence. These sum up to explaining about three-fourths (76.1%) of the divergence (Table 4 and Figure I). This is an understated conclusion since there are many additional policy factors that we have not been able to empirically assess: wage theft, guestworker programs, racial discrimination, industry deregulation, exploitable immigrants, forced arbitration, and anti-poaching agreements.

Excessive unemployment and the erosion of the minimum wage more than fully explain the (limited) growth of the wage gap in the bottom half and the substantial shortfall in the growth of the 10th percentile wage (which fell modestly even though net productivity rose by more than half). Other factors such as exploitable immigrants with limited legal rights, increased wage theft, and employee misclassification also depressed wages at the bottom.

In short, without all of these policy-driven changes in the U.S. economy, the bulk of the gap between typical workers’ pay and economywide productivity would not have occurred, and wages at the 10th percentile would have risen instead of fallen. Limited access to courts, collective bargaining, and employment and the inability to rely on government labor standards have systematically weakened the options that workers have to improve their wages, hours, and working conditions.

There are reasons to believe that the impact of these factors is larger than the sum of their individual effects. One way of understanding what has happened is to gauge all the ways that an individual workers’ options to obtain better employment conditions or to affect their current employment have been increasingly foreclosed—limiting both exit and voice. When workers want to improve their conditions of work, they have increasingly limited options to organize a union, rely on adequate and enforceable government standards (e.g., the minimum wage, safety and health, overtime, anti-discrimination, correct classification), or make employers accountable through litigation. Exit is more limited because of anti-poaching agreements, noncompetes, and generally higher unemployment, and the downward pressure on their wages is intensified by globalization, fissuring, and dominant buyer power. Increasingly, resistance is futile.

The lessons here are simple. Wage growth has been greatly directed by policy decisions and is a political variable. It responds—robustly—to big policy changes. But for decades these policy decisions have gone in the wrong direction. Policymakers can deliver prosperity to the vast majority of U.S. workers based on faster wage growth. Whether workers obtain a fair share of the economy’s gains in the future will depend not so much on abstract forces beyond their control but on demanding that their political representatives restore bargaining power to workers, individually and collectively.

Acknowledgments

We gratefully acknowledge the assistance of people who provided information or commented on particular issues: Eileen Appelbaum, Danny Blanchflower, David Cooper, Daniel Costa, Nick Hanauer, Thomas Lemieux, Mike Lipsitz, Celine McNicholas, Suresh Naidu, Heidi Shierholz, John Schmitt, Andrew Sharpe, Anna Stansbury, Evan Starr, Marshall Steinbaum, Thea Lee, David Weil, Nathan Wilmers, and Ben Zipperer. Melat Kassa provided essential, excellent research assistance. We also are appreciative of the funding of the Economic Policy Institute’s Unequal Bargaining Power initiative by the Nick and Leslie Hanauer Foundation, the William & Flora Hewlett Foundation, and the Bernard and Anne Spitzer Charitable Trust.

Appendix A: The failure of automation and skill gaps to explain wage suppression or wage inequality

That U.S. workers have “skills deficits,” that is, lack the skills necessary to deal with technological change, including primarily automation, has been the predominant explanation offered by economists, pundits, policymakers, and the media to explain sluggish wage growth and inequality in the United States, at least until recently. This is the skill-biased technological change hypothesis, which points to the increased use of computer equipment in the workplace and the onset of the information age.

This narrative is sometimes presented as explaining the wage gaps between “skilled” and ‘‘unskilled” (meaning those without a college degree) earners and the disappointing wage growth for the vast majority. This appendix will show that the skills deficit/automation claim has always been a weak explanation for the trends since 1979, and, since the mid-1990s, all indications are that there is no basis at all for considering automation as a significant factor in wage suppression or the growth of wage inequality. For this reason, center-left economists have increasingly stopped highlighting these factors in discussions of the wage problems we face, though the narrative lingers among the punditry.

The conventional wisdom on automation and wage inequality

The general view of the last 30 years is that inequality and wage stagnation are the result of technological change in the workplace, meaning automation, and globalization driven by either technological advances or the political decisions of U.S. trading partners (China’s decision to join the world trading system, for example). These trends were seen as not only inevitable but desirable, in that the harm to workers is the byproduct of forces one would neither want to nor could change. The only appropriate remedy is to adapt, primarily by upgrading workers’ skills and education and perhaps by providing a more adequate safety net.

We examined in the body of the paper the impact of globalization for wage trends and drew two conclusions: Globalization played a nontrivial role in lowering the wages of non-college-educated workers, and this downward pressure has been strengthened by policy decisions creating selective and regressive shaping of the global rules. This appendix focuses on the automation/skills deficit dimension.

The automation narrative in the 1990s (the Clinton years)

In the 1990s center-left economists settled on skill deficits as the dominant explanation for growing wage inequality. As noted by President Clinton’s Council of Economic Advisers (1995):

The sluggish growth of incomes is due to dramatic changes in technology and in global competition that have affected industrialized economies around the world, reducing the relative demand for workers with less education and training…. [M]ost economists believe that a shift in the demand for labor in favor of more highly skilled, more highly educated workers has played a key role. Intensifying global competition is also cited as a factor in putting downward pressure on the wages of less educated workers. However, a number of studies have found that the easily measured direct effects of trade on the wage distribution were small….

An array of economists adopted the automation narrative in the 1990s. Conservatives drew on the work of Kevin Murphy, Gary Becker, and others to conclude that skills deficits explained wage gaps, pointing to higher returns to both education and other, unobservable (meaning not captured by specific variables in survey data) skills. The center-left drew on the work of Katz and Murphy (1992) and, later, an important book by Goldin and Katz (2008), The Race Between Education and Technology. While conservatives have produced little new empirical work on wage inequality in the last 20 years or so, the center-left has focused, until very recently, on developing a newer version of skill-biased technological change centered on the polarization of occupational employment patterns (Autor, Katz, and Kearney 2006; Autor 2010; Acemoglu and Autor 2011; Kearney, Hershbein, and Boddy 2015).

The automation narrative in the 2010s (the Obama years)

This narrative was the dominant one offered by the Obama administration, and President Obama (USA Today 2018) was still offering it in September 2018:

This change has happened fast, faster than any time in human history. And it created a new economy that has unleashed incredible prosperity. But it’s also upended people’s lives in profound ways. For those with unique skills or access to technology and capital, a global market has meant unprecedented wealth. For those not so lucky, for the factory worker, for the office worker, or even middle managers, those same forces may have wiped out your job, or at least put you in no position to ask for a raise. As wages slowed and inequality accelerated, those at the top of the economic pyramid have been able to influence government to skew things even more in their direction: cutting taxes on the wealthiest Americans, unwinding regulations and weakening worker protections, shrinking the safety net. So you have come of age during a time of growing inequality, of fracturing of economic opportunity.

In this story, rapid technological change has led to substantial growth overall and for those at the very top—the ones with “unique skills” and “access to capital”—while those without unique skills and access to capital experience diminished demand and are unable to push for higher pay.

Similarly, a leading conservative labor economist and chairman of the George W. Bush Council of Economic Advisers from 2006 to 2009, Ed Lazear, offered the automation story in the Wall Street Journal in 2019:

How are American workers doing? Neither the middle class nor the poor have fared well in recent decades—but don’t blame tax cuts, a too-low minimum wage or the greed of the 1%. In rich countries around the world, the top half of the income distribution has been pulling away from the bottom half. Productivity growth among high-wage workers, driven by technological change, is the reason…. The likely explanation is that changes in trade and technology have raised the productivity of highly trained, highly educated workers relative to the less skilled. Wages tend to move with productivity, so that if differences in worker productivity grow, wage differences will also grow. (Lazear 2019)

The skills deficit story sounds logical, but it’s not true. It fails to explain wage patterns over the last four decades, and it is a prima facie implausible explanation for at least the last 20 or more years (since 1995 or 1999). Contrary to President Obama’s contention, technological change (or at least automation) has not been especially rapid in the last dozen years or so (there’s actually been a substantial deceleration). Further, the contention that there has been a shift in the demand for labor in favor of more highly skilled, highly educated workers was not true in the late 1990s and has happened far more slowly in the 2000s than in earlier periods.

Automation: A flawed explanation

There have been two versions of the automation narrative, one based on education wage differentials, used to explain the trend in 1980s, and one based on “polarization” in occupational employment, used to explain the 1990s and beyond. Both versions were recently offered by economist David Autor as being among four explanations of wage inequality (Greenhouse 2020). Autor was asked, “[D]espite productivity gains of around 75% since the 1970s, the average American worker saw their wages increased by only 20% since then. There are numerous theories as to what has caused this…. How do you think this happened?” He responded:

There’s no single answer to that question. I would say there are four really important forces. One of them is educational attainment—the rate of growth of educational attainment in the United States actually slowed in the early 1980s. So the rate at which people were completing college, but the demand was growing for college-educated workers. And that led to a lot of rising inequality, just because the wages of the more educated rose relative to others….

A second force is the direction of technological change, which has increased the value of abstract reasoning of creativity, of expertise, of judgment, and devalued a lot of skilled work that people did that followed well-understood rules and procedures. So that would be many clerical jobs, phone answering jobs, calculating, accounting, bookkeeping, copying and filing, but also many production jobs, which often involved skilled, repetitive tasks. But increasingly once we understand the rule book for that type of work, it’s feasible to encode it in software and have it executed by machines or by computers. [Autor mentioned two other forces: “trade pressure” and institutional changes that eroded worker protections.]

This section will briefly review the failures of both versions of the automation narrative to explain the key wage patterns identified in the paper. Our discussion focuses on automation and not “technology” or “technological change,” since the latter terms imply a more general dynamic than one in which the implementation of new technologies in workplaces substitutes software and equipment for labor.

The education narrative

The automation story based on educational differentials sees wage inequality as being driven by increasing education wage gaps. The reasoning is that workplace automation has had a “skill bias” in recent decades, meaning that automation has largely just reduced the demand for a subset of workers—those largely without four-year college degrees. In most discussions skill is equated with what people obtain with four-year college degrees. If the supply of college-educated workers keeps up with the demand for college graduates driven by ongoing automation, then the wage premium for having a college degree will be flat and there will be no increase in wage inequality. If supply fails to keep up with automation-driven relative demand—leading to “skill deficits”—then the relative price of college graduates will rise and drive up wage inequality. Some have argued (as mentioned earlier) that the supply of college graduates faltered around 1980 and failed to keep up with growing demand from ongoing automation.

But for automation to cause a change in wage patterns, technology/automation has to outrace skills.55 It is not enough for automation to be occurring in workplaces or to continue in the same manner as before. Automation, after all, has been a force in workplaces for over 200 years, while education levels have also grown rapidly. Moreover, it is possible for automation to be a large and ongoing force in shaping the pattern of jobs/occupations (rising white-collar and declining blue-collar employment shares) without it generating wage inequality. For instance, automation was ongoing in the 1950s and 1960s when real median wages rose and wage inequality did not increase.

For a hypothesis to have resonance it should connect the major observations within its purview, and the theory that skill-biased technological change explains growing wage inequality fails across many dimensions.56

Top 1% wage growth. The leading research promoting education wage differentials as the driver of overall wage inequality (Katz and Murphy 1992, Goldin and Katz 2008) does not address the redistribution to the very top. Some might attempt to explain this as increased returns to the skills of executives and professionals in finance corresponding to the rise in the college wage premium. But there is no persuasive evidence to support a skill explanation for rising top 1% wages and income or, specifically, the superlative executive compensation growth and expansion in the financial sector that lays behind it (Bivens and Mishel 2013).

“The more you learn the more you earn.” The education story, at least as applied to the 1980s in Katz and Murphy (1992), offers to explain both the rising 90/50 and 50/10 wage gaps as reflecting the rising relative wage differentials between every level of education: college over some college, some college over high school, high school over less-than-high-school. Such an explanation, however, cannot explain why automation did not generate rising education wage differentials between dropouts, high school graduates, and associate college graduates after the 1980s. The flat or declining 50/10 wage gap in the 30 years after 1987 is inconsistent with the skills-gap narrative, since middle-wage workers who have more education than low-wage workers have not reaped a growing advantage since then. Acemoglu and Autor (2012) identified this as a major failure of the education narrative in their review of Goldin and Katz (2008). Mishel, Bernstein, and Schmitt (1997a) made that same point years earlier. (The changed behavior of the 50/10 wage gap in the 1990s—stable or flat rather than growing, as in the 1980s—has been cited by proponents of the occupational employment polarization story as a motivation for adopting this new framework; see Mishel, Shierholz, and Schmitt 2013 for the history.)

The sharp deceleration in automation-driven relative demand for college graduates in the mid-1990s. The college wage premium flattened in the early and mid-1990s. Several studies, all by proponents of the automation explanation, found that the impact of automation on the relative demand for college graduates substantially declined after the mid-1990s relative to earlier decades. Autor (2017) updated the Katz and Murphy (1992) model and showed that the automation-driven relative demand for college graduates decelerated by a third in the early to mid-1990s. Goldin and Katz (2007, Table 1) also showed a large deceleration in 1990–2005 relative to earlier decades going back to 1950, noting “a slowdown in demand growth beginning in the early 1990s” (p. 6).

Autor, Goldin, and Katz (2020), updating the Goldin and Katz (2008) metrics from The Race Between Education and Technology, confirmed the dramatic deceleration of automation’s impact:

[T]he model’s results…divulge a puzzling slowdown in the trend demand growth for college equivalents starting in the early 1990s. Rapid and disruptive technological change from computerization, robots, and artificial intelligence is not to be found though the impact of these technologies may not be well captured by this two-factor setup.

Their results (based on Autor Goldin, and Katz 2020, Table A2) show a deceleration in growth of relative demand for college graduates in the 1999–2017 period relative to earlier periods: a 45.8%  deceleration relative to the 1979–1999 period and a 41.8% deceleration relative to the longer 1959–1999 period. The period since 1999, therefore, has been one featuring a historically small impact of automation on (relative) demand for college graduates.

If automation’s impact has been far less in the last 25 years than in earlier decades, then it cannot explain the ongoing strong or even faster growth of wage inequality in the top half, illustrated by the growth of the 95/50 and 90/50 wage gaps.

It is ironic that just as the education narrative was becoming dominant in the mid-1990s the actual automation-driven relative demand for college graduates became markedly slower, negating the story that automation’s impact was accelerating and causing inequality.

Growing within-group wage inequality. The rise of education wage differentials is, at best, only a partial explanation of rising wage inequality because roughly 60% of the increase is due to greater inequality within education groups (Mishel et al. 2012, Table 4.20; Autor, Goldin, and Katz 2020). Autor, Goldin, and Katz (2020) acknowledge that growing within-group wage inequality is a challenge to the automation narrative.57

Stagnation of wages for college graduates. The story that automation-induced unmet demand for college graduates is lifting the wages of those with more education while punishing the wages of those with less is belied by the actual labor market experience of college graduates. For one, the inflation-adjusted wages of college graduates did not rise between 2000 and 2014, making the “lifting of the most educated” story not very convincing. The widespread use of unpaid internships for college students and graduates provides further evidence that employers do not have “unmet needs.” There is also ample evidence that the wages of entry-level college graduates slumped in the 2000s (Gould 2020) and that many young college graduates filled jobs that did not require a college degree (Abel and Dietz 2014). The median annual wage of recent college graduates, according to the New York Federal Reserve Board, rose by only 1% from 2000 to 2019, hardly a sign of winning in a race between education and technology/automation.

The slow growth of the college wage premium since 1995 or 2000. This fact makes the education wage-gap narrative a prima facie implausible explanation for the growing wage gap in the top half. While the college wage premium grew minimally since 1995 and especially since 2000, the 95/50 wage gap continued to grow strongly. The log 95/50 wage gap rose 0.76 points per year in 1979–2000 and rose even faster, by 1.00 log points per year, over 2000–2019 (see Appendix Figure A). In sharp contrast, the key education wage gap, the college–high school wage premium, grew far more slowly in the latter 2000–2019 period—hardly growing at all, just 0.13 log points per year, 87% slower than in the 1979–2000 period.

Appendix Figure A

As noted above, the growth of the wage gap in the top half, illustrated by the 95/50 or 90/50 gap, is a key wage pattern that needs to be explained, as it was the only source of growing wage inequality other than that of the top 1% in the last three decades. These data show that any explanation of wage inequality based on education wage gaps is implausible for the period since 2000 (and probably since the mid-1990s). That is, it is implausible that a sharply decelerating growth of the college wage premium can help explain an accelerated growth of the 95/50 wage gap since 2000.

The slowdown of college completion. The education wage-gap narrative sometimes focuses on the slower growth of college completion post-1979 as the cause of wage inequality. This makes sense within the conventional framework that ignores the role of other factors—globalization, weaker unions, lower minimum wages, and so on—besides supply and automation-driven relative demand. Nevertheless, the slowdown of college completion cannot explain wage inequality since the early to mid-1990s, the era of slight growth in the college wage premium. After all, slow college completion is said to increase wage inequality as the relative supply of graduates falls short of relative demand. The growth of the relative supply of college graduates, according to Autor, Goldin, and Katz (2020, Table 3), was very slow in the most recent two decades (1999–2017), yet the college wage premium barely grew.

The framework put forth by Katz and Murphy (1992) and by Goldin and Katz in The Race Between Education and Technology (2008) relies on competitive labor markets driven solely by relative supply of education and automation-driven relative demand for education. The notion that the relative demand for education (as a proxy for automation) can be deduced from education wage premiums and supply trends presumes that factors other than automation (unions, globalization, minimum wages, corporate structure changes, and others) have no impact. The evidence presented in this paper indicates otherwise. Simply put, the college–high school wage premium goes up and down for lots of reasons besides automation and supply factors, and we cannot readily deduce automation’s impact from data on wages and education supply.

There have been many critiques of the skills deficit/education wage-gap narrative in the past, and the analyses have stood the test of time; these include Mishel and Bernstein 1994; Mishel, Bernstein, and Schmitt 1997a, 1997b; Mishel and Bernstein 1998; Howell and Wieler 1998; Galbraith 1998; Howell 2001; Card and DiNardo 2002; Mishel, Shierholz, and Schmitt 2013; and Howell and Kalleberg 2019.

The job polarization automation narrative

A new automation-causes-wage-inequality narrative emerged around 2005 to replace or supplement the education wage-gap narrative and to overcome one of the latter’s key weaknesses—the inability to explain why the 50/10 wage gap grew in the 1980s but flattened or declined in the 1990s. The job polarization story claims that the nature of automation changed starting in the 1990s such that automation replaced middle-wage occupations/jobs more than jobs in low-wage and high-wage occupations. This is demonstrated by mapping the changes in occupational employment patterns by wage level. That is, this story relies heavily on the expansion of low-wage occupations characterized by routine manual tasks, the expansion of high-wage occupations requiring abstract, nonroutine tasks, and a corresponding shrinkage of occupations in the “middle,” which perform routine manual tasks.

Autor, Katz, and Kearney (2006) articulated this narrative:

[T]hese models also imply a puzzling deceleration in relative demand growth for college workers in the early 1990s, also visible in a recent “polarization” of skill demands in which employment has expanded in high-wage and low-wage work at the expense of middle-wage jobs. These patterns are potentially reconciled by a modified version of the skill-biased technical change hypothesis that emphasizes the role of information technology in complementing abstract (high-education) tasks and substituting for routine (middle-education) tasks.

This occupational employment polarization narrative was introduced into the policy world in a Center for American Progress/Hamilton Project paper by David Autor (2010) and articulated as the necessary replacement for the education narrative in Acemoglu and Autor (2012).

But the job or occupational polarization story fails to explain key trends and wage patterns:

Growth in top 1% wages. The job polarization narrative, as with the education wage-gap narrative, fails to address the superlative wage growth of the top 1% and top 0.1%, a major dynamic driving wage inequality.

Lack of evidence of job polarization since 1999 or in the 2000s. This was demonstrated in Mishel, Shierholz, and Schmitt (2013) and Beaudry, Green, and Sand (2013, 2014) and further confirmed in Hunt and Nunn (2019). The key chart in Autor (2010, Figure 1) showed there was no job polarization in the 1999–2007 period, but that finding was not mentioned in the text. Autor’s paper (2015, 149–50) for the Kansas City Federal Reserve Board’s annual Jackson Hole conference acknowledged the lack of polarization in the post-1999 period through 2012, the latest data available at that time:

Although the polarization hypothesis can explain some key features of the U.S. and cross-national data, reality invariably proves more complicated than the theory anticipates. The clearest evidence for this general dictum is the unexplained deceleration of employment growth in abstract task-intensive occupations after 2000, which is discussed by Beaudry, Green, and Sand (2013, 2014) and Mishel, Shierholz, and Schmitt (2013)…. The final empirical regularity highlighted by Chart 7 is that growth of high-skill, high-wage occupations (those associated with abstract work) decelerated markedly in the 2000s, with no relative growth in the top two deciles of the occupational skill distribution during 1999 through 2007, and only a modest recovery between 2007 and 2012. Stated plainly, the U-shaped growth of occupational employ­ment came increasingly to resemble a downward ramp in the 2000s.

A “downward ramp” and the absence of the “U-shaped growth of occupational employ­ment” are acknowledgments, though expressed in a not-very-straightforward manner, that job polarization was not present between 1999 and 2012.

Obviously, a narrative based on the changing composition of employment by occupation (expanding low-wage and high-wage occupations and shrinking middle-wage occupations) cannot be relevant to explaining post-1999 wage patterns if those occupational employment patterns have not been evident. In fact, the Autor (2014) data show that nearly all of the employment growth was in low-wage occupations, a pattern that does not readily explain why the wage gap between the top and the middle kept expanding and did so at an accelerated pace. For occupational employment patterns to explain top-half wage-gap growth would require finding a rapidly expanding need for high-wage workers able to carry out abstract, nonroutine tasks, a pattern not present since 1999.

Lack of a relationship between occupational employment patterns and wage inequality. Remarkably, the job polarization narrative relies on mapping occupational employment patterns to explain wage inequality but has never presented evidence that these occupational employment shifts affect wages. In fact, Mishel, Shierholz, and Schmitt (2013) show that changes in occupational employment shares (whether an occupation expands or contracts employment relative to other occupations) are not related to changes in relative wages by occupation (whether wages in that occupation rose or fell relative to wages in other occupations). That is, one would expect that occupations that expand (contract) would have rising (falling) wages relative to other occupations. Looking at the relation between changes in occupational employment shares and the corresponding relative wages of occupations, Mishel, Shierholz, and Schmitt (2013) found no relationship in each of the decades of the 1980s, 1990s, and 2000s. It is also worth noting that middle-wage occupations have shrunk and higher-wage occupations have expanded since the 1950s, but median wages and wage inequality have risen and fallen over this time with no apparent correspondence to job polarization trends.

If occupational employment patterns do not affect occupational relative wages, then they certainly bear no relationship to changes in wage inequality, since presumably the mechanism for automation to cause changes in wage gaps is for automation-induced changes in occupational employment patterns to alter the relative wages of occupations. The effort to track occupational employment patterns has no implications for understanding wage patterns.

Footprints of automation

The discussion so far has relied on economic analyses that derive the pace and skill bias of automation from patterns of occupational employment growth or from wage and education supply trends. It is worthwhile to examine other perhaps more direct footprints of automation to discern the pace of automation in recent years compared to earlier periods. These data are illustrated in Appendix Figure B, drawn from Mishel and Bivens (2017) and Mishel and Shierholz (2017).

Appendix Figure B

Automation is what occurs as new technologies are incorporated along with new capital equipment or software to replace human labor in the workplace. Labor productivity and capital investment are both measures of automation in that they necessarily accompany the substitution of capital for labor. Thus, if there were a recent surge of robots or automation, we would expect to see the footprints in trends in productivity, capital investment, and software investment. The trends shown in the figure suggest that automation has been far slower since 2002 than in the three earlier periods: the early postwar years from 1973 to 1973; from 1973 to 1995; and over the high-tech boom years of 1995–2002. There is certainly no evidence of automation having accelerated. These data affirm the findings above that automation, given its slow pace in recent years, is unlikely to have been a major factor driving wage stagnation or wage inequality in the last two decades.

Automation (along with shifts in consumer demand and trade) would be expected to be a major factor in why employment in some occupations expands and employment in other occupations declines. Therefore, we can also examine the pace of change in the occupational composition of employment to deduce trends in automation.

Using the data in Atkinson and Wu (2017), Mishel and Bivens (2017) computed a metric to examine the pace of occupational employment shifts in each decade. Specifically, Mishel and Bivens examined the shares of total employment for each of the 250 occupations in the data for the beginning and end years of each decade and computed the changes in these shares. The sum of the occupational share gains will automatically be equal to the occupational share losses, so the metric of change in each decade is half the sum of the absolute change in employment shares. This metric adjusts for differences in the rate of employment growth in each decade and the absolute employment size of individual occupations. This metric of shifts in occupational employment measures the shares of total employment exchanged between occupations that gain and occupations that lose employment shares each decade.

Mishel and Bivens found that occupational employment changes were “fairly uniform over the 1940–1980 period and far more rapid than for any period since 1980. The period since 2000 has seen the lowest rate of change—half the rate of change of the 1940–1980 period.” These findings indicate that the pace of occupational employment shifts was extremely slow in the 2000s, affirming the deceleration found in the slower growth in productivity, software, and capital investments identified in Appendix Figure B.

Appendix B: The intentionality of macroeconomic policies

There is a growing recognition that unemployment was kept excessively high for the purpose of inflation control over long periods after 1979. Could this have been simply a technocratic mistake, an instance of policymakers underestimating the sacrifice that would be required to keep inflation in check? We think this proposition lets policymakers off the hook too easily.

First, it was completely predictable who would bear the brunt of a contractionary macroeconomic shock, and Fed policymakers, charged with maintaining high employment as well as stable prices, were obligated to construct some portfolio of policies designed to cushion the shock.

Second, macroeconomic policymakers and particularly the Federal Reserve consistently overshot their own conservative estimates of the NAIRU—the nonaccelerating inflation rate of unemployment—during most of the 1979–1995 period. Rather than aiming for stable inflation after the 1970s episode of accelerating inflation, they sought to reduce inflationary expectations to a very small number, and they were willing to tolerate the extended periods of unemployment this required.

Finally, if going the severe contractionary route was a technocratic mistake, it was a foolish one. Keeping unemployment rates above the “natural rate” is an expensive way to reduce inflation, especially since evidence that inflation rates like those experienced in the United States in the 1970s are deeply damaging is very hard to find. In the U.S. postwar experience, there is no consistent relationship between inflation rates and aggregate growth. But large distributional shifts can occur with unanticipated bursts of inflation. Inflation does indeed reduce the real value of wealth stocks, but it also reduces the real value of liabilities. In essence, it causes a redistribution from net creditors to net debtors. So, for example, Americans who bought a home with a fixed-interest-rate mortgage in the late 1960s or early 1970s saw a windfall wealth gain as inflation eroded the real burden of their mortgage obligation. Inflation hawks tend to emphasize particularly sympathetic economic actors who might be hurt by this—retirees living on fixed incomes, for example. Of course, even by the 1970s most retirees’ incomes mainly comprised Social Security benefits, which were largely shielded from price increases.

Against these uncertain aggregate benefits, policymakers should weigh the extremely large and regressive costs of recessions engineered largely to break inflationary expectations. The recessions of the early 1980s and 1990s were both caused in large part by interest rate hikes undertaken by the Federal Reserve to reduce inflation. The cumulative output loss of the two early 1980s recessions approached 80% of one year’s gross domestic product (GDP) at the time, while the early 1990s recession exacted a cumulative cost of nearly one-third of one year’s GDP.

Were these costs necessary for reducing the 1970s inflation? While the 1970s inflation was the first in post-1900 U.S. history to not to be associated with a war effort, its cause was not mysterious: It was the combination of oil price shocks caused by political unrest in the Middle East and the global slowdown in productivity growth. The real price of oil tripled in 1973 (the Yom Kippur War), declined after 1975, and doubled again in 1979 (the Iranian Revolution). These exogenous oil shocks were amplified by wage-price spirals, as both firms and workers tried to raise the nominal prices under their control (product prices and wages, respectively) to avoid bearing the full brunt of adjusting to higher input costs, all without realizing that the economy’s underlying capacity to deliver income growth (productivity) had slowed markedly.

Implicit in this analysis is a view that inflation is at least in part an outcome not just of over-accommodative macroeconomic policy (the conservative view), but of distributional conflict between capital and labor. When bargaining power is more equal, exogenous price shocks take longer to propagate through the economy and cause higher and more persistent inflation. After an exogenous price shock, firms raise their price level to preserve profit margins, at the expense of real (inflation-adjusted) wages. Workers who have some degree of bargaining power can respond by demanding higher nominal wages to claw back the lost ground. Firms then pass on the higher wage costs into higher prices and so on. The longer this process goes on, the steeper and more persistent is the inflation. Conversely, if firms are able to pass on higher prices in response to the initial cost shock and workers lack the bargaining power to demand higher nominal wages in response, then the shock is muted and leads to lower and less persistent inflation.

What made the oil price shocks especially effective in generating wage-price spirals in the 1970s were the atypically strong perceptions held by American workers about their own bargaining power, as well as expectations of real wage growth fostered by decades of rapid and equal economic growth.

Coming into the 1970s, American workers had experienced wage growth on par with productivity growth for most of the past three decades, and this productivity growth averaged 2% per year. They had also come off a decade from 1959 to 1969 when unemployment averaged 4.8% and reached a low of 3.5% in 1969. Further, key labor standards in the U.S. labor market were strong in historical terms. Private-sector unionization rates were 24.2% in 1973, more than twice as high as they are today. The inflation-adjusted value of the minimum wage reached its highest point ever in 1968 after three decades of rising (roughly) with economywide productivity.

An objection to an analysis putting the 1970s inflation at the feet of wage-price spirals has traditionally been that it implicitly blames workers for inflation’s rise, and seems to acknowledge the need for reducing workers’ bargaining power as an anti-inflation strategy. This line of thinking is unjustified and unfair. The root causes of the inflationary episode of the 1970s were two oil price shocks, and a wage-price spiral following such an exogenous shock requires, by definition, both wages (driven by workers’ desire to protect wages’ purchasing power) and prices (driven by firms’ desire to protect their profit margins) to rise. Blaming this spiral on the workers’ side alone is unfair, but typical in our framing of labor-capital conflicts. We’re more likely to hear about “labor militancy” rather than “capital militancy” when the subject is wage-price spirals, but this is probably because it is taken as given that capital will be militant and usually effective in protecting its share of income. Times must be truly remarkable when the economic context allows labor to push back and protect its gains.

Galbraith (1997) makes the obvious inference from this episode:

It would therefore be reasonable to approach anti-inflation policy in general as a matter, first and foremost, of designing circuit breakers for shock episodes, so as to reduce the cost of adjusting to a new pattern of relative prices and therefore the need to do it through the brute-force method of mass unemployment. Some simple steps, like coordinating the timing of wage bargains and providing the president with limited discretion over cost-of-living adjustments in Social Security, federal pensions and other payment streams might help a great deal….

In short, the fact that the 1970s inflation was tamed by the severe contractionary approach does not mean that the approach was most efficient way to tame this inflation, or that the benefits outweighed the costs imposed on the broader economy.

Notes

1. The Republican Party platform reads: “Our economy has become unnecessarily weak with stagnant wages. People living paycheck to paycheck are struggling, sacrificing, and suffering.” The Democratic platform reads: “But too many Americans have been left out and left behind. They are working longer hours with less security. Wages have barely budged and the racial wealth gap remains wide, while the cost of everything from childcare to a college education has continued to rise.”

2. The 44.6% growth is the growth of wages on average over the 1979–2019 period and represents what each group would have experienced absent changes in the distribution of wages.

3. This acceleration is reflected in the growth of the log 95/50 wage gap, which grew 0.75 log points annually from 1979 to 1995 and by 1.0 log points from 2000 to 2019, using the same data as in the figures.

4. This college–high school wage premium is the premium estimated using log hourly wages, and the controls are for gender, race and ethnicity, education, age, and geographic division. The premium is the coefficient on “college,” having a four-year degree but no further education, as the omitted category is “high school.” The sample is wage and salary workers ages 16 and older in the CPS-ORG data; the data are available at https://www.epi.org/data/#?subject=wagegap-coll.

5. See Table A2 in Autor, Goldin, and Katz 2020. The relative supply of college graduates grew 1.96 log points annually from 1999 to 2017, down from 2.28 log points annually from 1979 to 1999. This deceleration of supply should have led to a faster growth of the college wage premium.

6. The most recent version of this chart can be found at https://www.epi.org/productivity-pay-gap/.

7. In addition to these two factors, Bivens and Mishel also identified differential trends in the price deflators used to measure productivity and real median hourly compensation. Since our focus is on the inequality-generating factors–inequality within compensation or a shift toward labor’s share of income­—we compute the divergence so as to exclude the impact of differing price deflators (using the CPI-U-Rs to deflate both productivity and median hourly compensation).

8. Restoring labor’s share to 82.4% means a 9.1% across-the-board increase (82.4/75.5 = 9.1%) in compensation); falling to 75.5% is an 8.4% cut ((75.5/82.4 – 1) = 8.4%).

9. This conclusion is based on regression-based estimates of the Black–white wage gap. Wages are adjusted into 2018 dollars using the CPI-U-RS. The regression-based gap is based on average wages and controls for gender, race and ethnicity, education, age, and geographic division. The log of the hourly wage is the dependent variable. See https://www.epi.org/data/#?subject=wagegap-bw.

10. See https://www.epi.org/data/#?subject=wage-education.

11. The range in each estimate brackets the regression results from using state-specific unemployment rates to predict wage growth, as Bivens and Zipperer (2018) did, and results from using national time-series data only. The lower state-specific results likely stem predominantly from smaller and noisier samples at the state level.

12. Estimates of the wage impact of unemployment on the median and 10th percentile wage are from Bivens and Zipperer 2018, Figure F. The impact of 1 percentage point higher unemployment lowers the median wage by 0.459 and 0.296 in the earlier and latter period and lowers the 10th percentile wage by 0.582 and 0.243 in the earlier and latter period.

13. See Bivens and Zipperer 2018 for a demonstration of how including the post-2008 period reduces the coefficient on unemployment in a wage Phillips curve, and Bivens 2019 for coefficient estimates for just the post-2007 period.

14. These data are the share with collective bargaining coverage for “1979,” calculated as an average of 1977–1980 shares from May CPS data.

15. This discussion draws heavily from Bivens et al. 2017.

16. Evidence for these effects can be found in Mishel et al. 2012: union impact and coverage by demographic groups (Table 4.33) and by education, occupation, and wage fifth (Table 4.37).

17. These estimates look at what wages would have been in 2013 had union density remained at its 1979 levels.

18. Fortin, Lemieux, and Lloyd (2021) find a 7.6 log point impact of deunionization on the median. This amount includes both direct and spillover impacts. Their estimate of the direct union impact on the 90/50 wage gap is 52.8% of the total union impact. Applying that share to 7.6 points yields a 4.0 log point estimate of a direct union effect on the median wage. Anna Stansbury provided a benchmark for comparisons to the Fortin, Lemieux, and Lloyd results based on her work with Summers (2020): “We estimate for the nonfinancial corporate sector a direct effect of union decline on labor rents for noncollege workers of 2.9 percent of compensation over 1984 to 2016.” The Fortin, Lemieux, and Lloyd estimates line up close to, even below, the Stansbury-Summers estimate if one adjusts the Stansbury-Summers results for the longer time period used by Fortin, Lemieux, and Lloyd, 1979–2017. There was a very steep decline in union membership from 1979 to 1984, comprising 40% of the total decline from 1979 to 2016. If one scales the Stansbury-Summers estimate to the total decline since 1979, their estimated impact would be 4.8% (2.9% × [1/60.4%]), an even larger impact than Fortin, Lemieux, and Lloyd.

19. Western and Rosenfeld find that deunionization’s impact (direct and spillover) on wage inequality explained 33.9% of male wage inequality and 20.4% of women’s wage inequality. The comparable estimate for Fortin, Lemieux, and Lloyd (2021) is deunionization explaining 28.8% and 6.7%, respectively, of men’s and women’s wage inequality.

20. Farber et al. (2021, 33) note: “We show that a combination of low-skill composition, compression, and a large union income premium made mid-century unions a powerful force for equalizing the income distribution.” Specifically (p. 24): “During our long sample period, the union premium has remained between ten and twenty log points, with the less-educated receiving an especially large premium. Moreover, the negative effect of unions on residual income variance is large and also relatively stable over time. By contrast, selection into unions is not constant across time. In the Great Compression period, when unions were at their peak and inequality at its nadir, disadvantaged households were much more likely to be union members than either before or since.”

21. This conclusion leaves open, however, the question of how useful or good the status quo rules of the game governing globalization are, even aside from any downward pressure they may put on American workers’ wages. For example, these rules’ effects on the economic development possibilities of poorer countries or the ability of all countries to tax mobile capital are incredibly important issues in their own right. Saying that the downward wage pressure of the globalization status quo could be overcome with an aggressive domestic policy response does not imply that this status quo should be preserved.

22. This estimate is based on the impact of raising the minimum to $15 in 2025 and including the impact on those who received minimum wage increases at the state or local level since 2017. This estimate understates the share of earners affected since it ignores those in states that had a higher minimum than the federal threshold minimum wage in 2017 but did not increase it further since then.

23. This calculation refers to net productivity growth (output per hour taking depreciation into account). Average wage data are drawn from NIPA data, deflated by the CPI-U-RS.

24. Increase from July 2019 to October 2025, as per analysis in Cooper 2019.

25. This estimate assumes 0.5% real wage growth in the median worker’s wage from 2019 to 2025. One obtains similar results if one examines the real value of the minimum wage relative to the average hourly earnings of production/nonsupervisory workers.

26. See https://www.dol.gov/agencies/whd/minimum-wage/history.

27. The analysis provides estimates of the impact of the higher threshold on weekly hours (Table 21) and weekly wages (Table 22). These data allow a computation of the before and after hourly wages that are the basis for these estimates.

28. Those affected on the post-1979 erosion of the salary threshold include three groups. The first are those directly affected by the raising of the salary threshold as proposed by the Department of Labor, the 4.2 million receiving a 1% hourly wage boost. The second group are the broader group, the 8.9 million affected by the 2016 rule change. There was probably a smaller than 1% wage impact for this group, since it is generally more educationally downscale than those directly affected (see Shierholz 2019 analysis of the education composition of those enjoying new protections versus “strengthened protections” from the 2016 rule change). The remaining group are those who earned above the 2016 rule change and the salary threshold in 1979.

29. See Table 1 in Gould 2018 for the decile cutoffs in 2017.

30. Judged by the share of the workers covered in 1979, 39.8%, compared to those who would have been covered by the 2016 rule in 2015, 32.7% (Kimball and Mishel 2016).

31. This section draws heavily on Shierholz 2021.

32. Calculated as 0.68 * 0.1495 = 0.102.

33. “I was also struck by commentary from leaders at digital platforms in Silicon Valley, as well as our own discussions with executives, venture capital companies, and workers. Most indicated that they viewed an independent contractor status as the default option for employment” (Weil 2017).

34. For instance, a rise of 3.1 percentage points of the entire private nonagriculture workforce yields a 0.5% or 0.9% wage reduction if misclassified workers are paid, respectively, 15% or 30% less than regular W-2 workers. Similarly, a rise of 4.6 percentage points among the bottom two-thirds of the private nonagriculture workforce yields a 0.7% or 1.4% wage reduction if misclassified workers are paid, respectively, 15% or 30% less than regular W-2 workers.

35. This discussion borrows heavily from Costa 2019.

36. For additional discussion and context on OECD’s findings, see Costa and Martin 2019.

37. Also see the literature review in Ortega and Hsin 2018 for evidence of wage disparities between unauthorized and authorized workers.

38. Starr, Prescott, and Bishara (2020, 15) state: “The ability to enforce noncompetes should encourage greater noncompete use, more investment, and higher wages, but employers use noncompetes virtually as often in states where they are clearly unenforceable.” Starr, Prescott, and Bishara (2019) find “that noncompetes are associated with reductions in employee mobility and changes in the direction of that mobility (i.e., toward noncompetitors) in both states that do and do not enforce noncompetes…. [W]e find that employees with noncompetes—even in states that do not enforce them—frequently point to their noncompete as an important reason for declining offers from competitors.”

39. BLS data on “median weekly earnings of full-time wage and salary workers by detailed occupation and sex” at https://www.bls.gov/cps/cpsaat39.pdf show weekly earnings for full-time workers in construction and installation, maintenance, and repair occupations at, respectively, 94% and 102% of U.S. weekly earnings. The noncompete ban estimates are in Figure 5 using fixed effects and controls. Mike Lipsitz kindly provided the exact estimates.

40. The incidence of noncompetes in construction occupations (12%) was a bit below the average (14%), while that for installation, maintenance, and repair occupations was above average (18%) (Figure 5, Starr, Prescott, and Bishara 2020).

41. As Starr (2019b) suggests, “it’s not clear if different establishments owned by separate franchises should be considered competing entities.”

42. Technology plays a role in that communications and other technologies better enable firms to monitor dispersed economic activity and monitor the fulfillment of standards. Key researchers in this field have noted the enabling role of technological changes. Weil (2014) wrote, “Information and communications technologies have enabled this hidden transformation of work.” Bernhardt et al. (2016) write: “New information and communications technologies (ICT) have facilitated outsourcing and the decentralization of producing goods and services because ICT lowers the costs of information processing and coordination of work across organizational boundaries, thereby reducing the cost advantages of internal production. ICT also enhances firms’ capabilities to monitor and enforce contracts with external suppliers, thereby reducing the relative advantages of hierarchy. ICT allows firms to achieve control over productive activities—the advantages of vertical integration—without assuming the risks of actual ownership or the inflexibility of bureaucracy.”

43. Self-employment or independent contracting has been stable for the last 25 years, contrary to the hype about how we are all becoming gig workers and freelancers. The Contingent Worker Survey (CWS) (BLS 2018), the gold standard of data tracking “alternative work arrangements,” found that in 2017 those who were independent contractors (and do not themselves have any employees) on their main job comprised about 7% of all employment, the same as in 1995 and 2005 (BLS 2018). Analysis of tax data through 2016 (Collins et al. 2019) showed that “consistent with the 2017 CWS results, we find no evidence that ‘traditional’ work arrangements are being supplanted by independent contract arrangements reported on 1099s.”

44. There was a modest growth in employment in “staffing firms” as the share of private payroll employment in “employment services” grew from 2.5% in 1995 to 2.9% in 2018 (see Bureau of Labor Statistics Current Employment Statistics (CES) Series CES0500000001 and CES6056130001). Staffing employment in 2018 was actually lower than what prevailed at the start of the last two business cycles in 2000 (3.5%) and 2007 (3.1%). However, the composition of staffing employment shifted as the presence increased greatly in manufacturing (Mishel 2018).

45. Franchising is a part of fissuring to the extent that the franchisee is not simply a retail outlet for the franchisor (“traditional franchising”) but actually “relies on the transfer of a complete business format and methods—including extensive business support (e.g., training of the franchisee) and ongoing monitoring from the franchisor,” called “business template franchising” (Kosová and Lafontaine 2012). Unfortunately, it is hard to track trends in franchising because there are no recent data beyond 2012 (yet), and the data for 2007 and 2012 are not necessarily comparable to what is available in the 1980s and 1990s. Business-format franchisees provided 5.5% and 5.3% of total private-sector employment, respectively, in 2007 and 2012 (this assumes that business-format franchising represents the same share of total franchising employment in 2012 as in 2007). We do not know how much business-format franchising has grown since 2012. We have some historical information on total franchising but not for business-format franchising. According to a discontinued Department of Commerce series of publications, Franchising and the Economy, the share of (all) franchising in private employment was 6.3% in 1978 and 7.9% in 1986, somewhat higher than the 6.8% and 6.5% employment shares in, respectively, 2007 and 2012. Given that traditional franchising has fallen since the late 1970s, we can assume that business-format franchising has increased since the late 1970s, though we do not know how much. (Lafontaine and Blair [2009] say that “the value of goods sold via outlets of business-format franchisors has increased from 2.3% to 3.5% of total GDP” between 1972 and 1986, but we could not locate in the Department of Commerce data they cite any reference to business-format franchising. If Lafontaine and Blair are correct, then business-format franchising in the 2007–2012 period was on the same scale as in the 1972–1986 period.) In sum, we know that the type of franchising associated with fissuring represented 5% or so of total private employment in the most recent data available (for 2012) and did not necessarily significantly grow in the prior decades. This type of fissured franchising may have grown since 2012, however.

46. Analysis of the Census Bureau’s Longitudinal Business Database.

47. Weil bases this estimate on an analysis undertaken for investors by A/B Bernstein.

48. Stansbury and Summers (2020, 27–28) argue: “In addition, while a fall in worker rent-sharing power should not have any implication for firms’ underlying markups (which are determined by their product market power), it does have implications for measured markups. This is because measures of aggregate markups used in recent literature depend on firms’ costs, including firms’ labor costs—even if the labor costs partly represent rents accruing to labor as well as the true marginal cost of production. This implies that markups, as they have been measured in recent papers, cannot be used to distinguish between a story of rising product market power and a story of falling worker power: a rise in measured markups could reflect a fall in worker rent-sharing power just as much as it could reflect a rise in true markups and firms’ monopoly power.”

49. Measurement of labor’s share described in Bivens 2019 and the data are available at the Nominal Wage Tracker: https://www.epi.org/nominal-wage-tracker/.

50. Personal communications with Wilmers. Data based on Figure 1 of Wilmers 2018.

51. Based on personal communications and Wilmers 2018 (p. 230): “[G]rowing buyer power from 1979 to 2014 accounts for around 10 percent of the [35 percentage point] decline in earnings growth relative to the 1955 to 1978 trend.”

52. This discussion borrows heavily from Mishel, Schmitt, and Shierholz 2014.

53. “The percentage of male workers in the upper middle of the distribution, with hourly wages between $7.50 and $12.50 in 1979 dollars, fell from 52 percent to 38 percent in deregulated industries, whereas it fell only from 33 percent to 26 percent in other industries” (Fortin and Lemieux 1997, 85).

54. The ratio of compensation to wages, reflecting the scale of nonwage benefits, rose from 1.220 in 1979 to just 1.229 in 2017. See Bivens and Mishel 2015 for details on the data construction.

55. Mishel and Bernstein (1994) referred to this as the need to show “acceleration of technology/automation.”

56. The term “college graduate” in this discussion is explicitly restricted to those with a four-year bachelor’s degree and excludes those with “some college” or an associate degree and those with an advanced degree beyond a bachelor’s. Some discussions lump the college and advanced degree returns or supply together. That can be misleading when the evidence is used to suggest we increase the number of college graduates if, in fact, the evidence may suggest we need more post-college graduates.

57. Some researchers have asserted, without any empirical backup, that within-group inequality reflects the returns to unobservable (not captured by any metric in our regular data) skills. That is, of course, the only way to preserve the skill-biased technological change story without having to look beyond pure supply-and-demand factors. It is not, however, persuasive without further evidence or even some conjecture about what patterns we would expect returns to unobserved skills to display.

References

Abel, Jaison R., and Richard Deitz. 2014. “Are the Job Prospects of Recent College Graduates Improving?” Liberty Street Economics (Federal Reserve Bank of New York blog), September 4.

Acemoglu, Daron, and David Autor. 2011. “Skills, Tasks and Technologies: Implications for Employment and Earnings.” In Handbook of Labor Economics, edited by Orley Ashenfelter and David Card, 4, part B: 1043–171. Elsevier.

Acemoglu, Daron, and David Autor. 2012. “What Does Human Capital Do? A Review of Goldin and Katz’s The Race between Education and Technology.” Journal of Economic Literature 50, no. 2: 426–63.

Apgar, Lauren A. 2015. “Authorized Status, Limited Returns: The Labor Market Outcomes of Temporary Mexican Workers.” Economic Policy Institute.

Appel, Sam, and Carol Zabin. 2019. “Truck Driver Misclassification: Climate, Labor and Environmental Justice Impacts.” UC Berkeley Labor Center.

Appelbaum, Eileen. 2017. “Domestic Outsourcing, Rent Seeking, and Increasing Inequality.” Review of Radical Political Economics 49, no. 4: 513–28.

Appelbaum, Eileen, and Rose Batt. 2014. Private Equity at Work: When Wall Street Manages Main Street. Russell Sage Foundation.

Atkinson, Robert D., and John Wu. 2017. “False Alarmism: Technological Disruption and the U.S. Labor Market, 1850–2015.” Information Technology & Innovation Foundation.

Autor, David. 2010. “The Polarization of Job Opportunities in the U.S. Labor Market: Implications for Employment and Earnings.” Center for American Progress and Brookings Institution: The Hamilton Project.

Autor, David. 2014. “Polanyi’s Paradox and the Shape of Employment Growth.” Working Paper 17040. National Bureau of Economic Research. https://doi.org/10.3386/w20485.

Autor, David. 2015. “Polanyi’s Paradox and the Shape of Employment Growth.” Reevaluating Labor Market Dynamics (Federal Reserve Bank of Kansas City, Economic Policy Proceedings), 129–77.

Autor, David. 2016. “Will Automation Take Away All Our Jobs?” TEDxCambridge.

Autor, David. 2017. “How Long Has This Been Going On? A Discussion of ‘Recent Flattening in the Higher Education Wage Premium: Polarization, Skill Downgrading, or Both?’ by Robert G. Valletta.” Massachusetts Institute of Technology.

Autor, David H., David Dorn, and Gordon H. Hanson. 2013. “The China Syndrome: Local Labor Market Effects of Import Competition in the United States.” American Economic Review 103, no. 6, 2121–68.

Autor, David, David Dorn, Lawrence F. Katz, Christina Patterson, and John Van Reenen. 2017a. “Concentrating on the Fall of the Labor Share.” American Economic Review: Papers & Proceedings 107, no. 5: 180–85.

Autor, David, David Dorn, Lawrence F. Katz, Christina Patterson, and John Van Reenen. 2017b. “The Fall of the Labor Share and the Rise of Superstar Firms.” Working Paper. Massachusetts Institute of Technology.

Autor, David, Claudia Goldin, and Lawrence F. Katz. 2020. “Extending the Race Between Education and Technology.” AEA Papers and Proceedings 110: 347–51.

Autor, David H., Lawrence F. Katz, and Melissa S. Kearney. 2006. “The Polarization of the U.S. Labor Market.” American Economic Review 96, no. 2: 189–94.

Autor, David H., Lawrence F. Katz, and Melissa S. Kearney. 2008. “Trends in U.S. Wage Inequality: Revising the Revisionists.” Review of Economics and Statistics 90, no. 2: 300–23.

Autor, David, Alan Manning, and Christopher Smith. 2010. “The Contribution of the Minimum Wage to U.S. Wage Inequality over Three Decades: A Reassessment.” Massachusetts Institute of Technology Working Paper.

Autor, David, Alan Manning, and Christopher Smith. 2016. “The Contribution of the Minimum Wage to U.S. Wage Inequality over Three Decades: A Reassessment.” American Economic Journal: Applied Economics 8, no. 1: 58–99.

Azar, José, Ioana Elena Marinescu, and Marshall Steinbaum. 2017. “Labor Market Concentration.” National Bureau of Economic Research Working Paper no. 24147. https://doi.org/10.3386/w24147.

Azar, José, Ioana Marinescu, and Marshall Steinbaum. 2019. “Measuring Labor Market Power Two Ways.” AEA Papers and Proceedings 109.

Balasubramanian, Natarajan, Jin Woo Chang, Mariko Sakakibara, Jagadeesh Sivadasan, and Evan Starr. 2020. “Locked In? The Enforceability of Covenants Not to Compete and the Careers of High-Tech Workers.” Journal of Human Resources no. 1218-9931R1.

Barkai, Simcha. 2020. “Declining Labor and Capital Shares.” Journal of Finance 75, no. 5: 2421–63.

Bassier, Ihsaan, Arindrajit Dube, and Suresh Naidu. 2020. “Monopsony in Movers: The Elasticity of Labor Supply to Firm Wage Policies.” Working Paper 27755. National Bureau of Economic Research.

Batt, Rosemary, Virginia Doellgast, and Hyunji Kwon. 2006. “Service Management and Employment Systems in U.S. and Indian Call Centers.” In Brookings Trade Forum 2005: Offshoring White-Collar Work—The Issues and Implications, edited by Susan M. Collins and Lael Brainard. Brookings Institution.

Batt, Rosemary, David Holman, and Ursula Holtgrewe. 2009. “The Globalization of Service Work: Comparative Institutional Perspectives on Call Centers.” Industrial & Labor Relations Review 62, no. 4: 453–88.

Batt, Rosemary, and Hiro Nohara. 2009. “How Institutions and Business Strategies Affect Wages: A Cross-National Study of Call Centers.” Industrial & Labor Relations Review 62, no. 4: 533–52.

Beaudry, Paul, David A. Green, and Benjamin M. Sand. 2013. “The Great Reversal in the Demand for Skill and Cognitive Tasks.” Working Paper 18901. National Bureau of Economic Research.

Beaudry, Paul, David A. Green and Benjamin M. Sand. 2014. “The Declining Fortunes of the Young Since 2000.” Papers and Proceedings of the 126th Annual Meeting of the American Economic Association 104, no. 5: 381–6.

Benmelech, Efraim, Nittai Bergman, and Hyunseob Kim. 2018. “Strong Employers and Weak Employees: How Does Employer Concentration Affect Wages?” Working Paper 24307. National Bureau of Economic Research.

Bensman, David. 2009. Stuck on the Low Road: Deregulation Turned Truck Driving from a Good Job to a Bad One. American Prospect, September.

Bergen, Mark, and Josh Eidelson. 2018. “Inside Google’s Shadow Workforce.” Bloomberg News, July 25.

Berlinski, Samuel. 2008. “Wages and Contracting Out: Does the Law of One Price Hold?” British Journal of Industrial Relations 46, no. 1: 59–75.

Bernhardt, Annette, Rosemary Batt, Susan Houseman, and Eileen Appelbaum. 2016. “Domestic Outsourcing in the United States: A Research Agenda to Assess Trends and Effects on Job Quality.” Working Paper 16-253. Upjohn Institute.

Bernhardt, Annette, Ruth Milkman, Nik Theodore, Douglas Heckathorn, Mirabai Auer, James DeFilippis, Ana Luz González, Victor Narro, Jason Perelshteyn, Diana Polson, and Michael Spiller. 2009. Broken Laws, Unprotected Workers: Violations of Employment and Labor Laws in America’s Cities. Center for Urban Economic Development, National Employment Law Project, and UCLA Institute for Research on Labor and Employment.

Bivens, Josh. 2010. Failure by Design: The Story Behind America’s Broken Economy. Ithaca, N.Y.: Cornell Univ. Press.

Bivens, Josh. 2013. “Using Standard Models to Benchmark the Costs of Globalization for American Workers Without a College Degree.” Economic Policy Institute.

Bivens, Josh. 2017a. Adding Insult to Injury: How Bad Policy Decisions Have Amplified Globalization’s Costs for American Workers. Economic Policy Institute.

Bivens, Josh. 2017b. “President Trump’s Alternative Facts Have Foreigners and Bureaucrats, Not the Top 1 Percent, Reaping the Gains from Economic Growth.” Working Economics (Economic Policy Institute blog), January 24.

Bivens, Josh. 2018. “The Fed Shouldn’t Give Up on Restoring Labor’s Share of Income—and Measure It Correctly.” Working Economics (Economic Policy Institute blog), January 30.

Bivens, Josh. 2019. “Predicting Wage Growth with Measures of Labor Market Slack: It’s Complicated.” Working Economics (Economic Policy Institute blog), March 19.

Bivens, Josh, Lora Engdahl, Elise Gould, Teresa Kroeger, Celine McNicholas, Lawrence Mishel, Zane Mokhiber, Heidi Shierholz, Marni von Wilpert, Valerie Wilson, and Ben Zipperer. 2017. How Today’s Unions Help Working People: Giving Workers the Power to Improve Their Jobs and Unrig the Economy. Economic Policy Institute.

Bivens, Josh, Elise Gould, Lawrence Mishel, and Heidi Shierholz. 2014. Raising America’s Pay: Why It’s Our Central Economic Policy Challenge. Economic Policy Institute, June 4.

Bivens, Josh, and Lawrence Mishel. 2013. “The Pay of Corporate Executives and Financial Professionals as Evidence of Rents in Top 1 Percent Incomes.” Journal of Economic Perspectives 27, no. 3: 57–78.

Bivens, Josh, and Lawrence Mishel. 2015. “Understanding the Historic Divergence Between Productivity and a Typical Worker’s Pay: Why It Matters and Why It’s Real.” Economic Policy Institute.

Bivens, Josh, Lawrence Mishel, and John Schmitt. 2018. It’s Not Just Monopoly and Monopsony: How Market Power Has Affected American Wages. Economic Policy Institute.

Bivens, Josh, and Ben Zipperer. 2018. The Importance of Locking in Full Employment for the Long Haul. Economic Policy Institute.

Blanchard, Olivier, and Dani Rodrik, eds. 2021. Combating Inequality: Rethinking Government’s Role. MIT Press.

Block, Sharon, and Benjamin H. Harris, eds. 2021. Inequality and the Labor Market: The Case for Greater Competition. Brookings Institution Press.

Bloom, Nicholas, Faith Guvenen, Benjamin Smith, Jae Song, and Till von Wachter. 2018. “Is the Large Firm Wage Premium Dead or Merely Resting?” AEA Papers and Proceedings 108, no. 1: 317–22.

Bronfenbrenner, Kate. 2009. No Holds Barred: The Intensification of Employer Opposition to Organizing. Economic Policy Institute and the American Rights at Work Education Fund.

Bureau of Labor Statistics (BLS). 2018. “Contingent and Alternative Employment Arrangements, May 2017” (press release). June 7.

Callaci, Brian. 2019. “Fissuring in Flight: Consolidation and Outsourcing in the U.S. Domestic Airline Industry, 1997–2018.” Unpublished.

Callaway, Brantly, and William J. Collins. 2017. “Unions, Workers, and Wages at the Peak of the American Labor Movement.” Working Paper 23516. National Bureau of Economic Research.

Card, David. 1991. “The Effect of Unions on the Distribution of Wages: Redistribution or Relabeling?” Working Paper 287. National Bureau of Economic Research.

Card, David. 1996. “The Effect of Unions on the Structure of Wages: A Longitudinal Analysis.” Econometrica 64, no. 4: 957–79.

Card, David, Ana Rute Cardoso, Joerg Heining, and Patrick Kline. 2018. “Firms and Labor Market Inequality: Evidence and Some Theory.” Journal of Labor Economics 36, no. S1: 13–70.

Card, David, and John DiNardo. 2002. “Skill-Biased Technological Change and Rising Wage Inequality: Some Problems and Puzzles.” Journal of Labor Economics 20, no. 4: 733–83.

Carré, Françoise. 2015. (In)dependent Contractor Misclassification. Economic Policy Institute.

Census Bureau. 2020. “Legacy BDS Firm Characteristics Data Tables 1977–2014” [data table], Longitudinal Business Database [February]. Accessed February 2020.

Collins, Brett, Andrew Garin, Emilie Jackson, Dmitri Koustas, and Mark Paynek. 2019. “Is Gig Work Replacing Traditional Employment? Evidence from Two Decades of Tax Returns.” Internal Revenue Service.

Colvin, Alexander J.S. 2018. The Growing Use of Mandatory Arbitration: Access to the Courts Is Now Barred for More Than 60 Million American Workers. Economic Policy Institute.

Colvin, Alexander J.S., and Heidi Shierholz. 2019. Noncompete Agreements: Ubiquitous, Harmful to Wages and to Competition, and Part of a Growing Trend of Employers Requiring Workers to Sign Away Their Rights. Economic Policy Institute.

Cooper, David. 2019. Raising the Federal Minimum Wage to $15 by 2025 Would Lift Wages for Over 33 Million Workers (fact sheet). Economic Policy Institute, July 17.

Cooper, David, Elise Gould, and Ben Zipperer. 2019. Low-Wage Workers Are Suffering from a Decline in the Real Value of the Federal Minimum Wage. Economic Policy Institute.

Cooper, David, and Teresa Kroeger. 2017. Employers Steal Billions from Workers’ Paychecks Each Year: Survey Data Show Millions of Workers Are Paid Less Than the Minimum Wage, at Significant Cost to Taxpayers and State Economies. Economic Policy Institute.

Costa, Daniel. 2019. Employers Increase Their Profits and Put Downward Pressure on Wages and Labor Standards by Exploiting Migrant Workers. Economic Policy Institute.

Costa, Daniel and Ron Hira. 2020. “EPI Comments on USCIS Preference Allocation System for H-1B Visas by Prevailing Wage Level.” Public comments, U.S. Citizenship and Immigration Services, Department of Homeland Security, Modification of Registration Requirement for Petitioners Seeking to File Cap-Subject H-1B Petitions, Notice of proposed rulemaking, CIS Docket No. 2674-20, DHS Docket No. USCIS-2020-001, RIN: 1615-AC61, December 2.

Costa, Daniel, and Philip Martin. 2019. “OECD Highlights Temporary Labor Migration: Almost as Many Guestworkers as Permanent Immigrants.” Working Economics (Economic Policy Institute blog), December 4.

Costa, Daniel, and Jennifer Rosenbaum. 2017. Temporary Foreign Workers by the Numbers: New Estimates by Visa Classification. Economic Policy Institute.

Council of Economic Advisers. 1995. Economic Report of the President (1995). U.S. Government Printing Office.

Daly, Mary, and Bart Hobijin. 2014. “Downward Nominal Wage Rigidities Bend the Phillips Curve.” Journal of Money, Credit and Banking 46, no. 52.

De Loecker, Jan, Jan Eeckhout, and Gabriel Unger. 2020. “The Rise of Market Power and the Macroeconomic Implications.” Quarterly Journal of Economics 135, no. 2: 561–644. https://doi.org/10.1093/qje/qjz041.

Denice, Patrick, and Jake Rosenfeld. 2018. “Unions and Nonunion Pay in the United States, 1977–2015.” Sociological Science 5: 541–61.

Department of Commerce. Various years. “Franchising and the Economy.”

Department of Labor. No date. History of Changes to the Minimum Wage Law.

Department of Labor (Wage and Hour Division). 2016. Defining and Delimiting the Exemptions for Executive, Administrative, Professional, Outside Sales and Computer Employees, 81 Fed. Reg. 32391–32552 (May 23).

Deutsch, Rachel, Kate Hamaji, Elizabeth Nicolas, Celine McNicholas, Margaret Poydock, and Heidi Shierholz. 2019. Unchecked Corporate Power: Forced Arbitration, the Enforcement Crisis, and How Workers Are Fighting Back. Center for Popular Democracy and Economic Policy Institute.

DiNardo, John, Nicole M. Fortin, and Thomas Lemieux. 1996. “Labor Market Institutions and the Distribution of Wages.” Econometrica 64, no. 5: 1001–44.

Dube, Arindrajit, Laura Giuliano, and Jonathan Leonard. 2019. “Fairness and Frictions: Impact of Unequal Raises on Quit Behavior.” American Economic Review 109, no. 2: 620–63.

Dube, Arindrajit, Jeff Jacobs, Suresh Naidu, and Siddharth Suri. 2020. “Monopsony in Online Labor Markets.” American Economic Review: Insights 2, no. 1: 33–46.

Dube, Arandajit, and Ethan Kaplan. 2010. “Does Outsourcing Reduce Wages in the Low-Wage Service Occupations? Evidence from Janitors and Guards.” Industrial and Labor Relations Review 63, no. 2: 287–306.

Economic Policy Institute. 2012. “Minimum-Wage Impact on 50/10 Wage Gap, 1979–2009” [chart]. The State of Working America. Washington, D.C.: Economic Policy Institute.

Economic Policy Institute. 2019a. “Productivity–Pay Gap.” Last updated July 2019.

Economic Policy Institute. 2019b. “Black–White Wage Gap.” State of Working America Data Library.

Economic Policy Institute. 2019c. State of Working America Data Library, “College Wage Premium.” State of Working America Data Library.

Economic Policy Institute. Various dates. “Wages.” State of Working America Data Library.

Eisenbrey, Ross, and Lawrence Mishel. 2015. “The New Overtime Salary Threshold Would Directly Benefit 13.5 Million Workers: How EPI’s Estimates Differ from the Department of Labor’s.” Economic Policy Institute.

Eisenbrey, Ross, and Lawrence Mishel. 2016. “Explaining the Differences Between EPI and DOL Estimates of Workers Affected by the New Overtime Salary Threshold.” Working Economics (Economic Policy Institute blog), May 20.

Erickson, Christopher L., and Daniel J.B. Mitchell. 2008. “Monopsony as a Metaphor for the Emerging Post-Union Labour Market.” International Labour Review 146, no. 3–4 (January): 163–87. https://doi.org/10.1111/j.1564-913X.2007.00012.x.

Farber, Henry S., Daniel Herbst, Ilyana Kuziemko, and Suresh Naidu. 2021. “Unions and Inequality over the Twentieth Century: New Evidence from Survey Data.” Working Paper 24587. May 2018, revised April 2021. National Bureau of Economic Research.

Farhi, Emmanuel, and François Gourio. 2018. “Accounting for Macro-Finance Trends: Market Power, Intangibles, and Risk Premia.” Brookings Papers on Economic Activity (Fall): 147–250.

Federal Reserve Bank of New York. 2020. The Labor Market for Recent College Graduates (interactive web feature). Last updated February 12, 2020.

Fortin, Nicole M., and Thomas Lemieux. 1997. “Institutional Changes and Rising Wage Inequality: Is There a Linkage?” Journal of Economic Perspectives 11, no. 2: 75–96.

Fortin, Nicole M., Thomas Lemieux, and Neil Lloyd. 2021. “Labor Market Institutions and the Distribution of Wages: The Role of Spillover Effects.” Working Paper 28375. National Bureau of Economic Research.

Freeman, Richard. 1991. “How Much Has De-Unionization Contributed to the Rise in Male Wage Inequality?” Working Paper 3826. National Bureau of Economic Research.

Freeman, Richard B., and James L. Medoff. 1984. What Do Unions Do? Basic Books.

Galbraith, James. 1997. “Time to Ditch the NAIRU.” Journal of Economic Perspectives 11, no. 1: 93–108.

Galbraith, James K. 1998. Created Unequal: The Crisis in American Pay. Free Press.

Goldin, Claudia, and Lawrence F. Katz. 2007. “The Race Between Education and Technology: The Evolution of U.S. Educational Wage Differentials, 1890 to 2005.” Working Paper 12984. National Bureau of Economic Research.

Goldin, Claudia, and Lawrence F. Katz. 2008. The Race Between Education and Technology. Harvard Univ. Press.

Goldschmidt, Deborah, and Johannes Schmieder. 2017. “The Rise of Domestic Outsourcing and the Evolution of the German Wage Structure.” Quarterly Journal of Economics 132, no. 3: 1165–217.

Goos, Maarten, Alan Manning, and Anna Salomons. 2014. “Explaining Job Polarization: Routine-Biased Technological Change and Offshoring.” American Economic Review 104, no. 8: 2509–26.

Gould, Elise. 2018. The State of American Wages 2017: Wages Have Finally Recovered from the Blow of the Great Recession but Are Still Growing Too Slowly and Unequally. Economic Policy Institute.

Gould, Elise. 2020. State of Working America Wages 2019: A Story of Slow, Uneven, and Unequal Wage Growth over the Last 40 Years. Economic Policy Institute.

Greene, Jay. 2021. “Amazon Shortchanged Drivers $61.7 Million in Tips. Now It Has Settled with the FTC to Pay Up.” Washington Post, February 2.

Greenhouse, Steven. 2020. “02: The Future of Work.” The World as You’ll Know It (podcast transcript), 32:38.

Greenwald, Daniel L., Martin Lettau, and Sydney C. Ludvigson. 2019. “How the Wealth Was Won: Factors Shares as Market Fundamentals.” Working Paper 25769. National Bureau of Economic Research.

Groshen, Erica L., and Harry J. Holzer. 2019. “Improving Employment and Earnings in Twenty-First Century Labor Markets.” Russell Sage Foundation Journal of the Social Sciences 5, no. 5.

Grullon, Gustavo, Yelena Larkin, and Roni Michaely. 2015. “Are U.S. Industries Becoming Concentrated?” Swiss Finance Institute Research Paper 19-41, forthcoming in Review of Finance, posted online May 31, 2015.

Hacker, Jacob S., and Paul Pierson. 2011. Winner-Take-All Politics: How Washington Made the Rich Richer—and Turned Its Back on the Middle Class. Simon & Schuster.

Hacker, Jacob S., and Paul Pierson. 2020. Let Them Eat Tweets: How the Right Rules in an Age of Extreme Inequality. Liveright Publishing.

Hooper, Peter, Frederic S. Mishkin, and Amir Sufi. 2019. “Prospects for Inflation in a High Pressure Economy: Is the Phillips Curve Dead or Is It Just Hibernating?” Working Paper 25792. National Bureau of Economic Research.

Howell, David R. 2001. “The Skills Myth.” American Prospect, December 19, 2001.

Howell, David R., and Arne L. Kalleberg. 2019. “Declining Job Quality in the United States: Explanations and Evidence.” Russell Sage Foundation Journal of the Social Sciences 5, no. 4: 1–53.

Howell, David R., and Susan S. Wieler. 1998. “Skill-Biased Demand Shifts and Wage Collapse in the United States: A Critical Perspective.” Eastern Economic Journal 24, no. 3: 343–66.

Hunt, Jennifer, and Ryan Nunn. 2019. “Is Employment Polarization Informative About Wage Inequality and Is Employment Really Polarizing?” Working Paper 26064. National Bureau of Economic Research.

Johnson, Matthew S., Kurt Lavetti, and Michael Lipsitz. 2020. “The Labor Market Effects of Legal Restrictions on Worker Mobility.” Federal Trade Commission, Bureau of Economics.

Katz, Lawrence, and Alan Krueger. 1999. “The High-Pressure Labor Market of the 1990s.” Brookings Papers on Economic Activity 1.

Katz, Lawrence F., and Kevin M. Murphy. 1992. “Changes in Relative Wages, 1963–87: Supply and Demand Factors.” Quarterly Journal of Economics 107 (February): 35–78.

Kearney, Melissa S., Brad Hershbein, and David Boddy. 2015. “The Future of Work in the Age of the Machine.” Brookings Institution: Hamilton Project.

Kimball, Will, and Lawrence Mishel. 2016. Nearly One-Third of Salaried Workers Will Be Guaranteed Overtime Under the New Rule, but That Falls Short of the Half Covered in the Mid-1970s. Economic Snapshot. Economic Policy Institute, June.

Kochan, Thomas A., and Will Kimball. 2019. “Unions, Worker Voice, and Management Practices: Implications for a High-Productivity, High-Wage Economy.” Russell Sage Foundation Journal of the Social Sciences 5, no. 5.

Kochan, Thomas A., Will Kimball, Duanyi Yang, and Erin L. Kelly. 2018. “Worker Voice in America: Is There a Gap Between What Workers Expect and What They Get?” ILR Review 72, no. 1: 3–38. https://doi.org/10.1177/0019793918806250.

Kopczuk, Wojciech, Emmanuel Saez, and Jae Song. 2007. “Uncovering the American Dream: Inequality and Mobility in Social Security Earnings Data Since 1937.” National Bureau of Economic Research Working Paper no. 13345, August 2007.

Kosová, Renáta, and Francine Lafontaine. 2012. “Much Ado About Chains: A Research Agenda.” International Journal of Industrial Organization 30: 303–8.

Krogstad, Jens Manuel, Jeffrey Passel, and D’vera Cohn. 2019. “5 Facts About Illegal Immigration in the U.S.” Pew Research Center.

Krueger, Alan B. 2017. “The Rigged Labor Market.” Milken Institute Review, April 28.

Krueger, Alan B. 2018. “Reflections on Dwindling Worker Bargaining Power and Monetary Policy.” Luncheon Address at the Jackson Hole Economic Symposium, August 24.

Krueger, Alan B., and Orley Ashenfelter. 2018. “Theory and Evidence on Employer Collusion in the Franchise Sector.” Working Paper 24831. National Bureau of Economic Research.

Lafontaine, Francine, and Roger D. Blair. 2009. “The Evolution of Franchising and Franchise Contracts: Evidence from the United States.” Entrepreneurial Business Law Journal 3, no. 2: 381–434.

Langella, Monica, and Alan Manning. 2020. “The Measure of Monopsony.” Video, 56:28. Marshall Lecture, Annual Congress of the European Economic Association.

Lazear, Edward P. 2019. “Mind the Productivity Gap to Reduce Inequality: It Isn’t Only an American Problem, but the U.S. Has Lessons to Learn from Other Wealthy Countries.” Wall Street Journal, May 6.

Lettieri, John W. 2020. A Better Bargain: How Noncompete Reform Can Benefit Workers and Boost Economic Dynamism. American Enterprise Institute.

Levy, Frank, and Peter Temin. 2007. “Inequality and Institutions in 20th Century America.” Working Paper 13106. National Bureau of Economic Research.

Lipsitz, Michael, and Evan Starr. 2020. “Low-Wage Workers and the Enforceability of Non-Compete Agreements.” Forthcoming in Management Science. Online version October 20, 2020.

Lipsky, David B. 2007. “Conflict Resolution and the Transformation of the Social Contract.” Labor and Employment Relations Association Presidential Address, January.

Locke, Mandy, and Franco Ordonez. 2014. “Taxpayers and Workers Gouged by Labor-Law Dodge.” McClatchy Washington Bureau, September 4.

Lopez, German. 2019. “I Was Skeptical of Unions. Then I Joined One.” Vox, August 19.

Maestas, Nicole, Kathleen J. Mullen, David Powell, Till von Wachter, and Jeffrey B. Wenger. 2017. Working Conditions in the United States: Results of the 2015 American Working Conditions Survey. RAND Corporation.

Manduca, Robert. 2018. “Income Inequality and the Persistence of Racial Economic Disparities.” Sociological Science 5: 182–205.

McAdams, John M. 2019. “Non-Compete Agreements: A Review of the Literature.” Federal Trade Commission, Bureau of Economics. http://dx.doi.org/10.2139/ssrn.3513639.

McNicholas, Celine, Zane Mokhiber, and Adam Chaikof. 2017. Two Billion Dollars in Stolen Wages Were Recovered for Workers in 2015 and 2016—and That’s Just a Drop in the Bucket. Economic Policy Institute.

McNicholas, Celine, Margaret Poydock, Julia Wolfe, Ben Zipperer, Gordon Lafer, and Lola Loustaunau. 2019. Unlawful: U.S. Employers Are Charged with Violating Federal Law in 41.5% of All Union Election Campaigns. Economic Policy Institute.

Mishel, Lawrence. 2014. “Low-Wage Workers Have Far More Education Than They Did in 1968, Yet They Make Far Less.” Economic Snapshot. Economic Policy Institute, January 23.

Mishel, Lawrence. 2018. Yes, Manufacturing Still Provides a Pay Advantage, but Staffing Firm Outsourcing Is Eroding It. Economic Policy Institute, March.

Mishel, Lawrence. 2019. “Private Sector Union Density Dynamics, 1951–2017.” Presentation at Columbia University, September 12.

Mishel, Lawrence, and Jared Bernstein. 1994. “Is the Technology Black Box Empty? An Empirical Examination of the Impact of Technology on Wage Inequality and the Employment Structure.” Technical Paper. Economic Policy Institute.

Mishel, Lawrence, and Jared Bernstein. 1998. “Technology and the Wage Structure: Has Technology’s Impact Accelerated Since the 1970s?” Research in Labor Economics 17: 305–56.

Mishel, Lawrence, Jared Bernstein, and John Schmitt. 1997a. The State of Working America 1996–1997.  An Economic Policy Institute book. M.E. Sharpe.

Mishel, Lawrence, Jared Bernstein, and John Schmitt. 1997b. “Did Technology Have Any Effect on the Growth of Wage Inequality in the 1980s and 1990s?” Economic Policy Institute.

Mishel, Lawrence, and Josh Bivens. 2013. The Pay of Corporate Executives and Financial Professions as Evidence of Rents in Top 1 Percent Incomes. Economic Policy Institute.

Mishel, Lawrence, and Josh Bivens. 2017. “The Zombie Robot Argument Lurches On: There Is No Evidence That Automation Leads to Joblessness or Inequality.” Economic Policy Institute.

Mishel, Lawrence, Josh Bivens, Elise Gould, and Heidi Shierholz. 2012. The State of Working America, 12th Edition. An Economic Policy Institute book. Cornell Univ. Press.

Mishel, Lawrence, and Jori Kandra. 2020. “Wages for the Top 1% Skyrocketed 160% Since 1979 While the Share of Wages for the Bottom 90% Shrunk: Time to Remake Wage Pattern with Economic Policies That Generate Robust Wage-Growth for Vast Majority.” Working Economics (Economic Policy Institute blog), December.

Mishel, Lawrence, and Melat Kassa. 2019. “Top 1.0 Percent of Earners See Wages Up 157.8% Since 1979.” Working Economics (Economic Policy Institute blog), December 18.

Mishel, Lawrence, Lynn Rhinehart, and Lane Windham. 2020. “Explaining the Erosion of Private-Sector Unions: How Corporate Practices and Legal Changes Have Undercut the Ability of Workers to Organize and Bargain.” Economic Policy Institute.

Mishel, Lawrence, John Schmitt, and Heidi Shierholz. 2014. “Wage Inequality: A Story of Policy Choices.” SAGE Journals, August 4.

Mishel, Lawrence, and Heidi Shierholz. 2017. “Robots, or Automation, Are Not the Problem: Too Little Worker Power Is.” Economic Policy Institute.

Mishel, Lawrence, Heidi Shierholz, and John Schmitt. 2013. “Don’t Blame the Robots: Assessing the Job Polarization Explanation of Growing Wage Inequality.” Economic Policy Institute.

Mishel, Lawrence, and Julia Wolfe. 2018. “Top 1.0 Percent Reaches Highest Wages Ever—Up 157 Percent Since 1979.” Working Economics (Economic Policy Institute blog), October 18.

Mitchell, Daniel J.B., and C L. Erickson. 2005. “Not Yet Dead at the Fed: Unions, Worker Bargaining, and Economy-Wide Wage Determination.” Industrial Relations, 44, no. 3: 565–606.

New York Times Editorial Board. 2018. “The Trouble with Trucking.” New York Times, August 11.

Naidu, Suresh, Eric A. Posner, and E. Glen Weyl. 2018. “Antitrust Remedies for Labor Market Power.” Coase-Sandor Working Paper Series in Law and Economics, no. 850.

Naidu, Suresh, and Aaron Sojourner. 2020. Employer Power and Employee Skills: Understanding Workforce Training Programs in the Context of Labor Market Power. Roosevelt Institute, December.

Organisation for Economic Co-operation and Development (OECD). 2019a. International Migration Outlook 2019.

Organisation for Economic Co-operation and Development (OECD). 2019b. Negotiating Our Way Up: Collective Bargaining in a Changing World of Work.

Ormiston, Russell, Dale Belman, and Mark Erlich. 2020. An Empirical Methodology to Estimate the Incidence and Costs of Payroll Fraud in the Construction Industry. National Alliance for Fair Contracting.

Ortega, Francesc, and Amy Hsin. 2018. “Occupational Barriers and the Labor Market Penalty from Lack of Legal Status.” IZA Institute of Labor Economics No. 11680.

Pierce, Brooks. 1999. “Compensation Inequality.” Working Paper 323. Department of Labor, Bureau of Labor Statistics.

Piketty, Thomas, and Emmanuel Saez. 2003. “Income Inequality in the United States, 1913–1998.” Quarterly Journal of Economics 117, no. 1. (Longer updated version published in Top Incomes over the Twentieth Century, A.B. Atkinson and T. Piketty, eds., Oxford Univ. Press, 2007; tables and figures updated to 2017 in Excel format, March 2019.)

Posner, Eric A., Glen Weyl, and Suresh Naidu. 2018. “Antitrust Remedies for Labor Market Power.” Harvard Law Review 132, no. 2: 535–601.

Prescott, J.J., Norman D. Bishara, and Evan Starr. 2016. “Understanding Noncompetition Agreements: The 2014 Noncompete Survey Project.” Michigan State Law Review, no. 2: 369–464.

Rinz, Kevin. 2018. “Labor Market Concentration, Earnings Inequality, and Earnings Mobility.” Working Paper 2018-10. U.S. Center for Administrative Records Research and Applications, Census Bureau.

Rosenfeld, Jake. 2014. What Unions No Longer Do. Harvard Univ. Press.

Rosenfeld, Jake. 2021. You’re Paid What You’re Worth and Other Myths of the Modern Economy. Belknap Press.

Rosenfeld, Jake, Patrick Denice, and Jennifer Laird. 2016. Union Decline Lowers Wages of Nonunion Workers: The Overlooked Reason Why Wages Are Stuck and Inequality Is Growing. Economic Policy Institute.

Schmitt, John, and Alexandra Mitukiewicz. 2012. “Politics Matter: Changes in Unionization Rates in Rich Countries, 1960–2010.” Industrial Relations Journal 43, no. 3: 260–80.

Shambaugh, Jay, Ryan Nunn, Patrick Liu, and Greg Nantz. 2017. Thirteen Facts About Wage Growth. Economic Facts. Brookings Institution: Hamilton Project.

Shierholz, Heidi. 2019. “More Than Eight Million Workers Will Be Left Behind by the Trump Overtime Rule: Workers Would Receive $1.4 Billion Less Than Under the 2016 Rule.” Working Economics (Economic Policy Institute blog), September 25.

Shierholz, Heidi. 2021. “The Importance of Enforcement in Designing Effective Labor Market Tools.” In Combating Inequality: Rethinking Government’s Role, edited by Olivier Blanchard and Dani Rodrik. MIT Press.

Smith, Matt, Jennifer Gollan, and Adithya Sambamurthy. 2014. “Job Brokers Steal Wages, Entrap Indian Tech Workers in U.S.” Reveal News, October 27.

Sokolova, Anna, and Todd Sorensen. 2020. “Monopsony in Labor Markets: A Meta-Analysis.” Washington Center for Equitable Growth.

Staiger, Douglas, James H. Stock, and Mark W. Watson. 1997. “The NAIRU, Unemployment and Monetary Policy.” Journal of Economic Perspectives 11, no. 1: 33–49.

Stansbury, Anna M., and Lawrence H. Summers. 2020. “Declining Worker Power and American Economic Performance.” Brooking Papers on Economic Activity: BPEA Conference Drafts.

Starr, Evan. 2019a. “Consider This: Training, Wages, and the Enforceability of Covenants Not to Compete.” ILR Review 72, no. 4: 783–817. https://doi.org/10.1177/0019793919826060.

Starr, Evan. 2019b. The Use, Abuse, and Enforceability of Non-Compete and No-Poach Agreements: A Brief Review of the Theory, Evidence, and Recent Reform Efforts. Economic Innovation Group.

Starr, Evan. 2020. Non-Competes in the Workplace: Examining Antitrust and Consumer Protections Issues. (Testimony to the Federal Trade Commission). Federal Trade Commission: Panel 2. January 9.

Starr, Evan, J.J. Prescott, and Norman D. Bishara. 2019. “The Behavioral Effects of (Unenforceable) Contracts.” Forthcoming in Journal of Law, Economics, and Organization, online version September 18, 2020.

Starr, Evan, J.J. Prescott, and Norman D. Bishara. 2020. “Noncompete Agreements in the U.S. Labor Force.” Forthcoming in Journal of Law and Economics, online version May 7, 2020.

Stiglitz, Joseph E. 2012. The Price of Inequality: How Today’s Divided Society Endangers Our Future. W. W. Norton.

Stiglitz, Joseph E. 2021. “Fostering More-Competitive Labor Markets.” In Inequality and the Labor Market: The Case for Greater Competition, edited by Sharon Block and Benjamin Harris. Brookings Institution Press.

Stockman, Farah. 2013. “Teacher Trafficking: The Strange Saga of Filipino Workers, American Schools, and H-1B Visas.” Boston Globe, June 12.

Stone, Katherine V.W., and Alexander J.S. Colvin. 2015. The Arbitration Epidemic: Mandatory Arbitration Deprives Workers and Consumers of Their Rights. Economic Policy Institute.

Stuyven, Dann. 2018. “Superstar Firms, Super Profits, and Shrinking Wage Shares.” U.S. Economics Analyst, Goldman Sachs Economics Research Note, February 4 (not publicly available).

Summers, Lawrence. 2015. “The Future of Work in the Age of the Machine: A Hamilton Project Policy Forum.” Brookings Institution: Hamilton Project, February 19.

Uber Technologies (Uber). 2019. “Form S-1 Registration Statement, United States Securities and Exchange Commission.” April 11.

USA Today. 2018. “Read Transcript of Former President Obama’s Speech, Blasting President Trump.” September 7.

USA Today Editorial Board. 2017. “Rigged System Rips Off Port Truckers: Modern-Day Indentured Servitude Thrives in Southern California.” June 20.

U.S. Government Accountability Office (U.S. GAO). 1989. “Tax Administration: Information Returns Can Be Used to Identify Employers Who Misclassify Workers.” GAO/GGD-89-107.

U.S. Government Accountability Office (U.S. GAO). 2009. “Employee Misclassification: Improved Coordination, Outreach, and Targeting Could Better Ensure Detection and Prevention.”

Valletta, Robert G. 2017. “Recent Flattening in the Higher Education Wage Premium: Polarization, Skill Downgrading, or Both?” In Education, Skills, and Technical Change: Implications for Future U.S. GDP Growth. National Bureau of Economic Research.

Viscelli, Steve. 2016. The Big Rig: Trucking and the Decline of the American Dream. Univ. of California Press.

Webber, Douglas A. 2015. “Firm Market Power and the Earnings Distribution” Labour Economics 35: 123–134.

Webber, Douglas A. 2018. “Employment Adjustment Over the Business Cycle: The Impact of Competition in the Labor Market.” IZA Discussion Papers 11887, Institute of Labor Economics (IZA).

Webber, Douglas A. 2020. “Labor Market Competition and Employment Adjustment over the Business Cycle.” Journal of Human Resources 56, no. 2.

Weil, David. 2014. The Fissured Workplace: Why Work Became So Bad for So Many and What Can Be Done About It. Harvard Univ. Press.

Weil, David. 2017. “Lots of Employees Get Misclassified as Contractors. Here’s Why It Matters.” Harvard Business Review, July 5.

Weil, David. 2019. “ The Present and Future of Work in the Fissured Workplace Context.” Russell Sage Foundation Journal of the Social Sciences 5, no. 5.

Western, Bruce, and Jake Rosenfeld. 2011. “Unions, Norms, and the Rise in U.S. Wage Inequality.” American Sociological Review 76, no. 4: 513–37.

Wilmers, Nathan. 2018. “Wage Stagnation and Buyer Power: How Buyer-Supplier Relations Affect U.S. Workers’ Wages, 1978 to 2014.” American Sociological Review, 1–28. https://doi.org/10.1177/0003122418762441.

Wilson, Valerie. 2015. The Impact of Full Employment on African American Employment and Wages. Full Employment Project at the Center on Budget and Policy Priorities.

Wilson, Valerie. 2020. “Racism and the Economy: Focus on Employment.” Working Economics (Economic Policy Institute blog), November 21.

Wilson, Valerie, and William M. Rodgers III. 2016. Black–White Wage Gaps Expand with Rising Wage Inequality. Economic Policy Institute.

Windham, Lane. 2017. Knocking on Labor’s Door: Union Organizing in the 1970s and the Roots of a New Economic Divide. Univ. of North Carolina Press.

Woodman, Spencer. 2015. “Exclusive: Amazon Makes Even Temporary Warehouse Workers Sign 18-Month Non-Competes.” The Verge, March 26.

A small but dedicated group of economists, legal theorists, and political thinkers has promoted the argument that little if any labor market regulation is required to ensure the proper level of protection for occupational safety and health (OSH), because workers are fully compensated by higher wages for the risks they face on the job and that markets alone are sufficient to ensure this outcome. In this paper, we argue that such a sanguine perspective is at odds with the history of OSH regulation and the most plausible theories of how labor markets and employment relations actually function. In particular, the imbalance of power between employers and employees and lack of information alone prevent employees from obtaining safe working conditions in market transactions with employers. We also find that the empirical claims purportedly buttressing the free-market view do not stand up to scrutiny, and that the profound shortcomings of OSH performance in the United States, brought to vivid light by the current pandemic, are attributable to too little public intervention in labor markets, not too much.

In the English-speaking world, OSH regulation dates to the Middle Ages. Modern policy frameworks, such as the Occupational Safety and Health Act in the United States, are based on the presumption of employer responsibility, which in turn rests on the recognition that employers generally hold a preponderance of power vis-à-vis their workforce such that public intervention serves a countervailing purpose. Arrayed against this presumption, however, has been the classical liberal view that worker and employer self-interest, embodied in mutually agreed employment contracts, is a sufficient basis for setting wages and working conditions and ought not be overridden by public action—a position we dub the “freedom of contract” view. This position broadly corresponds to the Lochner-era stance of the U.S. Supreme Court and today characterizes a group of economists, led by W. Kip Viscusi, associated with the value-of-statistical-life (VSL) literature.

The theoretical basis for the VSL approach assumes a full-employment labor market in which no worker would willingly exchange their job for another they would be qualified to fill. This situation has little similarity to real world labor markets, which are typically in a state of excess supply (unemployment), where compensation for equally qualified workers varies substantially by industry and employer, and where outcomes are determined not only by initial contracts but also worker and employer performance over time. OSH is one dimension of employer performance.

The problem of ensuring safe and healthy work is not just theoretical. Although workplaces are much less dangerous now than they were 100 years ago, more than 5,000 people died from work-related injuries in the U.S. in 2018. The U.S. Department of Labor’s Bureau of Labor Statistics (BLS) reports that about 3.5 million people sustained injuries at work in that year. However, studies have shown that the BLS substantially underestimates injury incidence, and that the actual number is most likely in the range of 5-10 million. The vast majority of occupational diseases, including cancer, lung diseases, and coronary heart disease, go unreported. A credible estimate, even before the Covid-19 pandemic, is that 26,000 to 72,000 people die annually from occupational diseases.

Many high-risk workers are also poorly paid. For example, the 1.5 million nursing assistants, whose injury rates are over twice the workforce average, have median annual pay that barely exceeds the poverty level for a family of four, and the 2.16 million janitors and cleaners have median annual pay below the poverty level. In addition, temporary and short-term employment is associated with elevated injury rates, even within the same occupation.

There is also growing evidence that workers of color, particularly Black workers, have elevated injury risk because they are overrepresented in relatively hazardous occupations. This is the case even for workers of the same age, education, and sex as their white counterparts. Structural racism is a likely cause of these disparities.

When workers are injured, they incur substantial financial and nonfinancial burdens. Injuries involving a week or more off work lead to average lost earnings of around 20% over at least 10 years following injury. Only between one-half and three-fourths of injured workers eligible for workers’ compensation cash benefits actually receive them. For those who do, less than a fifth of lost wages are replaced.

Many families do not have adequate liquid savings or access to credit. As a result, financial losses caused by occupational injuries and illnesses can lead to difficulty covering everyday expenses, including rent or mortgage payments, car payments, and even food. Latinx and Black workers are much less likely than white workers to have adequate savings to cover these expenses.

Injury impacts stretch beyond the financial. Injured workers experience limitations in household roles, including those as spouse and parent. They suffer post-injury depression and opioid use disorder. They also die at higher rates than their fellow employees from causes including suicide and drug overdose.

The United States stands poorly in international comparisons of work-related fatal injury rates. The U.S. rate is 10% higher than that of its closest rival, Japan, and six times the rate of Great Britain. This difference cannot be explained by differences in industry mix: The U.S. rate for construction is 20% higher, the manufacturing rate 50% higher, and the transportation and storage rate 100% higher than that of the E.U.

This is the real world context of statistical studies claiming to find wages compensating for voluntarily accepted OSH risk. Following Viscusi, such researchers employ regression models in which a worker’s wage, typically its natural logarithm, is a function of the worker’s demographic characteristics (age, education, experience, marital status, gender) and the risk of occupational fatality they face. Using census or similar surveys for nonrisk variables and average fatal accident rates by industry and occupation for risk, these researchers estimate the effect of the risk variable on wages, which they interpret as the money workers are willing to accept in return for a unit increase in risk. This exercise provides the basis for VSL calculations, and it is also used to argue that OSH regulation is unnecessary since workers are already compensated for differences in risk.

This methodology is highly unreliable, however, for a number of reasons:

• It rests on assumptions regarding risk perception and behavior at variance with most research.
• It assumes workers are accurately informed about the risks they face by employers who have an incentive to conceal them.
• It ignores extensive existing regulation of OSH risk and the effects this regulation has on choices made by workers and employers.
• The risk variable, on which the entire empirical strategy depends, is mismeasured: It applies large-group averages to the individual level, and it excludes fatal diseases, which are far more numerous than fatal accidents.

• It suffers from omitted variable bias; in particular, omission of employer- and industry-level variables likely results in very large overestimation of compensation for risk.
• It fails to adequately incorporate the different labor market experiences of men versus women, white workers versus workers of color, and other relevant distinctions.

Given these issues, it is striking that hazardous working conditions are the only job characteristic for which there is a literature claiming to find wage compensation.

While the findings of individual studies have little credibility, the pattern of results across different samples might convey useful information. If so, we note that greater compensation has been found for white, native-born, unionized, and higher-paid workers than for their Black, immigrant, nonunion, and lesser-paid counterparts—suggestive of the role of social and economic power differentials.

The shortcomings of the U.S. OSH system have been laid bare by the novel coronavirus pandemic. Workplaces in meatpacking and other industries have functioned as superspreading venues, and broader surveys reveal widespread worker concern over exposure. Key underlying factors include limited employer protection efforts, the failure of OSHA to assume regulatory responsibility, the inadequacy of occupational disease surveillance, the absence of a social safety net that would permit lower-paid workers to take time off from work if they test positive or have been exposed to identified cases, insufficient coverage by the workers’ compensation system, the lack of hazard pay for workers at greatest risk, and the unequal distribution of risk along lines of race and class. The policy measures needed to protect workers during the pandemic reflect a larger agenda that preceded this episode and will remain after it is over: Raising the bottom end of the labor market to acceptable levels of wages and productivity; revamping labor laws to encourage unionization; rebuilding the occupational and public health infrastructure; taking assertive action to combat discrimination on the basis of race, ethnicity, and immigrant status; and embarking on initiatives to enhance worker voice in the 21st-century economy.

II. Introduction

By early summer of 2020 wildcat strikes or other job actions were underway at over 900 workplaces across the United States, largely over issues of health and safety related to the novel coronavirus (Payday Report 2020). Workers were demanding timely information, protective equipment, and changes in work arrangements to reduce their risk of infection and resorting to collective action to bring the changes about. As the confrontational nature of these events makes clear, employers did not always go along.

How should we understand such conflicts over health and safety at work? What do they tell us more generally about fairness, efficiency, and the role of inequality and power in employment? This paper examines these questions by reviewing a longstanding debate about occupational safety and health (OSH): Is it an obligation that ought to be assumed by employers, or do competitive markets under a presumption of equal power between employers and employees ensure safe conditions and adequate compensation for any risks that would be too costly to eliminate? As we will see, in modern economics our answer to this question often comes down to whether or not we believe workers are fully compensated for the risks they encounter at work, compensation they might reasonably demand if they could negotiate with employers from a position of approximate equality.

In this paper we will argue:

• The “freedom of contract” view of occupational safety and health fails both theoretically and econometrically. It is based on an inherently implausible view of labor markets and the employment relationship, and it is not supported by the empirical evidence.
• The inadequacy of the free market view is not a secondary matter. Rather, the failure of unregulated markets to ensure adequately safe working conditions and health equity for the workforce has urgent health as well as economic consequences, including those now visible as a result of the ongoing pandemic.

We will begin by briefly reviewing the history of OSH regulation in the English-speaking world, highlighting the changing assumptions on which regulation has been based. This will lead to an examination of modern theoretical perspectives on the employment relationship and its health and safety aspects, and we will contrast the anti-regulatory view held by some economists with alternatives that recognize power imbalances between workers and employers. In the following section we will document the extent and distribution of occupational injury and disease, demonstrating that their impact reflects and exacerbates economic and social inequality. It will be clear from our review of the evidence that, even in a technologically advanced economy like that of the United States, the toll of workplace accidents, illnesses, and deaths remains unacceptably high. After this we will turn to the “value of statistical life” (VSL) literature, which supposedly documents substantial wage compensation for risk, and expose problems in theory, measurement, and methodology. Moreover, we will show that, even on its own terms, this literature presents evidence for the importance of centering workplace inequalities when assessing compensation for risk: More-vulnerable workers (i.e., low-wage, Black, and/or immigrant) not only face greater risk on the job but also receive less compensation for it. In the final section we will return to a discussion of health and safety at work in the age of Covid-19, demonstrating how our analysis is applicable to this extraordinary episode.

One clarification about terminology: In this paper we will use the term freedom of contract to refer to the labor market perspective underlying the presumption that OSH risks are fairly and efficiently negotiated between workers and employers on an equal footing. Strictly speaking, this usage is not quite precise, since freedom of contract is a concept drawn from legal theory and is intrinsically normative, not necessarily descriptive of any particular labor market setting. In practice, however, there is a close correspondence between the legal and economic perspectives. Advocates of freedom of contract generally make the following assumptions about how labor markets work: (1) labor markets are in a more or less continuous equilibrium in which labor supply equals labor demand, (2) all relevant aspects of employment are determined in the initial moment of contractual agreement, (3) there is no regulation of the legal rights and obligations of the employment relationship on either side, (4) social norms and hierarchies are either nonexistent or can be regarded as fixed and are therefore ignorable background influences on individual choice, and (5) in the absence of regulation (or possibly excessive market concentration) the concept of power is inapplicable. In addition, in the context of OSH it is further assumed that risk to life and health is an ordinary consumption good, evaluated on a single dimension (from less to more) and without regard to additional social or psychological considerations. There is no generally accepted term that represents this constellation of assumptions, but, since certain strands of legal and economic thinking align so closely, we have chosen to use the legal designation. Of course, freedom of contract as an idealization applies to the entire range of economic outcomes in which markets play or could play a role, not just the labor outcomes we consider here (Atiyah 1979).

III. Two views of the risk of injury and illness at work

Under English common law dating from the Middle Ages, masters were deemed to be responsible for ensuring that work was acceptably safe for their servants (Henshaw et al. 2007). The underlying philosophy was paternalism, based on the view that masters were more knowledgeable, commanded more resources, and assumed reciprocal obligations because they benefited from their servants’ devotion. Acceptability meant that risks should be no more than typical of such employments, so if disputes made their way to court the relevant evidence consisted of comparisons between the working conditions of this particular master and his peers. Indeed, paternalistic entailments limited the scope of contractual freedom over a wide swath of social and economic life (Atiyah 1979).

Gathering force in the 17th and 18th centuries, however, was a new doctrine, freedom of contract. Under this view, parties are assumed to act in their best interest in the marketplace or, if they don’t, they must be made to suffer the consequences so they will learn to do so in the future. Thus, neither party is to be regarded as having obligations toward the other except insofar as they have agreed to them, and both sides to a contract are free to agree to any terms they choose. The only role of the state is to enforce the terms of such contracts on parties that resist carrying them out. Applied to matters of health and safety at work, this view entailed rejection of the paternalistic view that masters (employers) had inherent obligations to protect servants (workers), relying instead on the judgment both should exercise in pursuing their separate interests. Thus was born the notion of occupational risk as the outcome of a mutual, and presumably optimal, employment agreement (Atiyah 1979).

It is important to recognize that these new terms do not describe an evolution in which one view of OSH was progressively supplanted by another. On the contrary, with changes in the economic role of business, repeated revolutions in technology, and the emerging centrality of the employment relationship in modern life, a version of the older employer responsibility approach to OSH took on new life and became, in one form or another, the dominant understanding by the end of the 19th century (Henshaw et al. 2007). Rather than resting on the paternalism of medieval caste hierarchies, however, the modern view had these elements:

• It recognized that technology had become far more complex and opaque, with serious new dangers best regulated at their source. This meant that businesses had to take worker health and safety into account when devising new methods and products.
• It increasingly viewed the workplace as a sort of society, with workers at least partly in the role of citizens. Thus, the exercise of care by the employer was demanded because it reflected the reciprocity and mutual concern on which industrial cooperation ought to be based.
• It acknowledged the disparity of power between business owners and workers, especially in times of excess labor supply. Indeed, the worst-off workers, those with the least education and fewest outside options aside from their current job, were visibly subjected to barbarous conditions in the absence of social protection (Mill 1871).

For these reasons, the employer responsibility view of OSH freed itself from the ancient noblesse oblige paternalism of the law of masters and servants.

Employer responsibility was codified in the British Factory Acts beginning in 1833, and it crossed the Atlantic after the American Civil War. Massachusetts became the first state to regulate working conditions in 1877, and in the decades that followed nearly every other state followed suit (MacLaury, undated). This is not to say that regulation was sufficient; horrendous conditions could be found in mining, railroads, and meatpacking, as well as manufacturing. The political debate, however, was not over whether some measure of regulation was necessary but where and how much.

Workers often took the initiative to demand responsibility from their employers when the government failed to take action. Along with wages, the call for greater safety on the job was a prime motivator for the unionization drives of the late 19th and early 20th centuries. Eventually labor movement pressure resulted in the Occupational Safety and Health Act of 1970, which created the Occupational Safety and Health Administration (OSHA) and initiated a nationally harmonized system of workplace standards, record-keeping, inspection, and enforcement. It would not be an exaggeration to say that the employer responsibility view of OSH is the law of the land.

A small but dedicated group of economists, legal theorists, and political thinkers, however, resisted this tide.1 They drew on the contract-centered ideal of classical liberalism to argue that government regulation of working conditions was unnecessary, since they were already regulated efficiently by competitive markets. Workers who felt that a particular employer was too inattentive to safety could look for other work more to their taste. Employers, feeling the pressure of worker preferences, would have to invest in better working conditions, offer extra wages or other compensation for the risks they asked workers to undergo, or both. Indeed, labor market agreements drawn up by self-interested individuals aware of their own circumstances and preferences were believed to offer the best form of regulation; government interference could never improve such outcomes and would most likely make them worse.

There is debate among legal scholars over the extent to which such classically liberal ideas were influential in judicial oversight of the emerging regulatory state. In most cases courts upheld the regulation of working conditions, but the most important exception, Lochner v. New York (1905), established a precedent that held for three decades. John Lochner was a baker who challenged New York state regulation of working hours in his industry; state courts upheld the rule, but the U.S. Supreme Court deemed it an infringement on the right of individuals to “purchase and sell labor” (Rosner and Markowitz 2020). Yet even in this ringing assertion of freedom of contract the court was at pains to argue that employer responsibility for safety was unaffected: “To the common understanding the trade of a baker has never been regarded as an unhealthy one” (Justice Peckham, cited in Bale 1988). Regardless, the reversal of the Lochner precedent by the New Deal era court has generally been seen as the end of strict freedom of contract as applied to OSH and related aspects of employment.

It is remarkable, then, that beginning in the 1970s a research project arose in economics to estimate the amount of extra pay received by workers in exchange for bearing greater occupational safety and health risks. The research was based on the assumption that such risks were voluntarily agreed upon by self-interested parties and reflected their corresponding preferences. In other words, these economists assumed a Lochner, freedom-of-contract world and erected an edifice of theory and statistical research on top of it. This approach has proved to be an influential strand of analysis and advocacy, and one of the main purposes of this paper is to scrutinize it in the larger context of what we know about labor, employment relations, and the role of unequal bargaining power.

Here we will briefly summarize the freedom-of-contract approach to OSH that has gained a foothold in economics in order to see how it fits into the long evolution of thinking on this question—or doesn’t. In a later section of the paper we will return to give it much more careful scrutiny.

As presented by its chief advocate, W. Kip Viscusi, over a series of many books and articles, the contract-centered view takes the following forms. To begin with, it is assumed that the national labor market comprises innumerable markets for specific types of labor, each properly modeled as a relationship between supply (workers with particular productivity-related attributes offering their labor) and demand (employers offering pay for particular jobs). Each market clears where supply equals demand, establishing a single level of compensation for all labor of that type.

If there were no compensation except for money wages, a single wage would be paid to each type of worker; economists refer to this as the “law of one price,” which is thought to hold in competitive markets. (There are reasons why such a law might be violated even under conditions of unrestricted competition, but we won’t consider them here.) But now let us also suppose workers care not only about the wages they are paid but also the working conditions they will experience. From the employer’s point of view (in this theory), improvements in these conditions, including OSH, are expensive. Since productivity is assumed to be equal across firms (the result of competition), and all excess or deficient rates of profit will be competed away, all employers appearing as purchasers of labor in a given labor market will pay the same wages-plus-working-conditions expense. In this sense the law of one price remains in effect.

On the worker’s side it is a bit more complicated. Workers care about safety, so any job that is more dangerous than the available alternatives—the ones offered to workers of the same productivity type—has to pay a higher wage, otherwise it would get no takers. Of course, workers differ individually in how they trade off the benefits of income against the costs of risk, so in subjective terms there isn’t a law of one price anymore; a given worker will likely find one job offer more appealing than another, since it better matches that worker’s own subjective tradeoff. Nevertheless, in the new equilibrium of job offers and acceptances, no workers in the equation would willingly exchange their employment for any other offered to workers of their productivity. Any truly better job, one that offered a preferred combination of wages and working conditions, perhaps better in both respects, would be available only to those with more marketable productivity attributes than theirs.

Thus, in the freedom-of-contract world every labor market, comprising workers of the same type, would display a range of wages and working conditions corresponding to different costs of making work safer (employers) and subjective benefits from safety (workers). Riskier jobs would pay more and would be preferred by more risk-tolerant workers. The extra amount of money employers in this hypothetical world must pay to attract a labor force if their jobs are more dangerous is called a compensating wage differential (CWD). It is important to note that the story underlying this term requires that the CWD be fully offsetting; if it isn’t, the job in question won’t be filled. We will see that this backstory plays a critical role in the interpretation given to statistical findings: Economists wedded to the freedom-of-contract view regard any evidence of wage compensation for risk as a measure of the full value placed on safer work that extra wages are needed to replace. Since we are not beholden to this view ourselves, however, we are able to imagine partial compensation for risk—a monetary increment that might be given to workers laboring under more dangerous conditions that nevertheless leaves them worse off than they would have been had those conditions been improved instead.

Returning to the freedom-of-contract story, it’s worth spending a moment to consider how far-reaching the consequences of this set of assumptions must be. First, the level of safety in every single job would be efficient. Employers would make all investments that result in safety improvements that workers value (and are willing to accept correspondingly lower wages for) above the cost of implementing them, and none whose cost exceeds workers’ valuations. Employers with high costs of making jobs safe would offer more dangerous jobs but pay higher wages; workers with a higher tolerance for risk would accept more dangerous work but would be compensated with the additional wages they require. No regulator would be able to make any alteration that would improve the level and allocation of safety at work. Second, all workers who offer labor of the same type would end up at jobs they value at least as highly as any other they might qualify for. Some would get more dangerous jobs and more pay, others less on both counts, but no one would prefer to switch. Regulation would not be needed to protect anyone from anything. Third, by reflecting workers’ subjective valuation of occupational risk, CWDs could be used to estimate a value of statistical life for society in general insofar as workers are representative of the whole population. If the equilibrium CWD is $1,000 for a 1-in-10,000 increase in the risk of death on the job, the VSL would be $1,000 times 10,000 or $10 million. This third property of CWDs has given economists like Viscusi prominence in the world of public policy analysis, and estimates of compensating differentials are factored into regulatory decisions ranging from air pollution to vehicle speed to climate change mitigation.

Despite the publicity given to this strand of modeling, specialists in labor market research have long known that the law of one price for employers and the no-switch condition for workers are routinely violated; workers of similar observable skills doing essentially the same work under comparable conditions are typically paid quite differently, and sometimes those with lower wages also face greater occupational risks. There are better and worse jobs, and much of the modeling and econometric work of the past several decades has been devoted to trying to understand the forces behind these outcomes and what they imply for public policy. The starting point was the work of Richard Lester in the 1940s, such as “Wage Diversity and Its Theoretical Implications” (1946). It was clear that the simple supply-and-demand story could not be an adequate account of how labor markets operate, since competition should enforce equal compensation for equally productive workers.2

In subsequent years many approaches have been put forward. Some depend on dropping the assumption of perfect competition by dividing either firms, workers, or both into more- versus less-favored segments (Dickens and Lang 1993). Since the 1980s, one wage dispersion of interest has been that of inter-industry (and now also intra-industry) wage differentials. Other economists have revolutionized market analysis by looking more carefully into “search theory,” i.e., how workers search for jobs and employers search for workers. Two progenitors of this approach, Christopher Pissarides and Dale Mortensen, were awarded a Nobel Memorial Prize in Economics in 2010 in recognition of the centrality of the approach to modern labor economics, particularly at the macro level.3 In search models, as we will see, it is expected that the law of one price will be violated, a point emphasized in Mortensen (2003). Yet another approach foregrounds bilateral bargaining between employer and worker, either individually or in the context of collective bargaining. This process is referred to as “rent-sharing,” and bargaining models are often grafted onto underlying mechanisms, such as barriers to competition, that give rise to rents themselves. Meanwhile, closer collaboration between economists and management scholars has generated a range of modeling strategies that incorporate more complex and realistic employer choice options. These include “efficiency wage” models (pay for performance and not just showing up) and incorporation of on-the-job training and other investments or practices that use human resources as a basis for competitive advantage.4 Finally, there has been extensive research into the role of racial, gender, and other forms of discrimination in generating unequal outcomes for workers of comparable productivity.5

Nevertheless, despite all these important theoretical insights, a word of caution is in order. For institutional and other reasons we will not consider here, economists working on a particular departure from the freedom-of-contract model typically treat it as the only such departure; its significance is thought to be the difference it alone would make. Thus, we see an array of labor market models in the literature that are at variance with the freedom-of-contract approach only in one respect even though, taken together, they would point to a radically different vision. This strategy is important in practice because it is common for economists to propose policy responses that address only the single “imperfection” they are studying, and these frequently fall short of the scope of commitment embodied in regulatory interventions like OSHA. Without going into detail, we would like to explicitly stipulate a core set of characteristics of the employment relationship that, in combination, remove it decisively from the freedom-of-contract world:

• All employment contracts are necessarily incomplete. Worker and employer performance cannot be fully specified in advance, and thus an ongoing relationship with procedures for designating required, permissible, and prohibited activities is inescapable.
• The rights, opportunities, and obligations pertaining to each worker individually depend on those for their co-workers, including those above and below them in the chain of hierarchy. Thus, separate individual agreements (contracts) extending to all terms of the employment relationship are not feasible.
• Because of incomplete contracting, what workers sell is their subordination to the employer. This subordination is bounded, of course, by their freedom to terminate, but it is usually so costly for workers to exercise this option that they often accept substantial depredations and still remain on the job. This is the core power imbalance between worker and employer.6
• In all societies, employment is a legal status with associated rights and responsibilities for both parties. The ability to benefit from the provisions of this status also differs among workers for both legal and socioeconomic reasons, constituting another dimension of power.
• Provision of safe working conditions is largely a matter of ongoing performance and not a once-and-for-all commitment made at the moment of contract. In practice it entails such activities as housekeeping and maintenance, work pacing and scheduling, honest and timely risk communication, and the consideration of health and safety factors in the introduction of new materials and technologies. The determination of safe working conditions is subject to voice as well as exit influences from the workforce (Hirschman 1970).7
• Thus, regulatory systems like OSHA, which contribute to the legal status of employment, are essential. They reduce the effect that differences in power would otherwise have on safety and health outcomes, strengthening the ability of workers to engage in individual (whistleblowing) and collective action, and creating legal and financial incentives for improved employer performance. The necessity of regulation can mostly clearly be seen in its breach, as we will document in our concluding section on the coronavirus pandemic. If an economic model has the property that OSH regulation is generally unnecessary or even harmful, this points to a defect with the model, not regulation.

These foundations of our alternative view of OSH and employment do not negate the role of markets but supplement it. Employer costs and worker preferences do matter in the ways specified in market analysis, but they comprise only part of the story. We prefer to consider the theoretical departures of search theory, behavioral economics, and other branches of modern economics as extensions of an institutionally complex view of the employment relationship, not a stripped-down freedom-of-contract benchmark.

IV. The real world of dangerous work in the United States

A. The overall risk to the working population

In this section, we will summarize information about the extent, severity, and impact of injuries and illnesses related to work. We will see that there are a considerable number of work-related injuries and illnesses in the U.S. and that their financial and health impacts are substantial. We will also see that the impacts of these injuries tend to be unequally distributed, concentrated among lower-wage workers, Black and Latinx workers, and temporary, contracted, and online platform workers. In the United States, we have good information about fatal injuries at work, but we know less about nonfatal injuries. Moreover, we have only very indirect ways of trying to capture the extent of work-related illnesses, which are, by far, the largest category of risk.

1. Fatal injuries

Before 1980, the only source of occupational fatality data in the U.S. was the annual Survey of Injuries and Illnesses (SOII) conducted by the Bureau of Labor Statistics (BLS). In principle, the SOII covers all employers except for agricultural employers with fewer than 10 employees and federal government employees (whose injuries are reported separately). Employers are the only source of information for the SOII, so it depends on their diligence and willingness to report. From 1980 through 1991, the National Institute for Occupational Safety and Health (NIOSH) implemented a new source of occupational fatality data, the National Traumatic Occupational Fatality database (NTOF). The NTOF used death certificates, rather than employer reports, as a source of occupational fatality data. Unlike the SOII, it included all employment, including employees of small employers and the self-employed. It found much higher fatality numbers than reported by the SOII, particularly in the high-risk agriculture and construction sectors (Stout-Weigand 1988; Leigh and Garcia 2000).

Since 1992, U.S. national data about work-related fatalities have been collected by the BLS and published as the Census of Fatal Occupational Injuries (CFOI). The CFOI includes fatal injuries of both employees and independent workers (the self-employed and online platform workers). It is generally acknowledged to be a virtually complete accounting of these deaths and includes worker demographic information, industry, and injury characteristics. Figure A shows that, from 1992 through 2009, the annual number of fatal on-the-job injuries was generally declining, from over 6,000 to a little over 4,500. However, after four years of little change, the number of fatalities rose in 2014-2018. Because the number of people working rose during that period, the fatality rate remained about the same, 35 fatalities per million full-time workers. In 2018, 5,250 workers died from injuries at work. Another way of looking at this is that an American worker died on the job every 100 minutes.

Figure A

2. Nonfatal injuries

Surprisingly, no source provides an accurate count of the number of nonfatal injuries in the U.S. The only national source of information for the number and rate of nonfatal occupational injuries is the BLS’s annual SOII, which reported that 3.5 million work-related injuries and illnesses occurred in the private sector and state and local government in 2018. This translates into 3.1 injuries per 100 full-time workers. However, as noted above, the SOII omits self-employed and online platform workers, somewhere between 8% and 10% of total employment, and it also omits federal employees and workers on small farms. Although it does not omit household employees, their work-related injuries may escape reporting. Even for covered workplaces, researchers, the Government Accountability Office (GAO 2009), and the BLS itself (Ruser 2008) have agreed that it misses many injuries. This under-recording is not random, but is affected by many factors, including industry, occupation, duration of time off work, type of injury, and the state where the injury occurred (Boden, Nestoriak, and Pierce 2010; Boden and Ozonoff 2008; Fan et al. 2006). Also, as noted above, the SOII data are submitted to BLS and OSHA by employers only, a practice that may limit completeness of reporting (Rappin, Wuellner, and Bonauto 2016).

Injuries are underreported by workers, employers, and medical providers. Reasons for worker underreporting include fear of current and future employer retribution, employer bonuses for zero reported injuries (so-called safety bingo), and lack of understanding about reporting (Azaroff, Levenstein, and Wegman 2002; Frederick and Lessin 2000). Employers may underreport because of the economic incentives to do so (e.g., they wish to reduce workers’ compensation costs); because middle-manager evaluations include only reported, as opposed to actual, injury rates; because some contract bidding takes into account reported injury rates; or because of poor reporting systems and lack of understanding (Rappin, Wuellner, and Bonauto 2016; GAO 2009). Medical providers, who are workers’ compensation gatekeepers in addition to treating injured workers, have indicated that they frequently perceive pressure from employers to downplay injuries and illnesses (GAO 2009). These factors may be particular problems for low-wage and contingent workers, immigrants, and minorities (Castillo 2018; Premji and Krause 2010).

Researchers looking at different states have estimated that, on average, the SOII misses anywhere between 30% and 70% of injuries involving time lost from work (Boden and Ozonoff 2008; Rosenman et al. 2006). If we assume that the SOII misses half of all injuries, then 2018 saw about 7 million work-related injuries, over 6 per 100 full-time workers each year. Employers reported that about one-third of these injuries involved more than a day off work.

3. Occupational disease

Chronic occupational disease deaths are not covered by the CFOI. The SOII covers occupational illnesses in principle, however in practice chronic occupational illnesses are virtually uncounted (Leigh et al. 2000). There are many reasons for this, perhaps the most important of which is that these illnesses often go unrecognized by both workers and their physicians. Some states have programs targeting specific occupational diseases, like lead poisoning, work-related asthma, and pesticide poisoning. The only exception to the considerable undercount of chronic occupational diseases may be for coal workers’ pneumoconiosis, a disease that NIOSH has focused on for decades.

Because no record-keeping adequately captures occupational diseases, estimates of mortality from these causes have relied on epidemiological studies of specific diseases to estimate the fraction attributable to work. These estimates are then combined with overall U.S. mortality from these diseases (Steenland et al. 2003). The authors of an important study in this area estimated that about 49,000 people die annually from occupational illnesses, with a wide range of uncertainty—from 26,000 to 72,000. Even the low estimate of occupational disease deaths is much greater than the number who die from occupational injuries: If we take 5,500 as the annual average of fatal injuries, then occupational illnesses are responsible for roughly five to 13 times as many deaths each year.

The main causes of death from occupational exposures are cancers, noncancer respiratory disease related to dust and chemical exposures, and coronary heart disease, largely related to job strain (high job demands combined with low worker control) (Steenland et al. 2003).

4. Overall risk

In conclusion, except for traumatic fatal injuries, occupational injuries are substantially underreported in the United States. This underreporting varies by many categories, including worker characteristics, occupation, industry, injury severity, and state. Both incidence of and mortality from chronic occupational disease are barely reported at all. Experts can only guess at these risks, so it would seem unlikely that workers would have a reasonable sense of the probability of dying from disease hazards at work.8 One consequence of this lack of information is that workers can’t base their wage demands on the true risk they face on the job, a shortcoming that calls into question estimates of the value of a statistical life predicated on the assumption they can. Our best estimates are that, each year:

• 5,500 people die from work-related injuries.
• Another 49,000 people die from work-related illnesses.
• 7 million people are injured at work every year, and over 2 million of these lose at least a day from work.
• Many people suffer discomfort and disability from work-related illnesses, but we don’t know how many. However, work-related chronic lung and heart diseases can have life-altering consequences that last for decades before they result in premature death.

B. The distribution of risk

Of course, not all jobs are equally risky. Table 1 shows nonfatal injury rates for some high-risk occupations compared with the average for all occupations, and Table 2 shows fatal injury rates. Many of these high-risk jobs have relatively low wages. More generally, groups that have less power, knowledge, or ability to affect workplace hazards and groups that are discriminated against are more likely to be at risk of injury in the workplace.

Table 1

Table 2

Temporary, contracted, and online platform employment. Temporary agency workers often have higher fatal and nonfatal injury rates than do workers in standard jobs (Foley 2017; Julià et al. 2016; Smith et al. 2010). Similarly, studies of subcontracted employment have shown higher risks than in standard work (Kochan et al. 1994). Low-wage workers also tend to have higher injury rates. One study showed workers in low-income families had higher risk of injury, even accounting for industry and occupation (Dembe, Erickson, and Delbos 2004), another that people earning less than $50,000 per year reported a greater risk of injury than those with higher earnings (Fan et al. 2006).

Temporary and short-term workers, frequently hired through temporary employment agencies, may be particularly vulnerable to workplace safety risks. As noted in an OSHA white paper:

New workers often lack adequate safety training and are likely to be unfamiliar with the specific hazards at their new workplace. As a result, new workers are several times more likely to be injured in the first months on the job than workers employed for longer periods. Consistent with these findings, OSHA has investigated numerous incidents in recent months in which temporary workers were killed on their first days on a job.

Temporary workers are also likely to be newly assigned to unfamiliar workplaces multiple times in any given year and may carry this increased risk as long as they are in the temporary workforce. For employers, there is less financial incentive to invest training resources on temporary employees because shorter tenure will yield a lower return on investment than similar investments for permanent employees.

OSHA has encountered many situations, including some in which temporary workers have been killed, in which employers have chosen to not provide required safety training to temporary workers. And the temporary workers themselves, recognizing the precarious nature of their employment, are less likely to complain to their employers, or to OSHA, about the existence of even serious hazards (OSHA 2015).

Workers in temporary employment relationships are often subject to the same occupational hazards faced by others in the same work environments in standard employment relationships. In addition, these workers are likely to have little control over their work schedules or pace, may be hired only during periods of high demand, and have few social supports in the workplace. They may also have limited training in job tasks, job risks, and prevention of injury or adverse health exposures. They may not have access to personal protective equipment and training in its use. And, in some cases, they may be assigned to the most dangerous jobs (Mehta and Theodore 2006).

ProPublica reporters compared injury rates for jobs held by temporary workers with jobs held by regular employees, accounting for whether the jobs they held were particularly hazardous. For jobs with similar injury risk they found that the odds of injury for temporary workers were almost four times as high as for regular employees (Pierce, Larson, and Grabell 2013).

Short-term and seasonal workers may be more subject to job strain and its adverse health consequences and less likely to benefit from the workplace conditions that may mitigate these effects (Cummings and Kreiss 2008). Job strain can result in both physical and psychological disruption; prolonged job strain leads to increased cardiovascular disease, musculoskeletal disorders, sleep disruption, and psychological problems. Exposure to workplace conditions like job insecurity, low job control, high job demands, and low social support at work may explain a substantial proportion of observed inequality in life span in different demographic groups in the U.S. (Goh, Pfeffer, and Zenios 2015). The growth in temporary, contracted, and online platform employment relationships that results in increased exposure of less-educated and ethnic minorities to harmful workplace practices is likely to result in poorer health.

Also, temporary, contracted, and online platform workers may have less access to health insurance and workers’ compensation benefits (Asfaw 2014; Mehta and Theodore 2006), causing greater financial strain and interfering with recovery from injury or illness. For a more detailed look at these issues, see The Fissured Workplace (Weil 2014) and The Changing Structure of Work: Implications for Workplace Health and Safety in the U.S. (Boden, Spieler, and Wagner 2016).

Workers of color. There is also evidence that workers of color, particularly Black workers, have elevated injury risk. Loomis and Richardson found that Black workers were 30-50% more likely to die from an occupational injury compared to similar white workers (Loomis and Richardson 1998). The disparity in fatal injury risk was primarily accounted for by differences in the underlying hazards of their occupations. Another study indicates that Black men have greater fatality risk than white men working in the same industry. Black agriculture workers’ fatal injury rate is 26.9 per 100,000 workers, compared to 21.2 for whites (CDC/NIOSH 2004). Within the same industry Black and white workers may have very different jobs.

Another study examined six occupations, employing 16% of private-sector workers in the U.S., at high risk of nonfatal work-related injury. The authors found that Latinx, Black non-Latinx, and American Indian/Alaska Native workers were substantially overrepresented in these occupations (Baron et al. 2013). A more recent study found that non-Latinx Black workers and foreign-born Latinx workers were more likely to work in riskier jobs than white workers, leading to higher rates of work-related disability (Seabury et al. 2017). These results held even after accounting for education, age, and sex. Other research has shown increased risk of fatality and injury among immigrant Latinx construction workers (Dong and Platner 2004) and Latinx hotel workers (Buchanan et al. 2010).

Although studies of occupational disease in this context are rare, there is evidence that Black workers have elevated occupational disease risk. For example, a 1971 study found that, among steel workers, coke oven workers had the highest lung cancer mortality (Lloyd 1971). Moreover, Black coke oven workers were largely working on the topside of the coke ovens—hot, dirty work that resulted in heavy exposure to carcinogenic emissions. A more recent study of lung cancer among chromium smelter workers also found Black workers at higher risk (Rosenman and Stanbury 1996).

Some studies did not find increased injury rates among nonwhite workers. One found substantial excess risk of occupational fatalities among Latinx and immigrant workers, but not among Black non-Latinx workers (Marsh et al. 2013). Studies using the National Longitudinal Survey of Youth 1979 did not find higher injury rates among nonwhite workers, but one found longer absences from work (Berdahl and McQuillan 2008; Strong and Zimmerman 2005). However, most of the evidence supports the view that white non-Latinx workers have the lowest risk.

Using another way of looking at relative power, Peter Smith and his colleagues have studied the relationship between injury and worker vulnerability (Smith et al. 2015). Their concept of vulnerability combines potential exposure to hazards with the ability to mitigate those hazards. Mitigation resources include workplace policies and procedures, workers’ awareness of their rights and responsibilities, and workers’ empowerment to protect themselves. Workers who reported being more empowered had much lower injury rates (Lay et al. 2017).

C. The burden of injuries at work

1. Lost earnings

Although there is a history going back to the early 20th century of studies on earnings lost because of injuries at work, researchers have applied modern statistical methods only over the last two decades. Beginning in 1999, a series of studies using workers’ compensation data from California, New Mexico, Oregon, Wisconsin, and Washington has consistently shown that injured workers who received workers’ compensation benefits for lost wages had substantial earnings losses and that these losses persisted for as long as the injured workers were followed. One study focusing on permanent partial disability (PPD) cases in five states estimated that affected workers had lost about 20% of their earnings over a five-year period (Reville et al. 2001). A 2014 study following New Mexico workers for 10 years after injury found average losses of about 20% over the full 10-year period (Seabury et al. 2014). Indeed, all studies have found losses that continue for as long as the study can follow injured workers, and from these studies we conclude that PPD injuries cause a lifetime of substantial lost earnings.

Some studies of lost earnings have also looked at temporary disability (TD) cases, where benefits were paid for time lost from work but not for permanent impairment or disability. The New Mexico study just cited (Seabury et al. 2017) found that in TD cases with one to eight weeks off work workers lost an average of about 10% of earnings averaged over the 10 years after injury. These losses looked like they would continue far into the future, but the workers were followed only for 10 years. Those losing 8 weeks or more lost almost as much on average as those with PPD injuries. A California study found that workers with injuries occurring from 2005 to 2017 who lost at least four days from work lost about 20% of earnings over the two post-injury years (Rennane, Broten, and Dworsky 2020).

A study examining differences in injured workers’ lost earnings by income found that workers in the highest wage quartile had the lowest percentage loss in earnings, while workers in the second quartile had the greatest percentage loss in earnings (Rennane, Broten, and Dworsky 2020). Differences in lost earnings among earnings or racial/ethnic groups have not been well studied.

Workers’ compensation programs only partially cover the earnings losses of injured workers who receive benefits. The extent of this coverage depends both on whether injured workers receive benefits and the proportion of lost wages covered when benefits are paid.

2. Proportion of injured workers receiving workers’ compensation benefits

Studies of the take-up of workers’ compensation benefits have uniformly shown that many injured workers never receive any benefits. Because state systems vary substantially, the proportion of injured workers receiving benefits can vary widely. A six-state study of injuries eligible for cash benefits (lost-time cases) found that workers’ compensation benefits were paid to about 90% of eligible injured workers in two states but to just 50-70% of cases in the other four (Boden and Ozonoff 2008). To be eligible for cash benefits that replace lost earnings, states require that the number of days lost from work exceed the waiting period, which varies between three and seven days. A similar study in Michigan found that 66% of eligible injuries received workers’ compensation benefits (Rosenman et al. 2006). A study of workers’ compensation take-up in 10 states found that, of work-related injuries involving medical treatment or at least one day lost from work, workers’ compensation benefits were received in 47-77% of cases (Bonauto et al. 2010). Presumably all the injuries in this study would have been eligible for workers’ compensation medical benefits, but not necessarily for cash benefits.

Aside from specialized federal compensation programs (the Black Lung Program, the Radiation Exposure Compensation Program, and the Energy Employees Occupational Illness Compensation Program), virtually no chronic occupational disease victims receive compensation. This means that, since occupational illnesses have a far greater incidence than injuries, the bulk of occupational illnesses and injuries go uncompensated.

3. Proportion of lost wages replaced among injured workers receiving workers’ compensation benefits

Several of the studies of lost earnings that used modern statistical methods also calculated the proportion of injury-related lost wages replaced by workers’ compensation cash benefits. The earliest of these studies—the five-state PPD study referenced above—found that workers’ compensation benefits replaced 38-60% of after-tax lost earnings for PPD cases in these states (Reville et al. 2001). (After-tax earnings are used because workers’ compensation benefits are not taxed, while wages are. We note, however, that lost fringe benefits are not considered in this calculation, meaning that the actual income-replacement rate is even lower than what these studies report.) A more recent study in New Mexico, using more complete data, estimated replacement of only 16% of after-tax earnings lost in the 10 years after injury (Seabury et al. 2014). This study also found that workers’ compensation benefits replaced less than 10% of after-tax losses for injuries classified as temporary disability cases. The reason for this somewhat surprising result is that even people receiving less than eight weeks of TD benefits, on average, have losses that continue far into the future.

Some of these lost earnings may be partially covered by sick pay, private disability insurance, public short-term disability insurance (in some states), unemployment insurance, Social Security Disability Insurance, Supplemental Security Insurance, etc., but the information about this is limited. One study estimated that workers’ compensation cases annually account for about $9 billion in Social Security Disability Insurance benefit payments and about the same amount in Medicare payments (O’Leary et al. 2012).

4. Proportion of medical costs replaced among injured workers receiving workers’ compensation benefits

In principle, workers’ compensation pays all injury-related medical costs for covered injuries. Of course, a substantial number of workers with occupational injuries and almost all workers with chronic occupational diseases do not receive workers’ compensation, so their medical costs are not covered. There is virtually no information about the cost of uncovered injury-related medical expenses, nor is there information about how much of those expenses are covered by general health insurance, Medicaid, Medicare, workers and their families, or other sources.

5. Work-related injuries and financial distress

Many working families live paycheck to paycheck and have little or no savings to fall back on. A 2018 survey asked people whether they could cover an unexpected expense of $400; only 61% said they could cover it with cash, savings, or a credit card paid off by the next statement. A 2017 survey asked people whether they could cover a $2,000 unexpected expense within a month. Over 40% said they could not, even including cash from credit card advances, payday loans, or checking account overdrafts; even greater proportions—54% of Latinxs and 64% of Blacks—gave this response (Stavins 2020). Not surprisingly, the proportion who could not cover the expense fell as incomes rose.

Reductions in income after injury can lead to problems in covering rent or mortgages or car loan payments, increased credit card debt, and general difficulties in making ends meet (Keogh et al. 2000; Morse et al. 1998). Even spending on food declines after injury (Galizzi and Zagorsky 2009). Groups with no or minimal savings are particularly hard hit by occupational injuries and illnesses, with Latinx and Black workers much more likely than white workers to fall into this category. Generally, low-income workers with relatively little savings are more likely to be injured than are other workers (Galizzi and Zagorsky 2009).

6. Other costs

Lost household production. People with work disabilities may not be able to engage fully in household tasks like caring for children, cleaning, or taking out the trash. Several interview studies have shown the considerable extent of these impacts on home life (Imershein, Hill, and Reynolds 1994; Keogh et al. 2000; Strunin and Boden 2004).

Depression and drug and alcohol use disorders. A substantial number of injured workers can’t do the same work or household tasks, lose their jobs, live with intermittent or constant pain, or can’t participate fully in many aspects of family life. There is substantial and growing evidence that these injury effects can lead to depression (Asfaw and Souza 2012; Dersh et al. 2007; Kim 2013) and drug use (Asfaw and Boden 2020). These health consequences have recently been coined “diseases of despair” and have been cited as important reasons for recent declines in U.S. life expectancy (Case and Deaton 2017).

Deaths indirectly caused by workplace injuries. Deaths from suicide and drug overdose, in particular, have been growing in recent years and have become major public health concerns. Moreover, work-related injuries have been shown to increase the risk of dying from both of these causes (Applebaum et al. 2019; Martin et al. 2020).

D. International comparisons

Information from two studies of fatal workplace injury rates in the U.S. and other high-income countries strongly suggests that U.S. rates are higher. The first of these studies compared fatal injury rates among countries around the world. For reasonable comparison, we chose to compare the U.S. with other countries whose 2018 gross domestic product per capita was at least $40,000 (in U.S. dollars). As is evident in Figure B, the U.S. had the highest work-related fatality rate of these countries (Kharel 2016).

Figure B

It is possible that differences in industry composition skewed this comparison. That is, if the U.S. workforce was more concentrated in hazardous industries, one might expect the overall work-related fatality rate to be higher. A second study by Bureau of Labor Statistics economists takes this possibility into account by comparing work-related fatality rates between the U.S. and the European Union by industry sector (Wiatrowski and Janocha 2014). Table 3 shows that U.S. rates are higher than E.U. rates in 11 of 13 industry sectors. Indeed, in three of the four most hazardous sectors U.S. rates are about twice those of the E.U. U.S. rates are lower only in the financial and insurance sectors and the professional, scientific, and technical sectors, the two with the lowest fatality rates. Not only is the overall U.S. fatality rate higher, but disparities between the safest and most hazardous industries are greater than in the E.U.

Table 3

Because nonfatal injury rates are underreported in the United States, and because we do not know the proportion of injuries reported in other countries, we don’t know how U.S. nonfatal injury rates compare with those of other countries.

V. The value-of-statistical-life literature: What it says, what it gets wrong

The title of Kip Viscusi’s second book, Risk by Choice: Regulating Health and Safety in the Workplace (1983), in which he developed the perspective that has governed the VSL literature ever since, states bluntly its freedom-of-contract assumptions. Without referencing the actual legal, medical, and social history of OSH, it posits an ideal model of how occupational risk would be transacted if the market for labor services were determined solely by self-interested agents whose only concern was to strike the best possible bargain. Employers are assumed to maximize profits, treating both wages and safety as costs of production with no corresponding benefits. Workers are assumed to maximize utility derived positively from both safety and wages. Competition assures that employers all earn the same profit, i.e., zero in economic terms (net of the cost of capital), and workers of a given type all receive the highest level of utility attainable given the jobs they and their peers are offered. The market for each type of labor clears, so no employer faces an unfilled job or worker a spell of unemployment. Employers and employees have equal power to contract or walk away from any bargain.

This equilibrium can be depicted in Figure C, where wages are measured on the y-axis and risk on the x-axis.

Figure C

Here w represents the wage and r the level of risk of death on the job. For a given worker utility function, the curve u = u⁰ traces combinations of wages and risk that leave combined utility unchanged at the market-clearing utility level u⁰. The curve incorporates increasing marginal disutility of risk (the curve becomes nearly vertical as a critical level of risk is approached). The other curve, π = π⁰, traces the corresponding combinations that leave employer profit π unchanged at the competitively determined level of zero. The employer faces increasing marginal costs of reducing risks (diminishing marginal return to safety investments), so the curve becomes steeper as one goes from right to left. Profits and utility are jointly maximized at the tangency indicated by dw/dr, the marginal tradeoff between wages and risk for this particular contracting pair. At market equilibrium for a worker with this productivity (which determines u⁰) and utility map and for an employer with this cost-of-safety schedule, an equilibrium wage of w* will be established along with an equilibrium risk r*. In a more complete representation of this model there would be many iso-utility curves reflecting different worker attitudes toward money income and risk and many iso-profit curves representing different employer cost-of-safety conditions. More safety conscious workers would sort to be matched with employers with less expensive safety options and vice versa. An upward-sloping contract locus would connect all these individual tangencies.

Here we can see all three properties of compensating wage differentials in a purely contract-centered world:

1. The level of risk is perfectly efficient, since the marginal cost of an additional increment of safety is exactly equal to the marginal benefit (in monetary equivalent) it would provide to workers. Safety is neither too low (where improvements inexpensive for the employer to provide would be highly valued by workers) nor too high (where the cost of providing some of the improvements exceeds workers’ valuation of them). Regulation cannot improve this outcome.
2. Workers in dangerous jobs would be no better off if they switched jobs with equally productive workers in safer ones. The dangerous jobs pay higher wages so that each worker receives that worker’s u⁰ in overall compensation. True, with their subjective differences, some workers will end up more satisfied with their work than others, but there is no a priori reason why such differences should have any systematic relationship, positive or negative, to safety.
3. The marginal tradeoff between wages and risk, dw/dr, can be employed to calculate the implicit value of life voluntarily chosen by workers at this point in the contract curve. Multiply this incremental equivalent by the number of such increments required to sum to a single statistical life (1 divided by the marginal risk change) and the result is the aggregate money that many such workers would collectively require to reduce expected occupational deaths by one. If workers are viewed as representative of the larger society, this VSL can be applied to any policy change that reduces expected mortality but imposes other costs, e.g., mitigating climate change, reducing toxic emissions from power plants, or, for that matter, regulating risk in the workplace.

Before moving forward, it is important to emphasize that this model begins with the core assumptions of a freedom-of-contract world. Workers care only about their individual well-being, a function of their income and the disutility they experience from occupational risk, and they perform this calculation rationally and consistently, unaffected by social or ethical considerations. Similarly, firms care only about profits, which depend on the wages they pay and the safety investments they make; neither worker commitment nor market reputation plays a role. The labor market is in continuous equilibrium, defined as a state in which supply equals demand and there is no opportunity for any party to improve its outcome by altering the terms it offers to either supply labor or employ it. All relevant outcomes of employment, including actual safety on the job, are determined at the moment employment offers are accepted. Employers and employees have equal bargaining power, a claim increasingly contested by empirical research. There is no regulation of the terms of employment, either by government or collective bargaining agreements; in fact, the legal status of “employment” itself is undefined because none of the questions employment law resolves are at issue. These aren’t claims supported by historical, institutional, or empirical research; they are assumptions made prior to any theoretical or statistical analysis.

On the basis of the preceding model, VSL researchers estimate a wage equation of the following form:

ln w_i = α + βX_i + γr_i + ε_i

where w_i is the wage of worker i, α an intercept, X_i a set of control variables for worker i, β a vector of coefficients on these variables, r_i the occupational risk faced by worker i, γ the coefficient on this risk, and ε_i the error term. By incorporating control variables, the intent is to isolate the effect that risk has on wages for equivalent jobs, with the coefficient γ representing the wage effect of a unit change in this risk, which, when multiplied by the level of risk r becomes the compensating wage differential; for instance, if γ = 0.02, average w = $50,000, and average r = 1/10,000 per year, estimated VSL (at these averages) would be [2% of $50,000] x 10,000 = $10,000,000. (These numbers are only for purposes of illustration.) Referring again to Figure C, it is assumed that each worker is of a given productivity type, where this type is a function of attributes like age, gender, experience, education, race, union membership, and marital status. (Note that gender, race, and union membership are routinely interpreted as signaling differences in “productivity.”) Each productivity type applies its utility maps to the same set of wage offers in the competitive market, so including these variables is expected to strip away this potential confounder.

A large number of studies have been conducted along these lines, and most of them find positive coefficients γ on risk, representing wage compensation for more dangerous working conditions. When these marginal tradeoffs are multiplied by the mean change in risk of death, a VSL of several million dollars is typically estimated. Government agencies in the United States and a few other countries have officially adopted these estimates for use in benefit-cost analyses of regulations and public programs, thereby influencing the ability to regulate pollution, workplace safety and health, vehicle speeds, and every other regulatory issue affecting life and health. The VSL also enters into calculations of the “social cost of carbon,” used to justify and calibrate measures to combat climate change (Melillo, Richmond, and Yohe 2014).

The procedure is endorsed by leading lights in the “law and economics” movement, such as Cass Sunstein (e.g., Sunstein 2014), for what they regard as its rational approach to valuing lives. VSL calculations have made their way into introductory economics textbooks as examples of the insight that comes from applying “the economic way of thinking” to problems usually considered fraught or vexing (Mankiw 2014). For many whose only analytical exposure to labor issues occurs in the undergraduate economics classroom or in the use of VSL studies, the freedom-of-contract perspective represents the “economic” approach to OSH.

It is important to recognize, however, that the VSL literature and its underlying analytics are strongly disputed within the economics profession, particularly by those with a deep background in the study of labor markets. For a striking example, consider Labor Economics by Pierre Cahuc, Stéphane Carcillo, and André Zylberberg (2014), which has established itself as the leading graduate textbook in its field. Much of Chapter 5 in the second edition is devoted to disputing the modeling and empirical approach of VSL proponents, which the authors find simplistic and misleading. In the pages that follow we will explain some of these criticisms and add several of our own.

At the theoretical level it should already be clear that the modeling framework used by VSL proponents assumes away nearly all the problems that have preoccupied labor economists in the decades since the pioneering work of Richard Lester and have recently reemerged as central to our understanding of labor market trends. In Figure C, the determination of the slope of the contract curve by the marginal rate of substitution between wages and safety for each corresponding worker embodies the law-of-one-price constraint, but modern labor economics begins by asking why this law doesn’t hold: There is an array of wages paid to workers with similar skills or occupations. The most widely subscribed approach in the discipline today is search theory, in part because it provides a plausible, readily modeled explanation for both wage dispersion and involuntary unemployment. A formal proof that fully compensating wage differentials are a special rather than general case in a search framework was provided by Lang and Majumdar (2004).9 As they demonstrate, a search theoretic version of Figure C would portray many iso-utility and -profit tangencies, and the locus of them could as well be downward as upward sloping—lower wages and greater risk for workers of equivalent potential productivity. With search frictions there is no guarantee that competition will prevent the emergence of good and bad jobs corresponding to safer and more dangerous ones.

Similarly, a formal proof by Dorman (1998) shows that wage differentials will not be fully compensating if worker effort depends on the pay and working conditions provided by the employer—the efficiency wage case. This effect can only be magnified when the theorist takes into account the many other linkages between compensation, worker commitment, investments in training and promotion, and the role of problem-solving for flexibility and innovation characteristic of strategic human resource management (Bartling, Fehr, and Schmidt 2012). Compared to the more realistic models on which modern labor market analysis is based, the unadorned supply-and-demand framework of the VSL researchers can only be regarded as naive. Indeed, what is remarkable about employment as viewed through the freedom-of-contract lens is that it is modeled as a one-off bargaining game played by workers and employers, rather than a repeated game in which performance and not just acceptance or rejection of an employment contract is part of the strategy set. Worker performance is about how much effort is supplied as well as engagement in problem-solving and the exercise of initiative, while firms perform on such dimensions as job security, working conditions (including health and safety), opportunities for training and promotion, and their demonstration of respect and appreciation in general. It is clear that in a more realistic model the level of worker protection will likely play an important signaling role.

A second set of criticisms targets the behavioral assumptions of the freedom-of-contract approach. A useful starting point is the way occupational risk is represented in the VSL literature. Recall the regression equation above, for which it was explained that r_i is “the occupational risk faced by worker i.” This risk is typically measured as the annual occupational mortality rate faced by workers in a given job type, ranging from zero, if the job has no such mortality at all, to some number of deaths per 100,000 workers or other denominator. Each increment to this measure—each additional death tabulated for this job type—is modeled as having exactly the same effect on workers’ subjective perception of how dangerous their work is. This notion that goods are valued on a continuous, absolute scale violates one of the best-supported theories at the intersection of economics and psychology: prospect theory (Kahneman and Tversky 1979). According to prospect theory, evaluations are comparative: People typically have reference levels (norms) for goods and respond differently depending on whether the amount offered is above or below this level.

In most applications the reference quantity is assumed to be the status quo, with the behavioral distinction pertaining to gains or losses relative to it, but the theory itself is general. In the context of OSH, prospect theory suggests that workers may distinguish between largely acceptable (at or below the norm) and unacceptable risk. If this is correct, the risk variable in VSL studies is misspecified, failing to register risk differentials that matter most while (perhaps) exaggerating others. For instance, a small difference in fatal risk between two jobs may have little effect on workers’ valuation of the jobs if both are seen as acceptably safe, but if the difference pushes one of them over the acceptability threshold the effect may be substantial. This threshold may apply to a particular source of risk, one workers may think should not be a source at all; in other words, the norm for some types of risk may be zero. That people put up with substantial risks in some aspects of their life and work while vehemently opposing them in others is not irrational if we recognize the ethical—normative—dimension of risk.

Closely related is the observation that the behavioral model implicit in the treatment of risk in the VSL literature assumes that workers care only about the amount of risk they are exposed to and not the process by which the risk level was determined. This assumption has been widely contradicted by studies that show that individuals perceive less risk or are less perturbed by the risk they perceive when they believe they have control over it. 10

That is, self-imposed risk is evaluated as more acceptable than externally imposed risk—important when, as we will soon see, vehicle accidents play a large role in traumatic occupational fatalities. In this respect risk is like many other goods, where procedural utility supplements and modifies outcome utility (Frey and Stutzer 2005; Bartling, Fehr, and Herz 2014). The sensitivity workers often express toward how they feel themselves treated by employers in safety and other aspects of work extends two arguments we have already seen. First, insofar as employment is better modeled as a repeated rather than a one-off game, signaling becomes a crucial aspect of bargaining and performance, and how well employers mitigate safety and health risks can well serve as a signal of their overall intentions. Second, to the extent that a certain degree of risk can be viewed as intrinsic to the type of employment at issue, risk levels above it are likely to be regarded as discretionary—as imposed by actions or inactions deliberately chosen by the employer. Thus, the reference point at the heart of prospect theory can logically be the level of risk above which employers can be assigned responsibility, along the lines of the earlier discussion of the evolution of OSH law.

Finally, behavioral research has identified many cognitive processes that mediate between objective risks and subjective risk perception. In the field of OSH the most important is probably the avoidance of cognitive dissonance, among the most firmly established results in all of social psychology. The core idea is that if new information casts doubt on beliefs individuals rely on to make meaning of their world or undergird their sense of self-worth, consideration of it is painful. To avoid this pain individuals will often refuse to receive or accept such information. The common name for this behavior is denial, and responding to dangerous working conditions with denial is widespread enough to be regarded as normal. “You have to get it out of your mind” is a refrain often heard on the part of workers at risk, but the consequence is diminished risk perception and less disutility from dangerous working conditions than the standard regression model is predicated on (Akerlof and Dickens 1982; Nelkin and Brown 1984). 11

Aside from perception and interpretation issues, however, there is another obvious reason why subjective and objective risk may differ: Employers have an incentive to obscure the risks to which their workers are exposed. Indeed, the symbiotic interaction between the complex etiology of industrial diseases, employer disinformation, and worker denial may well place a substantial portion of true risk beyond the purview of labor markets. The problem is further exacerbated by the ability of employer-financed institutions to fund studies with the purpose of obscuring the hazards workers face from chemical exposures, repetitive trauma, and other risks identified by independent health professionals (Michaels 2008; Michaels 2020). Note that the complications associated with perverse incentives for risk disclosure rest on top of the irreducible complexities of expectation and trust that arise from safety being the result of performance over time and not an upfront, tangible commitment, as discussed earlier in this chapter.

Aside from simplistic economics and inappropriate behavioral assumptions, the freedom-of-contract approach to OSH encounters another problem: It is factually incorrect. It is based on the presumption that the relevant terms of the employment relationship are freely negotiated between workers and employers, but as we’ve seen, they aren’t. It is not true to say that workers “accept” the working conditions they encounter at their place of employment, not legally at least. As discussed earlier, occupational safety and health has been regulated in one form or another since the Middle Ages in the English-speaking world. The current legal framework is established by the Occupational Safety and Health Act, which proclaims at the outset that employers bear responsibility for ensuring that the workplace is free of hazards. Just how this legal framework affects worker perception and decision-making is a complex question, but the categorical answer that it doesn’t—the premise on which the freedom-of-contract approach rests—is surely false. To put it differently, it is not, nor should it be, unreasonable for workers to expect their employers to live up to the language of the law and address workplace hazards as they become evident.

Up to this point we have considered the theoretical underpinnings of the CWD model; now we turn to its statistical implementation. Consider once more the regression equation above: What are the pitfalls?

Perhaps the most obvious difficulty is that, when we apply such a model to a sample of thousands of workers, while we know with some confidence how much they earn, how old they are, how much education they’ve had, and so forth, we don’t have very accurate measures of the degree of risk they face at work. In U.S. studies the now-standard approach, epitomized by the many studies conducted by Viscusi and his coauthors, is to apply the following formula to each individual i:

where j is the occupation for i and k is the industry. This is an average mortality over that industry-occupation pair, and it is applied to each worker who shares those same occupational and industry codes. To give a feeling for the scale of this procedure, the regression reported in Viscusi’s 2013 paper, referenced below, assigns fatal injury risk to over 126,000 workers across 50 industries and 10 occupations, giving an average of just over 250 workers sharing the same injury rate. An obvious concern is that this generates a poor measure of individual risk, since some will face less than this average and others more. Indeed, if one accepts the notion that there are better and worse jobs even within the same intersection of industry and occupation, it is plausible that a negative relationship between risk and wages is being obscured at the individual level. (In this scenario, individuals with higher incomes given their industry and occupation would also enjoy safer jobs.) Such a presumption is supported by research that shows that a substantial portion of overall wage inequality is accounted for by variation across individual employers (Lane, Salmon, and Spletzer 2007; Appelbaum 2017; Card et al. 2018). Similarly, as we saw in the previous section, Black workers have been found to be concentrated in more dangerous jobs than their white counterparts even within the same industry; it would be optimistic to believe that such discrimination would be fully captured by adding occupational to industrial sorting. It is common for economists to brush aside these concerns under the presumption that any negative association between wages and safety found at the level of occupations and industries should also apply within these cells at the employer and individual levels, but this is simply an instance of the “ecological fallacy” (Piantadosi, Byar, and Green 1988).12 Aside from the potential for bias, even measurement error orthogonal to the variables of interest due to assigning an average risk measurement to individuals reduces the likelihood that a given estimated effect of risk on wages reflects a “true” process; if there is conscious or unconscious methodological selection on statistical significance, such as choosing control variables or sample exclusions that generate a low p statistic for the risk coefficient, the effect size—in this case the estimate of CWDs—will be biased upward (Loken and Gelman 2017).

A second problem with the risk variable employed by Viscusi et al. is even more dramatic. Recall from the previous section that, while accurate data exist for fatal occupational injuries, we have only the most limited data on fatal occupational diseases—and to the extent we can compare their magnitudes, the second is from five to 13 times the first. Worse, we know nothing about the distribution of these nontraumatic fatalities across industries and occupations, the basis VSL researchers use to assign occupational risk measures to individuals. Consider the two polar cases, and for simplicity let the number of disease fatalities be 10 times fatal injuries. Assume first there is a perfect negative correlation between fatal risks from injury and disease. In that case, the industry-occupation cells deemed most safe by Viscusi would actually be the most dangerous, and the wage-risk relationship would be nearly the opposite of what he claims to have found. At the other extreme, the distribution of these two risks across industries and occupations is perfectly coincident, which would preserve the CWD effect, but the measured wage increment would correspond to a risk differential that is approximately 10 times as great. This calculation means the VSL would be just a 10th of its current estimation. Of course, neither of these polar cases is likely, but the actual relationship between the distribution of the measured risk of a fatal injury and the unmeasured risk of a fatal disease is simply unknown—a revealing indicator of how little actual concern a true measure of risk at work arouses in this group of researchers ostensibly studying it.

In this connection it should also be noted that the subset of occupational fatalities captured by the CFOI only loosely corresponds to the commonplace notion of “dangerous working conditions.” Approximately 40% of all fatal injuries each year are the result of motor vehicle accidents, including both those in which the worker was behind the wheel, where driving was a component of work responsibilities, and many others where the worker was on the receiving end. Such injuries might be disproportionately born by better-paid workers whose work is less restricted to a single location, but if so should this be viewed as wage compensation for risk? Somewhat less than another 10th of all CFOI-recorded deaths were caused by homicides, some of which might well be the result of foreseeable risk (night shift at a convenience store), but others would be essentially random events (U.S. DOL-BLS 2020b). 13 One of the disadvantages of measuring risk with only traumatic fatalities and not the far more numerous toll of fatal illnesses is that the latter may more closely reflect what is commonly understood to be “hazardous” work.

Yet another problem is the likelihood of missing variables in the regression model. Nearly all the literature that addresses this issue considers only missing variables that might capture differences in worker productivity. Recall that the underlying model behind the VSL studies implies that wage compensation for risk should arise only within labor markets specific to a particular worker type. If the variables included in the vector of X_i fail to fully distinguish between these different types of workers—which inevitably they must—the model will be comparing wages and safety levels for at least some workers across different labor markets. Since we expect workers in a higher-productivity labor market to have both higher wages and better working conditions on average, this statistical shortcoming will introduce a downward bias to measured wage compensation within each market. Not surprisingly, the general view among economists is that the VSL literature struggles to capture labor-market-relevant worker differences and probably underestimates true CWDs.14

There is less recognition, however, of the reciprocal problem: The econometric model is likely to be missing variables that would capture relevant differences on the employer side of the market, meaning differences between factors that affect worker compensation at the industry and occupation level as well as those stemming from different types of firms. In particular, some industries, occupations, or firms will offer higher wages because of features like greater product market power, higher capital-labor ratios (which reduce the labor share of production cost), and a greater concentration of workers with above-average clout, like union members or simply white males. If some industry-occupation cells have both more fatal injuries and other attributes that lead to higher wages, estimates of CWDs would be biased upwards, and incorporating a fuller set of such variables would cause the estimates to fall. Indeed, that is exactly what Dorman and Hagstrom (1988) found in a study published in the mid-1990s based on data from about a decade earlier: Inclusion of variables found to have explanatory power in the interindustry wage differential literature, which looks at differences in wages at the industry level, reduced measured compensation to the point where the only discernible wage compensation for risk was acquired by members of labor unions, and even their CWDs were lower than the average differentials found by Viscusi and his coauthors using the same data. 15 It is interesting to note that, in the more than two decades since the importance of incorporating industry-level variables was first documented, no further study by economists specializing in VSL estimation has used them.16

The final issue that arises with the empirical specification employed by Viscusi and others is that it assumes by construction that all variables have the same wage effects for all workers; that is, workers are permitted to differ according to their age, race, gender, education, experience, and average measured risk, but the coefficients on these characteristics—the effects they are estimated to have on wage outcomes—are the same for all. For example, the effect that education is assumed to have on how much a worker earns is forced to be equal for male and female workers, white and Black, and young and old, and the same, of course, goes for exposure to risk. In this way, discrimination and other forms of unequal power or treatment in the labor market are simply assumed away. At the very least, conscientious modeling would permit personal and industry-level factors to have different effects based on race, gender, experience, education, union status, and similar power-related distinctions, and ideally a multilevel modeling approach would be employed to capture the systematic aspects of these forms of social and labor market diversity.

Considering all of these issues together, we have little reason to put much faith in the numbers that purport to represent a statistical value of life based on wage compensation for risk. This impression is only reinforced by the recognition that risk of fatal injury is just one of many aspects of work that affects the well-being of workers and, according to the simplest version of supply-and-demand theory, ought to generate compensating wage differentials; others include the stability of employment, workplace autonomy, the degree of general work comfort or discomfort, opportunities for training and advancement, provision of sick leave, employer-provided child care, flexibility in scheduling, and commuting distance. For several decades economists have searched for these wage differentials with little success, as indicated by the very title of one study, “The Pervasive Absence of Compensating Differentials” (Bonhomme and Jolivet 2009). A recent paper, for instance, asserts, “While the theory on the relationship between job characteristics and wages is clear…the empirical literature documenting the existence and magnitude of such tradeoffs has lagged behind, often finding the opposite relationship that theory would predict” (Maestas et al. 2018) 17. And indeed, their carefully conducted investigation into the value workers place on nine job attributes finds that, rather than offsetting wage inequality, as it would if less-desirable work paid more, bringing them into a combined wages-and-amenities measure of compensation actually increases it. Thus, the VSL literature, with all its red flags, stands out as the only corner of the compensating differential literature in which researchers claim to have regularly found that, all else equal, disagreeable work pays more.

But for now let us put all these doubts temporarily to the side and examine the results the VSL researchers claim to have uncovered. Why do this? Even if the absolute levels of compensation estimated by Viscusi et al. are unreliable, it is possible the patterns in their results may hold some meaning. This would be the case if the various confounding factors we reviewed apply with more or less equal force to each subsample of the workforce; greater or lesser wage compensation for risk across them might be discernible in the findings despite the misestimation they all share. Of course, it is also possible that the confounding factors don’t apply equally, so the patterns we will describe have to be taken with an extra dollop of caution.

What we will find, if these studies are to be believed, is that the relative position of different categories of workers in the social hierarchy are reflected not only in absolute wage and risk differentials, as we saw in the previous section, but also in the extent to which higher risk is compensated by higher wages. It turns out that the pattern of estimated compensating differentials corresponds to what would be expected if workers with less bargaining power are not fully compensated, or at least receive less compensation for comparable risks. Perhaps the place to start is with a rather basic study published by Kip Viscusi in 2013 to demonstrate the usefulness of the BLS’s Census of Fatal Occupational Injuries (Viscusi 2013). Using aggregate worker hours as the denominator of the fatality variable and regressing on the log of wages, and using the results for the average wage in his sample, he estimated a VSL of $9.9 million. This is a rather typical result in his oft-repeated wage-risk regressions, uncomplicated by the various adjustments and disaggregations he has sometimes made to his procedure, so it can serve as a baseline. Note that Viscusi interprets this sum as fully compensating the worker who receives it—an inference based on the theoretical model we have already considered, but, as we have seen, unwarranted if the model is adjusted for greater realism. What interests us in this context, however, is the effect of altering the regression model to accommodate a range of risk coefficients across the workforce or applying the same model to different subsamples of workers. To repeat, while we have ample reason to be skeptical of any particular result using this modeling approach, there may still be interest in the pattern of relative results that emerges from it.

Here we review studies of compensating differentials that have disaggregated the workforce into subsamples by wage levels, race, immigrant status, and unionization:

Position in the wage hierarchy. The econometric technique used to estimate wage compensation for risk already incorporates the worker’s wage level, since the dependent variable is the log of wages; this means the coefficient on the risk of fatal injury corresponds to a percentage increase or decrease in what the worker is paid. The VSL calculation converts this to a monetary amount by applying the percentage change to a particular wage within the distribution, such as the median or mean. But there is no rule that says there can be only one such VSL; on the contrary, the percent increment (or decrement) can be taken at multiple points along the wage spectrum. Doing so will produce a range of estimated values, but the elasticity of the VSL with respect to wages will remain constant, since the same percent change is used in each case.

In order to compute VSLs for different wage groups in a way that allows this elasticity to vary, Kniesner and his coauthors conducted a wage-risk regression using panel data; the data incorporated changes in wages and attributed risk levels for the same workers over time as they moved between jobs. (Since few workers change jobs in this fashion over a short multiyear period, their sample was only a bit over 2,000 unique individuals, rather than 126,000 as before.) Thus, since the methodology assumes worker utility is unchanged across moves—there are no good or bad jobs for workers of a given “sort”—there is a sort of compensating wage differential for each of these individuals. Establishing any sort of gradient over the entire sample, however, requires making some simplifying adjustments that modify the independence of these wage-risk relationships. There is no single best way to do this, and Kniesner, Viscusi, and Ziliak (2010) provide results for a range of such “tunings.” Taking one such choice, where the tuning parameter is set at 1, the coefficient on the risk variable, which measures the percentage effect of a change in fatal risk on worker pay, is almost 50% higher for those in the top fourth of the wage distribution compared to those in the bottom fourth. Thus, not only are lower-wage workers estimated to get a smaller pay boost from having to cope with work hazards, they also get a much smaller percentage increase, despite facing far greater risk, as the previous section documented.18 (Kniesner, Viscusi, and Ziliak).

This result is interpreted by the authors in psychological terms: Occupational safety is described by Kniesner et al. as a “normal good” whose value to the worker increases with income. It could just as well be understood in terms of the foundational observation of Lester and those who followed him that there are better and worse jobs for workers with the same skills and personal attributes. A bad job is one with relatively lower pay and higher risk, all else being equal. As long as job segments overlap in the labor market, it follows that at any level of the wage hierarchy higher pay means a higher likelihood of having a good job. One of the characteristics of a good job is that it compensates workers for bearing extra risk. It’s not necessarily that higher-paid workers care more about safety; they are treated as if they do compared to those earning less, a consequence of their greater market power and in contrast to the relative disempowerment experienced by lower-wage workers in bad jobs.

Race. A 2003 study by Viscusi focused specifically on differences between white and Black workers in the degree of risk they face at work and the amount of wage compensation they are estimated to receive for it. He used CFOI data for fatal injuries and SOII data for nonfatal, assigning to both white and Black workers the average risk for the industry in which they were employed.19 (Risk categories were based only on industries, not occupations, since many industry-occupation cells would be too small if disaggregated by race.) Not surprisingly, he found that Blacks are relatively more concentrated in the most dangerous industries and receive lower pay overall. In addition, however, he estimated substantially lower compensating differentials for Blacks: Their coefficient on fatal risk, which measures the percentage effect on their earnings from a unit change in the average risk of their industry, was not quite 60% that of whites, while the coefficient on nonfatal risk was just under two-thirds. Note that these smaller percentage effects apply to the lower wages Black workers already face.

If valid, these results are important for many reasons. They indicate that racial inequality in pay is not offset but exacerbated by un- or insufficiently compensated risk of injury and death. They offer further support for our view that the extent of OSH risks and possible compensation for them are strongly influenced by the relative power a particular group of workers can draw in bargaining and at the workplace. Finally, they undermine the case for the freedom-of-contract view as it might apply to any set of workers. According to these regression results, Black workers receive less compensation than whites for the risks to which they are exposed. If we agree it makes little sense to argue that Blacks are somehow less aware of or concerned by these risks, this reduced compensation shouldn’t be interpreted as evidence for a lower Black VSL. In that case, however, why should the larger coefficients on risk for white workers be viewed as the basis for their VSL? Indeed, faced with these racial disparities, one is forced to either accept both the VSL (freedom-of-contract) interpretation of them and a racially pejorative view of relative attitudes toward risk or, conversely, reject both of them, as we do.20

Immigrant status. This was explored in a 2010 paper by Joni Hersch and Kip Viscusi, and then in a followup 2020 study by Viscusi and Nick Marquiss.21 Since the second paper largely follows the methodology of the first and updates it, we will focus on the most recent version. The main innovation in both papers was to add to the standard wage-risk regression additional terms that interact the fatality risk variable with binary variables indicating whether the worker in question is an immigrant either in general or from Mexico in particular.22 (Hersch and Viscusi also interact with indicators for other regions.) This means that native workers have a coefficient on their estimated fatal risk, while immigrants have that same coefficient plus the coefficient on their interaction term. In their published specification, for example, the baseline coefficient on risk was 0.0029 for 2007, while the coefficient on the interaction between risk and being an immigrant from Mexico was -0.0075, so the overall coefficient for native workers remained at 0.0029 but was -0.0046 for Mexican immigrants and 0.0013 for non-Mexican immigrants. In Table 4 we give a selection from their results.23

Table 4

What is striking in this exercise is that the combined coefficient was strongly negative for immigrants in three of the four cases (two types of immigrants, for 2007 and 2015 each) and was essentially zero for the other case. In addition, when Hersch and Viscusi in the earlier paper included a variable for English fluency, they found that non-English-speaking immigrants from Mexico received less—which is to say more negative—wage compensation for risk than those with English language skills.24 A negative compensating differential, if you accept the interpretation Viscusi and his coathors give to coefficients on risk, would somehow imply that immigrants have a positive appetite for risk of death, a bizarre claim not supported by any independent assessments, such as survey questionnaires. Such results, however, are broadly consistent with the perspective proposed in this paper that those with less workplace power have difficulty obtaining compensation for risks. Of course, care should be taken in drawing any conclusions. It is noteworthy that the risk variable in both studies assigned fatal risk on the basis of industry only, again because of the difficulty in populating a large number of industry-by-occupation cells. Moreover, all the other caveats we examined regarding this stream of research apply to these papers as well.

Unionization. The relationship between unionization and wage compensation for risk is complex. From a bargaining power perspective, the effect ought to be clear: In general, being a member of a union or covered under a collective bargaining agreement should give workers a greater ability to demand extra pay if there are extra risks of injury or illness. On the other hand, unions often resist resolving OSH issues through hazard pay, demanding instead employer adherence to more stringent safety standards or the establishment of a joint health and safety committee; that is, unions attempt to lower the risks. Moreover, the likelihood of a union being established is itself related to the presence of OSH hazards, either in the past or currently. And unions vary greatly among themselves, of course, in the amount of attention they give to OSH risks and the leverage they provide workers for addressing them.

That said, one might suppose that, compared to the situation workers find themselves in when unions are absent, collective bargaining ought to increase the pressure on employers to compensate workers for risk exposure. This was the general pattern found in the most influential meta-analysis of VSL studies, which Viscusi and Aldy published in 2003:

Regardless of estimation strategy, most assessments of the U.S. labor market found higher risk premiums for union workers than for non-union workers….Of the ten U.S. labor market value of life studies we reviewed that evaluated the role of unions in risk premiums, nine found union workers enjoyed greater compensating differentials for bearing risk than nonunion workers. In contrast to [our] accepted theory, several of these papers found that non-union workers had insignificant or statistically significant negative compensating differentials for risk. (Viscusi and Aldy 2003, 44, with our clarification)

It is important to note, however, that this summary is now nearly two decades old, and no published studies have looked at the union effect on compensating differentials in the United States since then.

Overall, the pattern of empirical results in the VSL literature, taken at face value (as if they were correctly estimated), is consistent with the view that wage compensation for hazardous work is not an automatic feature of unregulated labor markets, as proposed by freedom-of-contract theory, but a benefit to be achieved according to the influence or power workers can bring to bear on employers; a higher status in the labor market and more power on the job mean more likelihood of receiving this compensation. It would be too much to regard these statistical results as convincing evidence, however, since the methodological assumptions underlying them are so problematic, and this uncertainty may extend to the pattern as well as the overall direction of results. To take just one example, the exclusion of fatal occupational diseases from risk measures may mean that, on average, industries and occupations deemed safer by VSL researchers might actually be more dangerous. Similarly, the omitted explanatory variables, like measures of capital intensity and market concentration, that characterize this literature might, if included, change the picture of which groups in society are most likely to receive compensation for risk. The most we can say is that the evidence, such as it is, does not contradict the view that the degree to which risk is rewarded with greater pay, as well as every other outcome of the employment relationship, is the product of economic institutions, public policy, and the relative power each side brings to the table.

When we pull back and consider VSL methodology as a whole, what stands out is its focus on presumed commonalities that cut across different strata of the labor force. In the environmental, health, and other policy arenas that employ monetary measures of the value of life-saving, the demand is for a single value for all lives, or at most a small set of values for the specific populations that might be affected by a policy choice.25 Not surprisingly then, the statistical procedure we have surveyed attempts to pick out percent changes in wages corresponding to changes in fatal occupational risk averaged over broad swaths of the workforce. Workers differ in the risk they face, but the coefficient γ on the risk variable r applies to all of them. At the end of our survey we looked at a handful of studies that disaggregated by wage levels, race, and immigrant status and found these groups receive less compensation for the risks they face, which is especially problematic since these groups face the most risks. We also found indications that workers with a union are able to increase the compensation paid for risks. Still, even these estimates average effects over very large portions of the population. One is reminded of the story about two statisticians who went deer hunting. One took aim but fired two feet to the left. The other then fired two feet to the right. “Congratulations,” they said to each other, “we got it!”

From a realistic labor market perspective it makes no more sense to ask what is “the” compensating differential for risk than to ask what is “the” level of pay or number of hours in the work week; compensation for risk is just one more dimension along which labor market experience varies. It is surely the case that some workers are amply rewarded for taking on additional risk, others rewarded somewhat but not fully, and still others not at all or even penalized, in the sense that they are stuck in bad jobs that are worse in both respects. Thus, the presumption that estimated compensating wage differentials always represent full compensation for risk is unsupportable. What ought to interest us is the variation of compensating differentials: who tends to get how much compensation and why. Answering that question would call for entirely different study designs, drawing on both the analysis of large data sets and case studies of specific work contexts to capture, among other factors of interest, differences in labor market and workplace power. In this context it is worth noting that case study research is nearly absent from the VSL literature.

VI. Occupational safety and health in the era of the coronavirus: Challenges and policy

Years can go by with little public notice given to issues in occupational safety and health, at least in higher-income countries. But the advent of the Covid-19 pandemic has moved OSH to front pages and Twitter feeds, since the risk of infection has made work suddenly more dangerous, and in some communities workplaces are the most important sites of transmission. In this as in other aspects of our society, the virus has brought once-hidden flaws out into full view.

Before we can discuss the role of the workplace in the pandemic, however, it will be useful to sum up the main conclusions from the preceding analysis.

• Despite the Occupational Safety and Health Act and the other programs created over the past century, poor working conditions remain a concern. Workers exposed to excessive risk of accidents and illnesses face not only a health but also an economic problem, due to uncompensated medical costs and the loss of current and future income. The network of state-level workers’ compensation programs provides benefits to too few of these workers and provides too little income support.
• OSH reporting systems are incomplete and inaccurate. While the Bureau of Labor Statistics now provides a comprehensive census of fatal occupational injuries, nonfatal injuries are often missed and most occupational diseases pass completely beneath the radar. This is a problem at the national level, since it limits the guidance we need for policy and implementation; it is also a problem at the enterprise level, where workers remain uninformed about the true risks they are exposed to.
• While there is sometimes economic and ethical justification for hazard pay, such payments are rarely offered explicitly, and there is little reason to believe they are commonly implicit in wage offers either, despite the claims of some researchers. This is not to say no workers ever receive compensating wage differentials for risk, but evidence for the view that such differentials are widespread does not survive scrutiny. Moreover, there is no reason to assume that when extra wages are offered the differential fully compensates workers for the risks they bear. For these reasons, it is likely that the existence and extent of hazard pay in any form, explicit or implicit, reflect the relative power workers are able to draw on to advance their interests.
• The view of the labor market associated with the freedom-of-contract perspective, which holds that OSH risks are efficiently negotiated between workers and employers, is at odds with nearly everything we know about how labor markets really work. It cannot accommodate the reality of good and bad jobs, workplace authority based on the threat of dismissal, discrimination, and the pervasive role of public regulation in defining what employment entails and what obligations it imposes. It also fails to acknowledge the social and psychological dimensions of work, which are particularly important in understanding how people perceive and respond to risk.

The current pandemic has thrown all of these observations into vivid relief. In the remaining portion of this paper we will show how abiding economic, social, and public health aspects of OSH have been magnified under the impact of the coronavirus. We will conclude by briefly sketching some of the measures that would mitigate the impacts on workers, their families, and their communities. The place to begin is with the recognition that the pandemic is, to a significant extent, a crisis in occupational safety and health. Transmission occurs mainly in social settings when uninfected people come into proximity with others who are shedding the virus; this can occur at home, in school, and while shopping, traveling, and recreating. But another prime venue is the workplace. Those at greatest risk are often either classified as “essential” workers or unable for economic reasons to avoid going to work no matter how great the risk.26

Particular work environments have been singled out as coronavirus hotspots. Health care facilities, of course, belong in this category, although extra effort has gone into measures like frequent testing, isolation, and provision of personal protective equipment. At the opposite extreme, meatpacking plants are by nature conducive to spreading, with their crowded production lines and cool temperatures, but many companies have failed to address the risk (Narea 2020; Schlosser 2020; Swanson, Yaffe-Bellany, and Corkery 2020; Thompson 2020). 27 Instead they have often used their political clout to prevent public health officials and journalists from learning the full extent of the problem (Corkery 2020; Foley 2020; Grabell, Perlman, and Yeung 2020; NC Watchdog Reporting Network 2020). A recent report from the Centers for Disease Control and Prevention (CDC) tallies over 16,000 confirmed cases and 86 deaths nationwide in meat and poultry facilities, but a brief concluding statement on “limitations” makes it clear their partial, self-reported numbers are likely to be a substantial underestimate (CDC 2020).28

Food production facilities in general have been recognized as spreading venues. Workers in these jobs are more likely to be immigrants, Latinxs, or both, and half again as likely to earn less than twice the official poverty cutoff—$12,760 for an individual and $26,200 for a family of four—making them more likely to keep coming to work even if they are infected (Artiga and Rae 2020). Of course, low-income, minority, and immigrant workers are also likely to come home to more crowded living conditions while also having fewer options to avoid contact in shopping and other off-work activities. The result is that each infection acquired at work means more transmission in the community. For instance, when it is reported that confirmed cases in Nebraska’s meatpacking plants constitute a fifth of the state’s total (Goodwin 2020), it should be taken as a lower bound for the contribution of this one sector to the state’s caseload. While the actual contribution is unknown, with plausible estimates of R0 (the transmission rate) for these workers, and recognizing some infections were transmitted to rather than from the workplace, it could be as large as half.

When we add the many other sectors potentially conducive to spreading—retail, in-person care, transportation, etc.—it becomes clear how large a role the workplace plays in the pandemic. Among the inadequacies exposed are:

• Insufficient employer protection. There is substantial evidence of worker discontent with the measures employers have, or have not, taken to reduce the risk of infection. To take one example, over 17,800 complaints have been filed by workers to the Los Angeles County Department of Public Health alone (Roosevelt, Martin, and Avery 2020). The Payday Report website documents over 900 wildcat job actions nationwide over pandemic-related issues (Payday Report 2020). Polls show a minority of workers working outside the home are dissatisfied with precautions taken by their employers (17% in the Washington Post-Ipsos Coronavirus Employment Survey, 8% in an Economic Policy Institute poll), but the samples are drawn from all occupations and industries and may significantly understate the level of concern in the most affected workplaces (Dorman and Mishel 2020; Washington Post-Ipsos 2020). It is also possible that, while many workers fear being exposed to the virus at work, they don’t hold their employers responsible for this risk.
• Regulatory failure. The primary regulatory agencies for OSH in the United States, the U.S. Occupational Safety and Health Administration and the state OSHAs that take on responsibility for regulation in the 25 states that have assumed this role, have largely been silent bystanders. As of July 21, 2020, OSHA had received over 7,000 Covid-19 complaints but issued citations to only two workplaces. In both cases at least six workers had been hospitalized before the citations were issued (OSHA 2020; Department of Labor 2020). Incredibly, a recent ruling from OSHA effectively absolves employers of responsibility for reporting any occupational exposures from Covid resulting in hospitalization—this despite the general obligation on employers to log and report occupational events resulting in hospitalization. In its September 30, 2020, guidance, OSHA interpreted the statute as requiring notification only if the hospitalization occurs within 24 hours of the worker’s initial exposure to the virus, virtually an impossibility in light of the delayed onset of Covid symptoms (Edwards 2020; Moyer 2020; Jamieson 2020). Above all, OSHA has not translated CDC guidelines into enforceable mandates for employers in line with its core responsibilities (Covert 2020; Michaels and Wagner 2020).
• Surveillance failure. As noted just above, employer resistance and weak public health authority, compounded by drastic shortfalls in testing capacity, have rendered workplace infection numbers unreliable. Although the workplace is a prime venue for coronavirus transmission, our knowledge of where the risks are concentrated is inadequate.
• Economic coercion. Low-income workers are afraid to stay home from work, even if they or another household member is infected; many are also afraid to criticize employers for policies that penalize workers for putting health and family obligations first. In one poll conducted by the Roosevelt Institute, over 40% of all workers who said they would be unable to pay all their bills if they took time off from work either largely or strongly agreed they would continue going to work if they came down with a fever (Hertel-Fernandez et al. 2020). In the same poll, only a third were sure they would receive paid sick leave if they stayed home due to the virus; this drops to about a quarter for the bottom half of the wage distribution. To some extent, the intimidation of low-wage workers was reduced by the additional unemployment insurance payments voted by Congress, but this program is set to lapse, and in any case laid-off or dismissed workers would also lose health insurance coverage and any other nonwage benefits from work, with no guarantee that another job would be available when unemployment insurance benefits expire.
• Noncoverage by workers’ compensation (WC). Although the WC system is state-based and varies according to jurisdiction, as we have seen there is little coverage of occupational disease anywhere. Few workers or their survivors have received benefits for contracting Covid-19 during the pandemic, and where they have the payouts have been inadequate (Bailey and Jewett 2020). Of the five states that have taken action to make their WC programs substantially more responsive to the ongoing pandemic, three have done so by executive order and two by legislative enactment; of these, two are for all workers and three for essential workers only (Cunningham 2020; National Council on Compensation Insurance 2020).29
• Limited financial compensation for risk. At the outset of the pandemic there was much publicity about wage boosts given the “heroes” working in essential positions. Indeed, some workers received hazard pay, as evidenced by the Economic Policy Institute survey that found 30% of all workers working outside their homes receiving a wage increase. This is corroborated by the previously cited Roosevelt Institute study in which less than half of all workers received a pay boost, with the average increase coming to just over a dollar per hour (Hertel-Fernandez et al. 2020). But a team of researchers using a real-time human resources management database also found that a 10th of all workers had their wages cut between the beginning of the pandemic and the end of May (Cajner et al. 2020). Later, from the vantage point of eight months into the publicly visible phase of the pandemic, a team of researchers at the Brookings Institution found that hazard pay in the retail sector was short-lived and dwarfed by increased profits. Examining data for the 13 biggest retailers whose markets grew during the pandemic and for which financial information has been reported, they found, while a few instituted permanent wage increases, none sustained hazard pay throughout the period, and (after dropping one outlier) the average date for eliminating this premium was the end of May. For the six of these firms reporting wage data, outlays for hazard pay were just 38% of the increase in profits during the pandemic (Kinder, Stateler, and Du 2020).
• Unequal burdens. Every study of the pandemic to date that has looked at the issue has found higher infection rates among Blacks, Latinxs, Native Americans, and immigrants. These disparities can be found in polls, studies, and reportage of workplace exposures specifically, some of which have already been referenced above. To take one example, the Roosevelt Institute poll found that, while 23% of white workers reported they were “extremely concerned about being infected at work,” the numbers for Latinx and Black workers were 33% and 42%, respectively (Hertel-Fernandez et al. 2020). Of course, this is not about the relative attraction of the virus to people of these different backgrounds; it reflects the greater concentration of minority groups in riskier jobs with less employer protection, subject to heightened economic coercion from low income and fewer benefits.

These are readily observable aspects of the pandemic as an occupational health crisis. Our findings in this paper, however, direct us to several underlying factors that help explain why these harsh outcomes have occurred.

• The extent of the low-wage economy. Decades of increasing wage dispersion have left the U.S. with a substantial portion of the labor force whose pay is at or approaching the poverty level even during times of economic expansion. Such workers are subject to economic coercion in all significant life decisions, including what jobs to accept and when to put family or similar obligations temporarily above work attendance.
• Limited employer and government benefits. Paid sick and family leave are privileges for the upper strata of the labor force, not rights all workers can rely on. Health insurance is also largely employment-based, with only partial supplementation through the Affordable Care Act. The state unemployment insurance systems are designed primarily to make it difficult to claim and to not reach as many of the jobless as possible, and they generally replace less than half the worker’s prior earnings.
• Deunionization. Unions help inform workers about the risks they face and use the lever of collective action to improve conditions at work. At the same time, they diminish economic coercion by giving workers procedural rights on the job, yet the share of the private-sector labor force covered under a collective bargaining contract has fallen to just 7%.
• Lax regulation. The federal agencies responsible for regulating working conditions, the Occupational Safety and Health Administration and the Department of Labor’s Wage and Hour Division, have been weakened by prolonged underfunding and little policy impetus from political leadership. OSHA in particular was completely unprepared to meet the challenge of the pandemic.
• Marginalization of minority, immigrant, and low-income communities. Social hierarchies corresponding to what this paper has called power differentials largely structure the country’s economy and politics. Workers concentrated in the worst jobs also tend to live in crowded residences, sequestered in neighborhoods with little political influence. Policies like sheltering in place and commercial reopening are chosen with little input from or attention to the needs of these communities.

With these considerations in mind, we can suggest several lines of reform that not only address the current pandemic but have the potential to set in motion a more fundamental reshaping of public health and labor market outcomes for the post-pandemic world.

• Workplace measures for impeding the spread of the novel coronavirus should be mandated by an emergency temporary OSHA standard. No new statutory language is needed. Indeed, the 25 states with OSHA state plans can take this step on their own, since the law requires them to meet federal standards but leaves them free to exceed them. Adequate guidance has already been issued by the CDC; there is no reason for either OSHA or the states to delay action.

Legislative action at the federal or state level could mandate the establishment of joint worker–management health and safety committees with jurisdiction over protection measures against the virus. These could take advantage of workers’ on-the-ground knowledge in translating OSHA standards into detailed workplace actions while promoting an atmosphere of public health collaboration (Lichtenstein 2020).

• There should be a default presumption of work-relatedness for coronavirus infection for workers’ compensation eligibility, applicable to all workplaces and not only those categorized as essential. This would conform to the practice in Germany (among other countries), where, early in the pandemic, the Labor Ministry identified Covid-19 as a potential occupational disease and stipulated that employer responsibility could be voided only by strict adherence to specified distancing and personal protective measures (Bundesministerium für Arbeit und Soziales 2020). Federal and state OSHAs, coordinating with public health authorities, should establish mandatory record-keeping for all employee infections, to be transmitted to the relevant agencies and open to public inspection. There should be no more uncertainty about workplace hotspots, either for the workers directly at risk or members of the public concerned about community health conditions.
• Adequate paid sick leave for any employee who is infected by the coronavirus and family leave for those with pandemic-related obligations in their household should be universal and mandatory. During the transition to such a system the federal government can assume part of the financing burden.
• While avoidable infection risks should be minimized through the enforcement of science-based standards, there will still be occupations more exposed to the pandemic, like health care, in-person education and other services, and some retail employment. Workers facing these residual risks should be identified and awarded extra compensation—hazard pay—as a matter of fairness. These extra wage payments should supplement and not replace existing wages. As with paid leave benefits, hazard pay can be partially financed by the government to facilitate implementation.

In addition to the immediate actions suggested above, we need a fundamental redirection of labor and public health policies so future challenges don’t find us as unprepared as the current pandemic has. These include:

• shrinking and eventually eliminating the low-wage sector through more comprehensive and higher-targeted minimum wage requirements along with training and other productivity-enhancing measures;
• revamping labor laws to encourage unionization under centralized systems of bargaining;
• rebuilding the occupational and public health infrastructure with more funding and a broader scope for standard-setting and regulation;
• taking assertive action to combat discrimination on the basis of race, ethnicity, and immigrant status, the most salient forms of bias in occupational health outcomes; and
• embarking on initiatives to enhance worker voice in the 21st-century economy, such as extending rights to workers in online platform or otherwise “fissured” occupations and promoting new institutions for worker participation at the worksite and top management levels.

It is beyond the scope of this paper to do more than enumerate the items on this agenda, but they are logical consequences of the perspective on labor markets we have developed in the preceding pages.

About the authors

Peter Dorman is an emeritus professor of political economy at Evergreen State College in Olympia, Wash. The author of Markets and Mortality: Economics, Dangerous Work and the Value of Human Life and a forthcoming book on the economics of climate change, he has worked extensively for the International Labor Organization on occupational safety and health, child labor, and HIV/AIDS, and has written numerous articles and book chapters on topics in economics and political economy. He holds a PhD in economics from the University of Massachusetts, Amherst.

Les Boden is a professor of public health at Boston University School of Public Health. Much of his research has focused on the economic and human consequences of occupational injuries and illnesses. He has also written on workers’ compensation, occupational safety and health regulation, gender and racial disparities, and the legal and public health use of scientific information. He has a Ph.D. in economics from MIT.

Acknowledgments

The authors are grateful for the extensive assistance they received from Lawrence Mishel and Melat Kassa of EPI, our copy editor Patrick Watson, and our reviewers Andy Garin, Kevin Lang, and J. Paul Leigh. Any remaining shortcomings are ours and not theirs.

Endnotes

1. This group includes W. Kip Viscusi, who will be discussed shortly, and his various coauthors, as well as legal theorists like Cass Sunstein, who has written extensively on the use of compensating wage differentials for the purpose of valuing statistical lives (VSL). An overview of Sunstein’s views, which address complexities in the application of VSL findings but largely accept the labor market analyses on which they are based, can be found in Sunstein (2014). The group also includes policy analysts like John Graham, who served as the director of the Office of Information and Regulatory Affairs of the Office of Management and Budget—the so-called regulatory czar—under the George W. Bush administration. This position was occupied by Sunstein after the election of Barack Obama.

2. Lester pointed out that the existence of involuntary unemployment is another anomaly supply-and-demand theory cannot account for. Although it remains germane, we will not pursue that critique here.

3. A third economist, Peter Diamond, shared the reward for his work on search in nonlabor contexts.

4. For a relatively recent example, see Bartling, Fehr, and Schmidt 2012.

5. The literature on discrimination in labor markets is far too large to be distilled even into a few reviews. For two influential examples, see Lang and Lehmann 2012 and Bertrand 2011.

6. It might be tempting to reclaim subordination for markets—treating it as a disutility of employment for which a compensating differential would need to be paid—along the lines of the VSL literature and as Nozick (1974) attempted to do with alienation. This fails, however, because the alternative to greater subordination is more delegation of authority to the workforce, and economic theory is clear that, to work, such a strategy requires greater pay, not less. The role of wages in self-regulation and intrinsic motivation is modeled under different conditions by efficiency wage and gift exchange theory (Akerlof 1984).

7. The relationship between safety performance and the degree of worker subordination is not uniform. Greater worker autonomy, for example in controlling the pace of work, is often conducive to safety. On the other hand, greater managerial supervision of work behavior, such as requiring personal protective equipment, can also be risk-reducing if it is motivated by safety concerns.

8. Clearly, if workers cannot determine the probability of having an occupational illness, this probability cannot be reflected in their wage demands. As a result, estimates of the value of a statistical life based on their wages will not be meaningful.

9. The main point of their paper is to demonstrate that safety regulation can improve efficiency if workers differ in their risk preferences, due to externalities resulting from statistical discrimination by employers during search. This does not bear on the argument developed here.

10. See for instance the essays collected in Slovic 2000.

11. Dorman 1996 provides a simple model in which the salience of cognitive dissonance avoidance in worker risk perception depends inversely on the degree of agency experienced at work.

12. Their important conclusion: “inferences should be confined to the level of observation” (p. 902).

13. The CFOI website shows the distribution of vehicular and homicide deaths across occupations but doesn’t provide denominator data.

14. Because the limited suite of observations on age, education, experience, and marital status is not sufficient to capture many employer-relevant differences in qualification and skill, VSL studies routinely exclude white-collar or managerial occupations from their sample. Of course, these filters bias their findings toward a more positive coefficient on risk, since the excluded occupations generally feature both higher wages and lower fatal accident rates. Such sample pruning is problematic, however, since individuals often move between blue- and white-collar jobs, and the potential for promotion to lower-level management is also an important incentive for line workers.

15. This result somehow did not make its way into the widely cited meta-analysis by Viscusi and Aldy (2003), although they did incorporate the larger positive coefficient on risk found by Dorman and Hagstrom, when, for the purpose of comparison, industry-level control variables were excluded.

16. There is some confusion on this score, because VSL studies commonly employ industry and occupational dummies, and advocates like Viscusi claim this procedure picks up non-OSH factors associated with where and how workers are employed. Such a claim is false, however, since the statistical effects of including variables like average capital-labor ratios and percent female are entirely different than those of binary categorical indicators—as they also are for average risk measures, which don’t duplicate the information conveyed by dummies.

17. By “theory,” the authors mean the simple supply-and-demand framework reflecting a freedom-of-contract view of the labor market.

18. A similar result was found in Evans and Schaur 2010: The bottom quarter of the income distribution, aggregated across age differences, received minimal if any wage compensation for risk, while compensation was substantial for the upper quarter. Their purely cross-sectional wage and risk data were less suitable for quantile regression, however, with no opportunity for identifying the worker’s occupation, compared to Kniesner,Viscusi, and Ziliak 2010.

19. Viscusi does not explain his decision to not take advantage of the breakdowns by race in these data sources. It is possible the data were not reported by race crossed with industry, and he didn’t have access to the detailed records to construct these crosses himself. Incidentally, calculations based on SOII should be taken with an extra grain of salt since, as we saw in the previous section, it is an incomplete and biased accounting of traumatic injuries.

20. In his 2003 paper, Viscusi too rejects the VSL interpretation of racial disparities in estimated risk compensation, arguing they are due to employer discrimination. It is not possible, however, to simultaneously acknowledge that white and Black workers typically are employed side-by-side in the same jobs, reject racial differences in risk preferences, and yet maintain the interpretation of risk coefficients as revealing both worker preferences and employer costs in the face of these disparate results. While we have criticized the standard VSL model, which makes no provision for different coefficients for demographic or identity groups, it is derived from a logically consistent model of a labor market in equilibrium (Rosen 1974). No such model can generate a VSL depiction of racially disparate risk coefficients without unacceptable assumptions about race and risk preference or getting mired in conundrums resulting from employers engaging in discrimination in hiring and pay while also being held to an iso-profit frontier in safety provision.

21. Viscusi and Marquiss (2020) examine other issues in addition to the wage-risk regressions discussed here, but we will not take them up.

22. They also allow the intercept of the estimated wage equation to shift by including a term with just the immigration dummy, uninteracted.

23. Adjustment for clustering was not reported in their results, and this may be consequential, since adjusting for clustering—mandatory because the risk variable is assigned the same value for all workers in each industry—rendered most of the corresponding coefficients “insignificant” in Hersch and Viscui 2010.

24. It is not clear why Viscusi and Marquiss (2020) did not report a comparable regression specification, since their recommendations all center on communication issues due to different native languages.

25. In some contexts, the demand is for the value of life-years saved rather than lives as such.

26. For just one of many examples, see Hubler et al. 2020.

27. Note that the increased risk of transmission in these facilities is associated with cool temperatures produced by air conditioning and doesn’t imply that risk of aerosol transmission is associated with lower temperatures per se.

28. See the disclaimer on p. 5 of the report.

29. We exclude reforms that identify Covid-19 only as a potential occupational disease without altering the burden of proof for establishing work-relatedness; address only health care workers, first responders, and public employees; or have provisions that have already expired.

References

Akerlof, George A. 1984. “Gift Exchange and Efficiency-Wage Theory: Four Views.” American Economic Review 74, no. 2: 79–83.

Akerlof, George A., and William T. Dickens. 1982. “The Economic Consequences of Cognitive Dissonance.” American Economic Review 72, no. 3: 307–319.

Appelbaum, Eileen. 2017. “What’s Behind the Increase in Inequality?” Center for Economic Policy Research.

Applebaum, Katie M., Abay Asfaw, Paul K. O’Leary, Andrew Busey, Yorghos Tripodis, and Leslie I. Boden. 2019. “Suicide and Drug-Related Mortality Following Occupational Injury.” American Journal of Industrial Medicine 62, no. 9: 733–741.

Artiga, Samantha, and Matthew Rae. 2020. “The COVID-19 Outbreak and Food Production Workers: Who Is at Risk?” Kaiser Family Foundation Issue Brief.

Asfaw, Abay. 2014. “Disparities in Access to Health Insurance and Workers’ Compensation Benefit Between Non-Contingent and Contingent Farm Workers in U.S. Agriculture.” Journal of Health Disparities Research and Practice 7, no. 3: Article 6.

Asfaw, Abay, and Leslie I. Boden. 2020. “Impact of Workplace Injury on Opioid Dependence, Abuse, Illicit Use, and Overdose: A 36-Month Retrospective Study of Insurance Claims.” Occupational and Environmental Medicine 77, no. 9: 648–654.

Asfaw, Abay, and Kerry Souza. 2012. “Incidence and Cost of Depression After Occupational Injury.” Journal of Occupational and Environmental Medicine 54, no. 9: 1086–1091.

Atiyah, P.S. 1979. The Rise and Fall of Freedom of Contract. Oxford, England: Oxford University Press.

Azaroff, Lenore S., Charles Levenstein, and David H. Wegman. 2002. “Occupational Injury and Illness Surveillance: Conceptual Filters Explain Underreporting.” American Journal of Public Health 92, no. 9: 1421–1429.

Bailey, Melissa, and Christina Jewett. 2020. “Families of Health Workers Killed by COVID Fight for Denied Workers’ Comp Benefits.” Kaiser Health News, July 13.

Bale, Anthony. 1988. “Assuming the Risks: Occupational Disease in the Years Before Workers’ Compensation.” American Journal of Industrial Medicine 13: 499–514.

Baron, Sherry L., Andrea L. Steege, Suzanne M. Marsh, Cammie Chaumont Menendez, and John R. Myers. 2013. “Nonfatal Work-Related Injuries and Illnesses—United States, 2010.” CDC Health Disparities and Inequalities Report—United States 62, no. 3: 35–40.

Bartling, Björn, Ernst Fehr, and Holger Herz. 2014. “The Intrinsic Value of Decision Rights.” Econometrica 82, no. 6: 2005–2039.

Bartling, Björn, Ernst Fehr, and Klaus M. Schmidt. 2012. “Screening, Competition, and Job Design: Economic Origins of Good Jobs.” American Economic Review 102, no. 2: 834–864.

Berdahl, Terceira A., and Julia McQuillan. 2008. “Occupational Racial Composition and Nonfatal Work Injuries.” Social Problems 55, no. 4: 549–572.

Bertrand, Marianne. 2011. “New Perspectives on Gender.” In Handbook of Labor Economics, Vol. 4B, edited by O. Ashenfelter and D. Card. Amsterdam, Netherlands: Elsevier.

Board of Governors of the Federal Reserve System. 2019. “Report on the Economic Well-Being of U.S. Households in 2018.” May.

Boden, Leslie I., Nicole Nestoriak, and Brooks Pierce. 2010. “Using Capture-Recapture Analysis to Identify Factors Associated with Differential Reporting of Workplace Injuries and Illnesses.” JSM Proceedings (2010), Section on Survey Research Methods.

Boden, Leslie I., and Al Ozonoff. 2008. “Capture–Recapture Estimates of Nonfatal Workplace Injuries and Illnesses.” Annals of Epidemiology 18, no. 6: 500–506.

Boden, Leslie I., Emily A. Spieler, and Gregory R. Wagner. 2016. “The Changing Structure of Work: Implications for Workplace Health and Safety in the U.S.” Future of Work Symposium, U.S. Department of Labor.

Bonauto, David K., Joyce Z. Fan, Thomas W. Largo, Kenneth D. Rosenman, Mandy K. Green, Jennifer F. Walters, B.L. Materna, J. Flattery, T. St. Louis, L. Yu, S. Fang, L.K. Davis, D.J. Valiante, K.R. Cummings, and J. J. Hellsten. 2010. “Proportion of Workers Who Were Work-Injured and Payment by Workers’ Compensation Systems—10 States, 2007.” MMWR Morbidity and Mortality Weekly Report 59, no. 29: 897–900.

Bonhomme, Stéphane, and Grégory Jolivet. 2009. “The Pervasive Absence of Compensating Differentials.” Journal of Applied Econometrics 24, no. 5: 763–795.

Buchanan, Susan, Pamela Vossenas, Niklas Krause, Joan Moriarty, Eric Frumin, Jo Anna M. Shimek, Franklin Mirer, Peter Orris, and Laura Punnett. 2010. “Occupational Injury Disparities in the U.S. Hotel Industry.” American Journal of Industrial Medicine 53, no. 2: 116–125.

Bundesministerium für Arbeit und Soziales. 2020. “SARS-CoV-2 Occupational Safety and Health Standard” (SARS-CoV-2-Arbeitsschutzstandard). April 16.

Cahuc, Pierre, Stéphane Carcillo, and André Zylberberg. 2014. Labor Economics. Second Edition. Cambridge, Mass.: MIT Press.

Cajner, Tomaz, Leland D. Crane, Ryan A. Decker, John Grigsby, Adrian Hamins-Puertolas, Erik Hurst, Christopher Kurz, and Ahu Yildirmaz. 2020. “The U.S. Labor Market During the Beginning of the Pandemic Recession.” Becker-Friedman Institute Working Paper No. 2020-58.

Card, David, Ana Rute Cardoso, Jörg Heining, and Patrick Kline. 2018. “Firms and Labor Market Inequality: Evidence and Some Theory.” Journal of Labor Economics 36, no. S1: S13–70.

Case, Anne, and Angus Deaton. 2017. “Mortality and Morbidity in the 21st Century.” Brookings Papers on Economic Activity. Spring: 397–476.

Castillo, Carla G. 2018. “What the Doctors Don’t See: Physicians as Gatekeepers, Injured Latino Immigrants, and Workers’ Compensation System.” Anthropology of Work Review 39, no. 2: 94–104.

CDC (Centers for Disease Control and Prevention). 2020. “Update: COVID-19 Among Workers in Meat and Poultry Processing Facilities—United States, April–May 2020.” Morbidity and Mortality Weekly Report, July 7.

CDC/NIOSH (Centers for Disease Control and Prevention/National Institute for Occupational Safety and Health). 2004. Worker Health Chartbook. NIOSH Publication no. 2004-146.

Corkery, Michael. 2020. “As Meatpacking Plants Reopen, Data About Worker Illness Remains Elusive.” New York Times, May 25.

Covert, Bryce. 2020. “How OSHA Went AWOL During the Pandemic.” American Prospect, October 6.

Cummings, Kristin J., and Kathleen Kreiss. 2008. “Contingent Workers and Contingent Health: Risks of a Modern Economy.” Journal of the American Medical Association 299, no. 4: 448–450.

Cunningham, Josh. 2020. “COVID-19: Workers’ Compensation.” National Conference of State Legislatures, July 28.

Dembe, Allard E., J. Bianca Erickson, and Rachel Delbos. 2004. “Predictors of Work-Related Injuries and Illnesses: National Survey Findings.” Journal of Occupational and Environmental Hygiene 1, no. 8: 542–550.

Dersh, Jeffrey, Tom Mayer, Brian R. Theodore, Peter Polatin, and Robert J. Gatchel. 2007. “Do Psychiatric Disorders First Appear Preinjury or Postinjury in Chronic Disabling Occupational Spinal Disorders?” Spine 32, no. 9: 1045–1051.

Dickens, William T., and Kevin Lang. 1993. “Labor Market Segmentation Theory: Reconsidering the Evidence.” In Labor Economics: Problems in Analyzing Labor Markets, edited by William Darity Jr. Dordrecht, Netherlands: Springer.

Dong, Xiuwen, and James W. Platner. 2004. “Occupational Fatalities of Hispanic Construction Workers from 1992 to 2000.” American Journal of Industrial Medicine 45, no. 1: 45–54.

Dorman, Peter. 1996. Markets and Mortality: Economics, Dangerous Work, and the Value of Human Life. Cambridge, England: Cambridge University Press.

Dorman, Peter. 1998. “Compensating Differentials for Dangerous Work in a Labour Discipline Model.” Metroeconomica 49, no. 3: 364–385.

Dorman, Peter, and Paul Hagstrom. 1998. “Wage Compensation for Dangerous Work Revisited.” Industrial and Labor Relations Review 52, no. 1: 116–135.

Dorman, Peter, and Lawrence Mishel. 2020. “A Majority of Workers Are Fearful of Coronavirus Infections at Work, Especially Black, Hispanic, and Low- and Middle-Income Workers.” Economic Policy Institute.

Edwards, Johnny. 2020. “OSHA Eases Reporting Requirements on Worker Infections.” Atlanta Journal-Constitution, October 5.

Evans, Mary F., and Georg Schaur. 2010. “A Quantile Estimation Approach to Identify Income and Age Variation in the Value of a Statistical Life.” Journal of Environmental Economics and Management 59, no. 3: 260–270.

Fan, Z. Joyce, David K. Bonauto, Michael P. Foley, and Barbara A. Silverstein. 2006. “Underreporting of Work-Related Injury or Illness to Workers’ Compensation: Individual and Industry Factors.” Journal of Occupational and Environmental Medicine 48, no. 9: 914–922.

Foley, Michael. 2017. “Factors Underlying Observed Injury Rate Differences Between Temporary Workers and Permanent Peers.” American Journal of Industrial Medicine 60, no. 10: 841–851.

Foley, Michael, John Ruser, Glenn Shor, Harry Shuford, and Eric Sygnatur. 2014. “Contingent Workers: Workers’ Compensation Data Analysis Strategies and Limitations.” American Journal of Industrial Medicine 57, no. 7: 764–775.

Foley, Ryan J. 2020. “Coronavirus Outbreak at Iowa Pork Plant Was Larger than State Reported.” Des Moines Register, July 22.

Frederick, James, and Nancy Lessin. 2000. “Blame the Worker: The Rise of Behavioral-Based Safety Programs.” Multinational Monitor 21, no. 11: 10.

Frey, Bruno S., and Alois Stutzer. 2005. “Beyond Outcomes: Measuring Procedural Utility.” Oxford Economic Papers 57, no. 1: 90–111.

Galizzi, Monica, and Jay L. Zagorsky. 2009. “How Do On-the-Job Injuries and Illnesses Impact Wealth?” Labour Economics 16, no. 1: 26–36.

GAO (Government Accountability Office). 2009. “Enhancing OSHA’s Records Audit Process Could Improve the Accuracy of Worker Injury and Illness Data.” GAO-10-10.

Goh, Joel, Jeffrey Pfeffer, and Stefanos Zenios. 2015. “Exposure to Harmful Workplace Practices Could Account for Inequality in Life Spans Across Different Demographic Groups.” Health Affairs 34, no. 10: 1761–1768.

Goodwin, Connor. 2020. “Nebraska Meatpacking Workers Continue to Struggle During the Pandemic.” American Prospect, August 11.

Grabell, Michael. 2013. “The Expendables: How the Temps Who Power Corporate Giants Are Getting Crushed.” ProPublica, June 27.

Grabell, Michael, Claire Perlman, and Bernice Yeung. 2020. “Emails Reveal Chaos as Meatpacking Companies Fought Health Agencies over COVID-19 Outbreaks in Their Plants.” ProPublica, June 12.

Henshaw, John L., Shannon H. Gaffney, Amy K. Madl, and Dennis J. Paustenbach. 2007. “The Employer’s Responsibility to Maintain a Safe and Healthful Work Environment: An Historical Review of Societal Expectations and Industrial Practices.” Employee Responsibilities and Rights Journal 19: 173–192.

Hersch, Joni, and W. Kip Viscusi. 2010. “Immigrant Status and the Value of Statistical Life.” Journal of Human Resources 45, no. 3: 749–771.

Hertel-Fernandez, Alexander, Suresh Naidu, Adam Reich, and Patrick Youngblood. 2020. “Understanding the Covid-19 Workplace: Evidence from a Survey of Essential Workers.” Roosevelt Institute Issue Brief.

Hirschman, Albert O. 1970. Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States. Cambridge, Mass.: Harvard University Press.

Hubler, Shawn, Thomas Fuller, Anjali Singhvi, and Juliette Love. 2020. “Many Latinos Couldn’t Stay Home. Now Virus Cases Are Soaring in Their Communities.” New York Times, June 26.

Imershein, Allen, A. Hill, and A. Reynolds. 1994. “The Workers’ Compensation System as a Quality of Life Problem for Workers’ Compensation Claimants.” Advances in Medical Sociology 5: 181–200.

Jamieson, Dave. 2020. “OSHA Doesn’t Expect Employers to Report COVID-19 Hospitalizations Anymore.” Huffington Post, October 15.

Julià, Mireia, Carlos Catalina-Romero, Eva Calvo-Bonacho, and Fernando G Benavides. 2016. “Exposure to Psychosocial Risk Factors at Work and the Incidence of Occupational Injuries: A Cohort Study in Spain.” Journal of Occupational and Environmental Medicine 58, no. 3: 282–286.

Kahneman, Daniel, and Amos Tversky. 1979. “Prospect Theory: An Analysis of Decision Under Risk.” Econometrica 47, no. 2: 263–291.

Keogh, James P., Iman Nuwayhid, Janice L. Gordon, and Patricia W. Gucer. 2000. “The Impact of Occupational Injury on Injured Worker and Family: Outcomes of Upper Extremity Cumulative Trauma Disorders in Maryland Workers.” American Journal of Industrial Medicine 38, no. 5: 498–506.

Kharel, Ujwal. 2016. “The Global Epidemic of Occupational Injuries: Counts, Costs, and Compensation.” Pardee RAND Graduate School, Dissertations.

Kim, Jaeyoung. 2013. “Depression as a Psychosocial Consequence of Occupational Injury in the US Working Population: Findings from the Medical Expenditure Panel Survey.” BMC Public Health 13, no. 1: Article 303.

Kinder, Molly, Laura Stateler, and Julia Du. 2020. “Windfall Profits and Deadly Risks: How the Biggest Retail Companies Are Compensating Essential Workers During the COVID-19 Pandemic.” Brookings Institution Report.

Kniesner, Thomas J., W. Kip Viscusi, and James P. Ziliak. 2010. “Policy Relevant Heterogeneity in the Value of Statistical Life: New Evidence from Panel Data Quantile Regressions.” Journal of Risk and Uncertainty 40, no. 1: 15–31.

Kochan, Thomas A., Michal Smith, John C. Wells, and James B. Rebitzer. 1994. “Human Resource Strategies and Contingent Workers: The Case of Safety and Health in the Petrochemical Industry.” Human Resource Management 33, no. 1: 55–77.

Lane, Julia I., Laurie A. Salmon, and James R. Spletzer. 2007. “Establishment Wage Differentials.” Monthly Labor Review 130, no. 4: 3–17.

Lang, Kevin, and Jee-Yeon K. Lehmann. 2012. “Racial Discrimination in the Labor Market: Theory and Empirics.” Journal of Economic Literature 50, no. 4: 959–1006.

Lang, Kevin, and Sumon Majumdar. 2004. “The Pricing of Job Characteristics When Markets Do Not Clear: Theory and Policy Implications.” International Economic Review 45, no. 4: 1111–1128.

Lay, A. Morgan, Ron Saunders, Marni Lifshen, F. Curtis Breslin, Anthony D. LaMontagne, Emile Tompa, and Peter M. Smith. 2017. “The Relationship Between Occupational Health and Safety Vulnerability and Workplace Injury.” Safety Science 94: 85–93.

Leigh, J.P., and J.A. García. 2000. “Some Problems with Value-of-Life Estimates Based on Labor Market Data.” Journal of Forensic Economics 13: 127–143.

Leigh, J. Paul, Steven B. Markowitz, Marianne Fahs, and Philip J. Landrigan. 2000. Costs of Occupational Injuries and Illnesses. Ann Arbor: University of Michigan Press.

Lester, Richard A. 1946. “Wage Diversity and Its Theoretical Implications.” Review of Economics and Statistics 28, no. 3: 152–159.

Lichtenstein, Nelson. 2020. “It’s Workers Who Should Determine When Their Workplace Is Safe.” American Prospect, April 29.

Lloyd, J. William. 1971. “Long-Term Mortality Study of Steelworkers: V. Respiratory Cancer in Coke Plant Workers.” Journal of Occupational and Environmental Medicine 13, no. 2: 53–68.

Loken, Eric, and Andrew Gelman. 2017. “Measurement Error and the Replication Crisis.” Science 355, no. 6325: 584–585.

Loomis, Dana, and David Richardson. 1998. “Race and the Risk of Fatal Injury at Work.” American Journal of Public Health 88, no. 1: 40–44.

MacLaury, Judson. Undated. “Government Regulation of Workers’ Safety and Health, 1877–1917.” U.S. Department of Labor.

Maestas, Nicole, Kathleen J. Mullen, David Powell, Till von Wachter, and Jeffrey B. Wenger. 2018. “The Value of Working Conditions in the United States and Implications for the Structure of Wages.” National Bureau of Economic Research Working Paper no. 25204.

Mankiw, N. Gregory. 2014. Principles of Economics, Seventh Edition. Boston, Mass.: Cengage Learning.

Marsh, Suzanne M., Cammie Chaumont Menendez, Sherry L. Baron, Andrea L. Steege, and John R. Myers. 2013. “Fatal Work-Related Injuries—United States, 2005–2009.” MMWR Supplements 62, no. 3: 41–45.

Martin, Christopher J., Chuanfang Jin, Stephen J. Bertke, James H. Yiin, and Lynne E. Pinkerton. 2020. “Increased Overall and Cause-Specific Mortality Associated with Disability Among Workers’ Compensation Claimants with Low Back Injuries. American Journal of Industrial Medicine 63, no. 3: 209–217.

Mehta, Chirag, and Nik Theodore. 2006. “Workplace Safety in Atlanta’s Construction Industry: Institutional Failure in Temporary Staffing Arrangements.” WorkingUSA 9, no. 1: 59–77.

Melillo, Jerry M., Terese Richmond, and Gary W. Yohe, eds. 2014. Climate Change Impacts in the United States: The Third National Climate Assessment. U.S. Global Change Research Program.

Michaels, David. 2008. Doubt Is Their Product: How Industry’s Assault on Science Threatens Your Health. Oxford, England: Oxford University Press.

Michaels, David. 2020. The Triumph of Doubt: Dark Money and the Science of Deception. Oxford, England: Oxford University Press.

Michaels, David, and Gregory R. Wagner. 2020. “Occupational Safety and Health Administration (OSHA) and Worker Safety During the COVID-19 Pandemic.” Journal of the American Medical Association 324, no. 14: 1389–1390.

Mill, John Stuart. 1871. Principles of Political Economy, Vol. 1.

Morse, Timothy F., Charles Dillon, Nicholas Warren, Charles Levenstein, and Andrew Warren. 1998. “The Economic and Social Consequences of Work-Related Musculoskeletal Disorders: The Connecticut Upper-Extremity Surveillance Project (CUSP).” International Journal of Occupational and Environmental Health 4, no. 4: 209–216.

Mortensen, Dale T. 2003. Wage Dispersion: Why Are Similar Workers Paid Differently? Cambridge, Mass.: MIT Press.

Moyer, Merriell. 2020. “OSHA Clarifies Reporting COVID Illnesses, Deaths.” Safety News Alert, October 6.

Narea, Nicole. 2020. “Why Meatpacking Plants Have Become Coronavirus Hot Spots.” Vox, May 19.

National Council on Compensation Insurance. 2020. “COVID-19 Regulatory and Legislative Activity.”

NC Watchdog Reporting Network. 2020. “How NC Chose Cooperation over Transparency on Meatpacking Plants with Virus Outbreaks.” Raleigh News & Observer, August 11.

Nelkin, Dorothy, and Michael S. Brown. 1984. Workers at Risk: Voices from the Workplace. Chicago, Ill.: University of Chicago Press.

Nozick, Robert. 1974. Anarchy, State, and Utopia. New York: Basic Books.

O’Leary, Paul, Leslie I. Boden, Seth A. Seabury, Al Ozonoff, and Ethan Scherer. 2012. “Workplace Injuries and the Take-Up of Social Security Disability Benefits.” Social Security Bulletin 72, no. 3: 1–17.

OSHA (Occupational Safety and Health Administration). 2015. “Adding Inequality to Injury: The Costs of Failing to Protect Workers on the Job.” (Citing Foley et al. 2014 and Grabell 2013).

OSHA (Occupational Safety and Health Administration). 2020. “COVID-19 Response Summary.”

Payday Report. 2020. “COVID-19 Strike Wave Interactive Map.”

Piantadosi, Steven, David P. Byar, and Sylvan B. Green. 1988. “The Ecological Fallacy.” American Journal of Epidemiology 127, no. 5: 893–903.

Pierce, Olga, Jeff Larson, and Michael Grabell. 2013. “How We Calculated Injury Rates for Temp and Non-Temp Workers.” ProPublica, Dec. 18.

Premji, Stéphanie, and Niklas Krause. 2010. “Disparities by Ethnicity, Language, and Immigrant Status in Occupational Health Experiences Among Las Vegas Hotel Room Cleaners.” American Journal of Industrial Medicine 53, no. 10: 960–975.

Rappin, Christina L., Sara E. Wuellner, and David K. Bonauto. 2016. “Employer Reasons for Failing to Report Eligible Workers’ Compensation Claims in the BLS Survey of Occupational Injuries and Illnesses.” American Journal of Industrial Medicine 59, no. 5: 343–356.

Rennane, Stephanie, Nicholas Broten, and Michael Dworsky. 2020. “Wage Loss Monitoring for Injured Workers in California’s Workers’ Compensation System: 2016–2017 Injury Year Findings (Third Interim Report).” RAND Research Report.

Reville, Robert T., Leslie I. Boden, Jeff E. Biddle, and Christopher Mardesich. 2001. An Evaluation of New Mexico Workers’ Compensation Permanent Partial Disability and Return to Work. Santa Monica, Calif.: RAND.

Roosevelt, Margot, Hugo Martín, and Taylor Avery. 2020. “Workers Fear Returning to Work. Many Are Resisting the Call.” Los Angeles Times, July 23.

Rosen, Sherwin. 1974. “Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition.” Journal of Political Economy 82, no. 1: 34–55.

Rosenman, Kenneth D., Alice Kalush, Mary Jo Reilly, Joseph C. Gardiner, Mathew Reeves, and Zhewui Luo. 2006. “How Much Work-Related Injury and Illness Is Missed by the Current National Surveillance System?” Journal of Occupational and Environmental Medicine 48, no. 4: 357–365.

Rosenman, Kenneth, and Martha Stanbury. 1996. “Risk of Lung Cancer Among Former Chromium Smelter Workers.” American Journal of Industrial Medicine 29, no. 5: 491–500.

Rosner, David, and Gerald Markowitz. 2020. “A Short History of Occupational Safety and Health in the United States.” American Journal of Public Health 110, no. 5: 622-28.

Ruser, John W. 2008. “Examining Evidence on Whether BLS Undercounts Workplace Injuries and Illnesses.” Monthly Labor Review 131, no. 8: 20-32.

Schlosser, Eric. 2020. “America’s Slaughterhouses Aren’t Just Killing Animals.” Atlantic, May 12.

Seabury, Seth A., Ethan Scherer, Paul O’Leary, Al Ozonoff, and Leslie Boden. 2014. “Using Linked Federal and State Data to Study the Adequacy of Workers’ Compensation Benefits.” American Journal of Industrial Medicine 57, no. 10: 1165–1173.

Seabury, Seth A., Sophie Terp, and Leslie I. Boden. 2017. “Racial and Ethnic Differences in the Frequency of Workplace Injuries and Prevalence of Work-Related Disability.” Health Affairs 36, no. 2: 266–273.

Slovic, Paul. 2000. The Perception of Risk. London, England: Earthscan Publications.

Smith, Caroline K., Barbara A. Silverstein, David K. Bonauto, Darrin Adams, and Z. Joyce Fan. 2010. “Temporary Workers in Washington State.” American Journal of Industrial Medicine 53, no. 2: 135–145.

Smith, Peter M., Ron Saunders, Marni Lifshen, Ollie Black, Morgan Lay, F. Curtis Breslin, Anthony D. LaMontagne, and Emile Tompa. 2015. “The Development of a Conceptual Model and Self-Reported Measure of Occupational Health and Safety Vulnerability.” Accident Analysis & Prevention 82: 234–243.

Stavins, Joanna. 2020. “Unprepared for Financial Shocks: Emergency Savings and Credit Card Debt.” Contemporary Economic Policy 36, no. 1: 59–82.

Steenland, Kyle, Carol Burnett, Nina Lalich, Elizabeth Ward, and Joseph Hurrell. 2003. “Dying for Work: The Magnitude of US Mortality from Selected Causes of Death Associated with Occupation.” American Journal of Industrial Medicine 43, no. 5: 461–482.

Stout-Weigand, N. 1988. “Fatal Occupational Injuries in U.S. Industries, 1984: Comparison of Two National Surveillance Systems.” American Journal of Public Health 78: 1215–1217.

Strong, Larkin L., and Frederick J. Zimmerman. 2005. “Occupational Injury and Absence from Work Among African American, Hispanic, and Non-Hispanic White Workers in the National Longitudinal Survey of Youth.” American Journal of Public Health 95, no. 7: 1226–1232.

Strunin, Lee, and Leslie I. Boden. 2004. “Family Consequences of Chronic Back Pain.” Social Science and Medicine 58, no. 7: 1385–1393.

Sunstein, Cass. 2014. Valuing Life: Humanizing the Regulatory State. Chicago, Ill.: University of Chicago Press.

Swanson, Ana, David Yaffe-Bellany, and Michael Corkery. 2020. “Pork Chops vs. People: Battling Coronavirus in an Iowa Meat Plant.” New York Times, May 10.

Thompson, Alexander. 2020. “Ohio Meatpacking Plants Stayed Open Despite Coronavirus Outbreaks.” Columbus Dispatch, August 12.

U.S. Department of Labor. 2020. “U.S. Department of Labor Cites Ohio Nursing Facilities for Failing to Fully Implement Respiratory Programs to Protect Employees from Coronavirus.” News Release, July 21.

U.S. Department of Labor, Bureau of Labor Statistics (U.S. DOL-BLS). 2019. “2018 Annual Averages, Employment Status of the Civilian Noninstitutional Population in States by Sex, Race, Hispanic or Latino Ethnicity, and Detailed Age.”

U.S. Department of Labor, Bureau of Labor Statistics (U.S. DOL-BLS). 2020a. “Case and Demographic Characteristics for Work-related Injuries and Illnesses Involving Days Away From Work.”

U.S. Department of Labor, Bureau of Labor Statistics (U.S. DOL-BLS). 2020b. “Census of Fatal Occupational Injuries (CFOI)—Current and Revised Data.”

U.S. Department of Labor, Bureau of Labor Statistics (U.S. DOL-BLS). 2020c. “Census of Fatal Occupational Injuries (CFOI)–Fatal Injury Rates.”

Viscusi, W. Kip. 1983. Risk by Choice: Regulating Health and Safety in the Workplace. Cambridge, Mass.: Harvard University Press.

Viscusi, W. Kip. 2003. “Racial Differences in Labor Market Values of a Statistical Life.” Journal of Risk and Uncertainty 27, no. 3: 239–256.

Viscusi, W. Kip. 2013. “Estimating the Value of a Statistical Life Using Census of Fatal Occupational Injuries (CFOI) Data.” Vanderbilt Law and Economics Research Paper no. 13-17.

Viscusi, W. Kip, and Joseph E. Aldy. 2003. “The Value of a Statistical Life: A Critical Review of Market Estimates Throughout the World.” Journal of Risk and Uncertainty 27, no. 1: 5–76.

Viscusi, W. Kip, and Nick Marquiss. 2020. “A Regulatory Policy Strategy for Protecting Immigrant Workers.” Seton Hall Law Review, forthcoming.

Washington Post-Ipsos. 2020. “Washington Post-Ipsos Coronavirus Employment Survey, April 27-May 4. 2020.” May 20.

Weil, David. 2014. The Fissured Workplace. Cambridge, Mass.: Harvard University Press.

Wiatrowski, William J., and Jill Janocha. 2014. “Comparing Fatal Work Injuries in the United States and the European Union.” Monthly Labor Review 137, no. 6.

Empirical research on the causes of the surprisingly persistent earnings gap between women and men often takes the form of statistical models that control for as many variables as possible—race, ethnicity, education, labor force experience, job tenure, hours of work, occupation, industry—but ignore any measures of bargaining power other than unionization. This paper challenges this neoclassical approach, focusing instead on how the institutional landscape of unequal bargaining power of employers and workers and men and women has created costly trade-offs that perpetuate gender inequality. Attention to the history of patriarchal and capitalist institutions—as well as efforts to mitigate or modify them—is crucial to an understanding of a persistent gender pay gap. From this perspective, outright discrimination represents only the tip of a larger iceberg that has frozen women into economic disadvantage, assigning them responsibility for tasks whose value is indispensable yet difficult to measure or monetize.

Because social institutions solidify differences in collective bargaining power, institutional change is difficult to achieve. Yet the choice to collaborate with others to challenge unfair social institutions is among the most important choices people make. Such commitments can be risky, but they also yield rich rewards for everyone. Like other disempowered groups, women are often able to overcome their differences, find allies, and bargain for change, and their history of hard-won but cumulative successes challenges mainstream economic thinking and validates the rallying power of appeals to social well-being rather than private profit.

Introduction

Despite considerable progress since 1973, employed women in the United States still earned about 20% less per hour than men in 2019 (EPI 2020). Conservative economists attempting to explain this surprisingly persistent gap often emphasize that women tend to have less continuous labor force experience, work fewer hours per week for pay, and enter less-well-paying occupations and industries than men. Perhaps, they speculate, women are simply choosing to trade higher productivity and pay in paid employment for more intrinsically rewarding commitments, including family care, and they should not complain unless they are paid less than equally productive men working exactly the same number of hours performing exactly the same job (Wall Street Journal 2019).

While few economists endorse such an extreme interpretation of the gender wage gap, empirical research on the topic often takes the form of statistical models that ignore any measures of bargaining power other than unionization, and control for as many variables as possible—race, ethnicity, education, labor force experience, job tenure, hours of work, occupation, industry—in order to isolate the unexplained gender effect (Blau and Kahn 2017). This paper challenges this neoclassical approach, focusing instead on a bigger social and historical picture in which institutions shaped by the unequal bargaining power of employers and workers and men and women (as well as other groups) have created costly trade-offs that perpetuate gender inequality. From this perspective, outright discrimination represents only the tip of a larger iceberg that has frozen women into economic disadvantage, assigning them responsibility for tasks whose value is indispensable yet difficult to measure or monetize (Folbre 2017, 2021). Attention to the history of patriarchal and capitalist institutions—as well as efforts to mitigate or modify them—is crucial to an understanding of a persistent gender pay gap. Many of the so-called “control variables” included in models of wage determination—hours of work, occupation, industry, etc.—cannot be explained purely as the result of individual choices. Rather, they reflect structural inequalities related to unequal bargaining power.

In standard human capital models, individuals make choices in an economic environment in which prices and incomes are determined primarily by technologies of production. These models require capital, but the fact that the institutional environment constrains access to the accumulation of both financial and human capital is largely ignored. As a result, the impact of collective efforts to challenge laws, public policies, and cultural norms is largely obscured. Yet the historical record shows that differences in the relative bargaining power of employers and workers have interacted with other dimensions of inequality, including gender, sexuality, race, and ethnicity, to limit the spaces of feasible choice. Yet these spaces have been, and will continue to be, reconfigured by collective action, offering lessons for the design of progressive strategies to bargain for economic justice.

The limits of neoclassical economics

Modern neoclassical economics emphasizes the role of individual choices and minimizes the relevance of collective identity and action. Pressures from inside and outside of the profession are weakening this emphasis, but it has shaped the trajectory of much research on gender inequality and continues to exercise an inertial influence that discourages attention to the ways in which individuals collaborate with others to encourage or resist institutional change. The end result is a story that, despite some insights, remains incomplete and misleading.

The canonical assumptions of mainstream economics are these: Individuals seek to maximize satisfaction of preferences that are taken as given (exogenous), they are not concerned about the satisfaction of others outside their family (no interdependent utilities), and they have perfect information along with the capability to act on such information effectively (they are rational actors). Their assets, including their levels of education, are also taken as a given. Firms seek to maximize profits and typically inhabit a perfectly competitive economic system in which they cannot directly affect wages or prices and have little influence on the institutional environment that affects their profitability. Workers and firms are treated as though they are on an equal footing in a labor market: One supplies labor, the other demands labor; both are equally powerless to resist impersonal market forces that determine wages, working conditions, and terms of employment. Because these assumptions are unrealistic, many economists are willing to depart from at least one of them, even as they stick with the larger package that deflects attention from both collective identity and distributional conflict.

A look at research on gender inequality illustrates this package effect. Conventional explanations often ascribe gender-based discrimination to anachronistic, pre-modern attitudes that should decline over time because they are manifestly inefficient. Gary Becker argued in the late 1950s that profit-maximizing employers should drive discriminating employers out of business (Becker 1957). Many economists who rejected this optimistic prediction nonetheless accepted its basic premise, scurrying to emphasize that the process could be slowed by information problems leading to motivationally benign forms of “statistical discrimination” such as basing decisions on stereotypes (Arrow 1974). Such discrimination is widely construed as efficient behavior consistent with profit maximization (though this explanation has not gone uncontested by mainstream economists; see Schwab 1986). Note that suspension of one assumption—in this case, perfect information—reconciles the basic argument with the observed persistence of employer discrimination while also freeing employers of culpability.

Economics has come a long way since Becker and Arrow, and plenty of mainstream economists have turned their attention to historical origins of gender inequality. Nathan Nunn, Alberto Alesina, and Paola Giuliano (2013) build on earlier research by Ester Boserup (1970) to argue that plow agriculture, more physically demanding than hoe agriculture, may have empowered men more than women in societies that relied on the plow. This is a plausible argument, especially because it suspends the assumption that preferences should be taken as a given and suggests the influence of social norms shaped by the experience of previous generations. This path dependence helps explain why a technology that has long since become obsolete may exercise a persistent influence. Yet this also implies that no one is to blame. It deflects attention from the ways in which powerful groups (in this case, men) develop social institutions—such as exclusion of women from political participation or cultural influence—that work in their favor. Inherited norms, like Adam Smith’s invisible hand, exert impersonal, decentralized power.

In mainstream economic reasoning, technology is often cause and inequality is effect. For instance, Claudia Goldin (1990) attributes gradual increases in the ratio of female to male wages in the 19th century United States to the expansion of jobs requiring mental skills rather than physical strength. Rick Geddes and Dean Lueck (2002) argue that the increasing importance of human capital gave men incentives to relinquish power over women in order to benefit from their potential contributions to household income. Jeremy Greenwood, Ananth Seshadri, and Mehmet Yorukoglu (2005) describe labor-saving household devices such as washing machines as “engines of women’s liberation.” Martha J. Bailey, Brad Hershbein, and Amalia R. Miller (2012) give the “contraceptive revolution” considerable credit for increases in women’s labor force participation in the U.S. in the mid-20th century.

Again, these are plausible claims, but they skip over the role of collective identity and action in shaping social institutions and patterns of social inequality. They ignore differences in the bargaining power between employers and among different groups of workers. If physical strength was terribly important to economic success in the 19th century, is there any evidence that it influenced earnings differences among men? Even in that time period, bankers, doctors, and lawyers earned more than farmers. Enslaved people—those who survived—were more physically robust than most of their owners and overseers, but this hardly worked to their advantage. Free women obviously had far more rights than women and men who were enslaved; nonetheless they were also subject to institutions devised and enforced by others: patriarchal laws, norms, and patterns of control over resources that limited their bargaining power relative to men.

Similar constraints are relevant to other claims regarding the technology/gender nexus. The increasing importance of education, increasing women’s potential contribution to household income, may well have reduced men’s resistance to keeping women in the home. However, it does not explain how women fought for and won changes in property law giving them legal rights to both inherited wealth and their own earnings. Washing machines were a great invention, but much of the time they saved was reallocated to better meal preparation and higher standards of cleanliness for homes and wardrobes rather than to earning individual income. Quantitative analysis of the effects of access to contraception relies entirely on variation in the timing of its legalization across states—the result of a difficult and uneven struggle to modify institutions.

Most introductory economics courses take pains to explain that workers generally get what they deserve because they are paid the marginal product of labor. As Gregory Mankiw puts it, “The rich earn higher incomes because they contribute more to society than others do” (Mankiw 2013, 29). By extension, men earn higher wages because they contribute more to society than women do. These claims are not based on empirical evidence; they simply follow from the assumptions of marginal productivity theory (Folbre 2016). Gary Becker (1991) famously argues that women are less productive employees because they devote so much more time and energy to nonmarket household work.

Outside of introductory texts, the theory of marginal productivity has been largely displaced by a human-capital theory based on more empirical reasoning: statistical links between quantity and quality of education, labor force experience, and earnings. Such links never fully explained wage variation, but, ironically, they did expose the empirical impact of race and gender on rates of return to measured human capital (Folbre 2016). Still, the theory’s proponents often insist that poor measurement is to blame: Quality of education is variable, and women choose the wrong college majors. Now that women in the U.S. average higher educational attainment than men, attention has shifted to gender differences in labor force experience and hours of work. Claudia Goldin and Lawrence Katz (2016) argue that women in top professional and managerial occupations earn less than men largely because they are unwilling to devote long hours to the highly paid jobs that require such commitment.

Most women do put in fewer hours of employment than men. They are more likely to be employed part time and much less likely to put in more than 40 hours per week (Cha and Weeden 2014). That employers are willing to pay significantly higher hourly wages for long work hours suggests that they help generate profits. But greater profitability does not necessarily imply greater productivity or efficiency. Among the institutional factors that influence the relationship between hours of work and hourly pay are public policies governing overtime requirements (quite weak for professional/managerial occupations) and work-related benefits such as health insurance and pensions. In the United States today, such benefits make up about 38% of total compensation and are greatest in high-paying jobs (BLS 2019a). U.S. labor market competition demands extremely long hours from workers competing for high-wage jobs and short, unpredictable “just-in-time” hours from those competing for many low-wage jobs. Given a meaningful choice, both women and men would prefer more balanced options.

The organization of many professional and managerial jobs with regard to time has been institutionalized on the establishment level, often in ways advantageous to senior managers. Because the prevalence of extremely long hours in top-paying jobs has been influenced by managerial preferences and power, rather than dictated by technology, it can be renegotiated. Growing demands for more family-friendly policies, along with stricter rules against sexual harassment and awareness of glass ceilings that limit women’s advancement, have had tangible, if partial, effects. In professions where women have gained a strong foothold, such as accounting, law, and medicine, pressure for change is mounting, particularly since many professional men now lack the stay-at-home spouse who once enabled them to clock long hours in paid employment (Goff and Le Feuvre 2017; Lott and Klenner 2018).

If individual choices and actions are deemed the primary movers of social change, then much depends on what individuals want. Yet political, cultural, and economic power both shape preferences and constrain the ways in which they can be acted upon. Historically, neoclassical economists have warned against ad hoc assumptions regarding differences in preferences, since such assumptions can be used in a circular way to explain virtually any behavior (“they did it, so they must have wanted to do it”) (Stigler and Becker 1977). Nonetheless, many researchers continue to echo the long-standing claim that women earn less money than men simply because they care less about their earnings, prioritizing family care even at the price of poverty (Fuchs 1988; Hakim 2000). This claim fits neatly into the theory of compensating differentials, which holds that workers are willing to pay a wage penalty in order to enter jobs with nonpecuniary characteristics that they value (Rosen 1986), and dovetails with the claim, described above, that women wage earners are less productive than male wage earners because they are unwilling to work as hard or as long.

Considerable evidence suggests that, on average, in the U.S. women have different preferences than men do, often related to the importance of money versus people (Croson and Gneezy 2009; Fortin 2008). However relevant such differences may be, they do not fully explain gender differences in earnings, for several reasons. A focus on averages can be misleading. Differences in specific preferences, such as those relating to risk, are often small, reflecting considerable overlap between women and men (Nelson 2016). Data from stylized experiments and surveys cannot fully capture the nuances of gendered behavior: Men may be more competitive than women in some contexts, but not in others (Bjorvatn, Falch, and Hernaes 2016). For instance, experimental evidence suggests that when rewards benefit offspring rather than players, mothers are just as competitive as fathers (Cassar, Wordofa, and Zhang 2016).

Preferences operate in a social context: Norms influence both the formation of preferences and the cost of expressing them. For instance, gender differences in competitiveness are not apparent in matrilineal societies (Gneezy, Leonard, and List 2009; Andersen et al. 2013). In the U.S., as in other countries, women who move into better-paying but stereotypically masculine occupations often face sexual harassment and disapproval. Preferences that pay off in the labor market can prove costly for heterosexual women in the dating market (Badgett and Folbre 2003). Gender norms work the other way as well: Evidence suggests that men entering traditionally female occupations have a harder time finding spouses (McClintock 2020). And when wives earn more than husbands, both spouses slightly tilt their reported earnings to conform to gender stereotypes, overstating the relative size of husbands’ earnings (Murray-Close and Heggeness 2018).

The economic context also invites attention. Why do certain preferences impose a penalty, while others do not? And what determines the size of the penalty? After all, some people land in jobs that offer them high pay as well as nonpecuniary benefits. Mainstream theory defines compensating differentials as a kind of psychic income determined by the impersonal forces of supply and demand; it assumes that workers maximizing their utility and employers maximizing their profits make efficient decisions. The real story is more complicated: Both market forces and social institutions influence the “price” of caring for others.

While the textbook theory of labor supply focuses primarily on the trade-off between hours of market employment and hours of leisure time, adult Americans devote about as much time, on average, to unpaid work such as housework, home repair, and child care as they do to paid employment (BLS 2019b). This work, while not included in measures of gross domestic product, creates profound economic benefits, producing and maintaining the human capabilities on which the market economy depends. Employers, however, have little economic incentive to help pay for it. Expenditures of time and money on children and other dependents, whether made by families or through the state, are considered just another form of consumption. The care of family, friends, and neighbors represents an “externality” that is easily taken for granted, like the natural environment and the global climate.

Many women self-select into traditionally female occupations because they consider these more compatible with the demands of family care (Charles and Grusky 2005). But while they may be aware that these jobs pay less than traditionally male jobs, they don’t choose the size of the resulting pay penalties. In the U.S., the percentage of women in an occupation is inversely related to its average pay, even controlling for human capital characteristics. Sociologists offer considerable empirical evidence of what they term a “devaluation” effect (England, Allison, and Wu 2007). Occupations that involve face-to-face or hands-on care for others also impose pay penalties (even independent of female percentage) (Bittman et al. 2003; Hodges, Budig, and England 2018; Pietrykowski 2017). These occupational penalties vary considerably across countries because they are significantly shaped by public policies, union membership, and labor market conditions (Budig and Misra 2011; ILO 2018). While workers’ choices of employment are often constrained, employers’ decisions are empowered not only by their ability to hire and fire but also by their ability to “invest” in political and cultural influence.

Across all occupations, many women experience wage penalties as a result of becoming mothers (England, Budig, and Folbre 2002). In recent years, this penalty has declined for married mothers in the U.S. but remains quite high for single mothers (Pal and Waldfogel 2016). Comparative international research shows that work–family policies, such as paid family leave, child allowances, and pension credits for unpaid work, can mitigate these penalties (Harkness and Waldfogel 2003; Boeckmann, Misra, and Budig 2015; Budig, Misra, and Boeckmann 2012). A recent review of research on the gender wage gap notes that women’s labor force participation rates in the U.S. have grown more slowly than those in northwestern European countries in recent years, probably as a result of differences in family policy (Blau and Kahn 2017).

This research both parallels and extends insights on the pay penalties for less-than-full-time employment that affect men as well as women. Such penalties vary across countries but tend to be higher for men than for women, partly because women’s earnings are already lowered by the factors outlined above. Labor force surveys in the U.S. show that workers who report being involuntarily part time pay a penalty larger than those who report their status as voluntary, a finding somewhat inconsistent with the compensating differentials approach (Golden 2020).

Here again, institutions matter. Employers may prefer to offer only part-time jobs because they can easily be excluded from benefit packages or contracted out (Weil 2017). In retail services, in particular, many businesses rely on short, unpredictably scheduled hours in order to match fluctuating levels of customer demand, a strategy that benefits employers more than workers (Lambert 2008). Like long hours, short and unpredictable hours can make it difficult for workers to meet family care responsibilities (Morsy and Rothstein 2015; Carrillo et al. 2017). Employers seize gains while families—also part of the economy—suffer losses that don’t show up on conventional balance sheets.

The persistent though diminished role of gender discrimination in the labor market testifies to the power that employers exert across a wide range of labor markets. The strongest evidence of discrimination in the not-too-distant past comes from studies that effectively isolate the effect of gender from assessment of actual or potential experience on the job. One well-known example explores the positive implications for women orchestra musicians of auditioning behind a screen concealing their gender identity (Goldin and Rouse 2000); another example, based on submission of otherwise quite similar resumes to high-end restaurants (an “audit” study), revealed a distinct preference for male applicants (Neumark, Bnak, and Van Nort 1996).

More recent studies using audit methodologies reveal a pattern of discrimination against mothers based on rather subtle cues, such as participation in a parent-teacher organization listed on a job resume (Correll, Benard, and Paik 2007). Some lawyers argue that the most consequential form of discrimination in the U.S. today takes this form (Bornstein, Williams, and Painter 2012). In many ways it illustrates the potential role of statistical discrimination, since employers may well assume that mothers of young children face other demands on their time that lower their performance in paid employment. Yet motherhood penalties also illustrate the effects of long-standing employment policies built around presumptions regarding “ideal workers” that are now grievously out of date.

Audit studies of penalties paid by gay men in the labor market seem to offer more straightforward examples of “preference-based” discrimination (Tilcsik 2011). Considerable evidence suggests that employers also discriminate on the basis of gender identity as well as sexual orientation (Badgett et al. 2009). The pattern of results in one recent audit study strongly suggests forms of implicit bias—bias of which respondents might themselves be unaware—linked to expectations regarding traditionally male and female behaviors (Gorsuch 2019). This finding is consistent with research showing that the same personality traits affect men’s and women’s earnings differently—men seem to be rewarded for assertiveness, while women are not (Mueller and Plug 2006).

Research building on standard neoclassical models has generated many little individual pieces that fit into a jigsaw puzzle that still has many glaring holes. Contrary to predictions, individual employers are often able to influence wages, and market forces do not necessarily erode discriminatory preferences. The textbook distinction between the supply side and the demand side of the labor market is weakened when employers adjust demand for workers based on assumptions regarding future labor supply, and women supply labor to the market based on assumptions regarding future demand. Furthermore, firms often have significant ability to influence wages in ways that buffer mechanisms of competitive supply and demand.

Traditionally, the term “monopsony” was used to describe the market power a particularly large employer could wield in a local labor market where worker mobility was limited. Today it is increasingly applied to broader circumstances in which the supply of labor does not respond in a sensitive way to wage offers. Research on linked employer–employee data in the U.S. shows that women’s labor supply is less elastic than men’s because women’s mobility between jobs is limited by obligations of family care (Webber 2016). Employers can easily take advantage of this difference, engaging in what might be termed “strategic discrimination.” That is, they pay women less than men not because they have a preference for hiring men but because they recognize that women are more likely to accept lower wages.

In other words, employer discrimination, while often based partly on attitudes, is heavily influenced by institutional context. Economic models that focus entirely on the individual choices of individuals and firms have little to say about the processes of collective action that influence laws, norms, and access to economic resources. Hence, it is important to plug many of the findings above into a broader picture of social institutions affecting the relative bargaining power of employers and different groups of workers.

Bargaining for change

Feminist theories of intersectional inequality have much in common with stratification economics in that they both emphasize multidimensional forms of group conflict over rules, norms, and the distribution of economic resources (Folbre 2021; Darity 2005). As suggested by the term “social contract,” these institutional arrangements are under virtually constant negotiation at many different levels. Bargaining often characterizes situations where two parties, whether individuals or groups, see potential gains from mutual cooperation but disagree over how those gains should be shared. Both parties can potentially benefit from coming to an agreement, and their share is likely to be strongly affected by their fallback position, or next-best option.

Bargaining is not always an agreeable, consensual process, and it is heavily influenced by threats and bribes. Indeed, the process often termed “cooperative conflict” can sometimes be described instead as “coerced cooperation” (Folbre 2021). Examples of individual bargaining include activities as diverse as negotiating a new job or deciding who should cook dinner. Examples of collective bargaining include efforts to regulate labor contracts or to join with other similarly positioned or like-minded people to challenge other institutional arrangements or implement new ones. When members of disempowered groups come together with sufficient conviction in sufficient numbers, they can sometimes change the balance of power.

Recent changes in the legal definition and prosecution of sexual harassment provide a vivid example. Sexual harassment at the workplace is a terrible experience, and research indicates that it has serious economic consequences, leading many young women to change jobs, lose the benefits of tenure and firm-specific experience, and end up in safer but lower-paying occupations (McLaughlin, Uggen. and Blackstone 2017). Women’s ability to protect themselves in the workplace has been significantly affected by hard-won legal protections that were a long time coming. While the 1964 Civil Rights Act outlawed explicit discrimination against women, sexual harassment was not included until the Equal Employment Opportunity Commission officially began to consider unwanted sexual advances as a form of discrimination in 1980. In 1986, a Supreme Court decision (Meritor Savings Bank v. Vinson) held employers responsible for the sexual harassment of their workers (if made aware and given a chance to remedy it). In 1993, the Eveleth Taconite Co. was found liable in a lawsuit filed by more than 100 women who were victims of sexual harassment, a story eloquently told in the documentary film North Country (Thomas 2016).

Employers quickly took advantage of two legal tools to buffer threats of prosecution. Confidential settlement agreements allowed them to buy their way out of protracted legal actions while protecting their reputations and concealing the true extent of actionable harassment (Baum 2019). The mandatory arbitration clauses that proliferated in employment contracts often tied women’s hands (Levinson 2019). Individual women often lacked sufficient bargaining power to effectively defend their legal rights. Investigative reporting and social media helped tip the balance: In 2017, Facebook reported more than 12 million posts, comments, and reactions regarding #Me Too, numbers that lent weight to individual complaints and challenged conventional social norms of the “boys will be boys” variety (Garcia 2017).

The prosecution and conviction of celebrities like Bill Cosby and Harvey Weinstein have shifted social norms of unacceptable behavior in intimate transactions away from attention to physical coercion toward emphasis on the need for unconstrained and explicit consent. This shift has fascinating implications for the analysis of other voluntary transactions, including those between employers and workers. Economic coercion can take many forms. Before the 1990s, threats such as “have sex with me or you’ll never work in this town again” may have been considered vile, but they were not deemed illegal. Consider the parallel with “sign this arbitration or noncompetition clause in this contract or you’ll never work in this town again.”

Feminist reasoning challenges the consensual nature of intimate transactions when one person wields far more power than the other, an extension of the legal doctrine labeled “abuse of position,” which often prohibits sex between police officers and those they arrest, health professionals and their patients, or managers and those they supervise. As Wall Street Journal reporter Stuart Green observes, hierarchical power makes it impossible to accurately assess genuine consent (Green 2020). The same reasoning can be extended to economic transactions between those in command of great wealth and workers lacking any viable alternatives to employment by them.

Bargaining is an intrinsically multilayered and circular process in which initial gains can be parlayed into a stronger bargaining position that can, in turn, guarantee even larger gains. Both individual and collective capabilities come into play, along with external factors (such as the unemployment rate) over which bargainers have little control. An individual’s fallback position in bargaining over earnings depends partly on individual characteristics (such as level of education and years on the job) that are often measured, and some characteristics (such as wealth ownership, job-specific skills, and potential support from other family members) that are not. However, fallbacks for workers as a group are indirectly and directly affected by the institutional environment: the minimum wage, the level of unemployment, the threat of outsourcing or technological obsolescence, and the social safety net, including access to unemployment benefits, health insurance, pension benefits, and other income transfers. Differences in fallbacks among different groups of workers are heavily influenced by factors such as gender, sexuality, race, and ethnicity, which affect personal histories (such as access to education), norms and values (such as commitments to family care), and employers’ perceptions of “ideal workers.”

Multiple sources of bargaining power

A first step toward a more comprehensive model of gender bargaining in the labor market reframes some of the findings of mainstream research reviewed above. For instance, differences in productive technology (such as reliance on the hoe versus the plow) could have implications for women’s relative productivity in agriculture, and their relative productivity could, in turn, affect their fallback position—their “outside” option. Levels of intergroup conflict are also relevant: Vulnerability to abduction, rape, or violence outside the community makes it difficult for women to challenge community restrictions on their economic opportunities. Differences in weapon technology can influence the scope of gendered bargaining power. In Naomi Alderman’s science fiction tale, The Power (2017), a mutation caused by toxic military waste gives young women the capacity to generate bio- electric shocks that give them a distinct advantage over young men in hand-to-hand combat. Women as a group are empowered by it.

But this story, like Margaret Atwood’s The Handmaid’s Tale, also points to the role of social institutions that crystallize and fortify disempowerment. In the early U.S.—as in many other countries—specific laws and public policies excluded white women as well as enslaved women and men from participation in political decision-making, restricted their property rights, and limited their access to forms of education and experience that would have enabled them to compete with men (Schloesser 2002). They also excluded women from leadership in institutions such as established churches that had influence over norms and preferences.

These institutions were established through processes of collective action—a topic that, with some important exceptions, neoclassical economists tend to ignore.1 Often such collective action required alliances between groups with somewhat disparate interests, resulting in complex patterns of legitimation and cooptation. Members of partially disempowered groups such as wealthy white women or white working-class men often found themselves in contradictory positions, advantaged in some respects, disadvantaged in others. The complexity of identities and interests in intersecting and overlapping hierarchies made it difficult to mobilize opposition to the status quo (Folbre 2021).

Women are not a homogenous group. The only way to fully understand the position of women in the labor market is to examine how the disadvantages they face interact or overlap with other dimensions of group membership such as race, ethnicity, class, and citizenship (McCall 2001). The same holds for people who identify as lesbian, gay, bisexual, transsexual, or queer. A broad approach to bargaining power calls attention to all the factors that may combine to influence relative fallback positions in the labor market. It has long been observed that the effects of discrimination based on race and gender cannot simply be added up: They interact in complex ways (Greenman and Xie 2008). Women wage earners, like all wage earners, are divided in some respects even while they are unified in others.

The level of unemployment is generally considered a significant determinant of the bargaining power for all employees aiming to improve their wages and working conditions. It also can also significantly influence the relative position of different groups of workers. For instance, when the labor market is slack because the unemployment rate is high, competition for available jobs can intensify efforts to exclude less-empowered workers from the market (Bonacich 1972; Williams 1987). On the other hand, when the labor market is tight, competition among workers is diminished, and the cost of discrimination to employers goes up. Biddle and Hamermesh (2013) provide evidence from Current Population Survey data over the 1979–2009 period that employers discriminate less against women in a tight labor market.

Differences in the relative earnings of women and men (and between other subgroups) are significantly influenced by the overall level of earnings inequality, especially the overall distance between those at the top and those at the bottom. In Sweden, for instance, women earn more relative to men than in the U.S. in part because there is less overall earnings inequality (Blau and Kahn 2017). Increases in the polarization of earnings in the U.S. have made it particularly difficult for women to catch up with men (Fortin, Bell, and Böhm 2017). At the same time, the small number of women who manage to “swim upstream” intensifies class divisions among women, weakening feminist coalitions. It hardly seems incidental that feminist public policies have flowered most profusely in Scandinavian countries characterized by relatively low levels of class and race inequality.

One of the most important paths to upward economic mobility in the United States for the past 50 years has been attainment of university and postgraduate degrees. When discriminatory barriers to entry were lowered, young women from relatively affluent—and typically relatively white—backgrounds were able to climb this path. Black and Hispanic families were less able to finance their children’s education. The historical effects of racial discrimination exert a strong effect on family income and intergenerational wealth transmission, as well as labor markets (Darity 2005). Partly as a result, the median Black–white earnings gap has followed a very different trajectory than the median woman–man earnings gap and is now about as large as it was in 1950 (Bayer and Charles 2018).

Yet gender also exerts an influence: Black women enjoy greater earnings mobility relative to their white counterparts than Black men do (Chetty et al. 2019). There’s no simple explanation for this divergence, but the difference between individual earnings and access to income seems relevant. Black women may be under greater pressure than white women to increase their earnings because their chances of marrying and pooling income with a high-earning man are much lower. Most women who make it to the top 1% of income—a very small group—do so through marriage, not by their own earnings (Yavorsky et al. 2019).

Family background also matters. In the U.S. today, class and race generally affect access to intergenerational transfers more directly than gender does. Children raised by single parents (typically their mothers) are especially vulnerable to economic disadvantage, but this family structure is far more common in the bottom half of the income distribution than the top half. Recent research using longitudinal data documents substantial differences in earnings among young people with different family backgrounds but similar educational attainment, even after taking the effects of college selectivity, college major, and academic performance into account (Witteveen and Attewell 2017; Laurison and Friedman 2016). Parents with significant financial resources convey a number of tangible benefits to their young adult children—support for additional training (such as unpaid internships), payment of tuition and fees, safety-net assistance while experiencing unemployment or engaging in a job search, and help with mortgage down payments.

Consideration of these multiple sources of relative bargaining power puts inequalities based on gender and sexuality in context and explains why women are not an automatically unified constituency for change. Feminist organizing efforts have often been most successful when allied with a larger vision of economic justice and sustainability. A closer look at the forms of institutional change that have increased women’s collective bargaining power in the U.S. illustrates the significance of intersectional alignments. It also helps explain why resistance to gender equality remains so persistent.

Bargaining for equality

Efforts to develop the coalitions necessary to undermine patriarchal institutions pepper the historical record (Amott and Matthaei 1999). A detailed chronology matters less than appreciation of the thrust of collective efforts to reduce social divisions and challenge multidimensional inequalities. Initially focused on gaining property rights, the franchise, and access to higher education, feminist activists began to challenge institutions governing paid employment. Even small successes proved significant, gradually peeling back layers of institutional resistance and, in the process, revealing many of the dynamics by which group-based inequalities are reproduced over time.

In the early 19th century, women who gained political experience in the abolitionist movement went on to hold the first Women’s Rights Convention in Seneca Falls, N.Y., in 1848. Their basic principles were based on the U.S. Declaration of Independence and the rights of all persons to life, liberty, and the pursuit of happiness. Yet extreme racial divisions were toxic. With the advent of the Civil War, Union victory, and a constitutional amendment to give Black men—but not women—the right to vote, predominantly middle-class white women became the loudest voice of feminist activism. The promises of abolition proved short-lived, broken by the force of white supremacy. Political maneuvering on the state level led to the development of Jim Crow laws that segregated and disenfranchised African Americans as a group. Racial divisions also undermined the labor movement. Solidaristic groups such as the Industrial Workers of the World (IWW) were eclipsed by a trade union movement largely fixated on potential gains for white men (Hill 1996).

During the early decades of the 20th century, campaigns to give women the franchise and expand their access to education found expression in the Progressive Era concept of “the New Woman” (Rich 2009). Social feminists like Florence Kelley and Jane Addams insisted that poverty and inequality were important issues for women. In 1920, woman’s suffrage was finally achieved—a major political event that led to significant shifts in legislative priorities, including greater public spending on health care (Miller 2008). With growing numbers of young women in wage employment, socialist feminists emerged within the trade union movement. While racial divisions remained extreme, efforts to build progressive coalitions were at least partially successful (Tax 2001).

The shock of the stock market crash of 1929 and the ensuing Great Depression led to major political realignments, and the New Deal laid the foundations of the modern welfare state. Women played a relatively important role in Franklin Roosevelt’s administration, and their efforts had a lasting impact (Cobble 2005). Yet policies implemented early in the Great Depression and during the New Deal treated women as secondary earners, exemplified by 1932 legislation that required government agencies to fire one member of each married couple working in government—a policy clearly directed at working wives (McGuire 2008). Furthermore, many New Deal policies explicitly discriminated against people of color, a story well-told in Ira Katznelson’s book, When Affirmative Action Was White (2006).

The concept of unfair discrimination itself only gradually entered political and legal discourse, largely as a result of the concerted efforts of disempowered groups. The very concept had subversive implications, because its application cannot be limited to one or even two categories of group identity. The processes by which such discrimination takes place and the outcomes it generates are similar, regardless of the characteristics of those affected. Thus, increased awareness of one form of discrimination—as well as adoption of measures to challenge it—typically has gradual spillover effects on other forms. The ability to discriminate in a competitive labor market is conditioned on unequal bargaining power.

Fast forward to the Equal Pay Act of 1963 and the Civil Rights Act of 1964, landmark legislation that prohibited discrimination on the basis of race and sex, initially defined as paying men and women working in the same place different salaries for similar work. Many factors, including relatively rapid economic growth in the preceding years, contributed to changes in norms, preferences, perceptions, and incentives that made this legislation possible. But the cumulative effect of persistent and insistent collective protests against sexism and racism was also significant, and the new legislation itself altered the moral landscape. In the vocabulary of neoclassical economics, perceptions and preferences are partially endogenous: People can and do influence what other people see, believe, and want—which is exactly why control over various means of influence is so important.

The Civil Rights Act catalyzed waves of political mobilization and legal challenge that gradually expanded its scope (Thomas 2016). Definitions of discrimination were expanded to prohibit differential treatment based on marital status and to encompass sexual harassment. Transformation of the legal landscape reached beyond private workplaces to public policies including taxes and Social Security benefits. Educational institutions were also a major site of transformation, as more women entered university and professional training, increasing their access to professional and managerial jobs. Potential to realistically take advantage of anti-discrimination law, while still limited, was expanded as recently as the Lily Ledbetter Fair Pay Act of 2009, extending the time period during which discrimination charges can be filed.

Affirmative-action policies designed to compensate for both implicit and explicit bias had tangible, if modest, economic consequences. Analysis of complaints filed with the Equal Employment Opportunities Commission over a 30-year period suggests that their effects diminished somewhat over time (Kurtulus 2016). Yet some evidence suggests that even temporary policies alter employers’ long-run decisions, exerting an influence even when they are no longer officially in effect (Miller 2017). Ironically, legal challenges to affirmative action in higher education have been accompanied by its relatively successful institutionalization in the U.S. military (Knowles 2013). It is perhaps not surprising that access to higher education is more heavily guarded than access to military combat.

Legal changes and political pressures have clearly helped boost the relative earnings of both Black and white women since the 1960s. And fierce opposition to affirmative action based on gender and race has spurred efforts to develop affirmative action policies based on economic disadvantage as a supplement (not a substitute) for them (Orentlicher 2016).

Lesbian, gay, bisexual, and transsexual people have also made important gains. In 2015, the U.S. Supreme Court struck down all state bans on same-sex marriage, legalized it in all 50 states, and required states to honor out-of-state same-sex marriage licenses. Some workplace protections were implemented in at least 22 states and many large cities even before the U.S. Supreme Court declared in June 2020 that federal law bans employment discrimination based on sexual orientation and gender identity.

The process of bargaining for equality through institutional change has obviously fallen short, in part because it has elicited powerful pushback from groups most threatened by it. Prominent conservative thinkers like Christopher Caldwell reject the Civil Rights Act itself as a travesty and swindle (Rauch 2020). Wealth in the U.S. is highly concentrated in predominantly white and male hands (Kijakasi 2019), and this wealth is systematically deployed in ways that shape access to information and the tenor of political discourse to defend the status quo. Still, even failed efforts to bargain for gender equality yield some important lessons, and new alliance-based bargaining strategies are emerging.

Bargaining lessons

Neoclassical economic thinking focused attention on the discriminatory preferences of employers, consumers, and workers, helping motivate efforts to outlaw discrimination and erode discriminatory norms. Yet this thinking remained superficial, avoiding consideration of deeper institutional inequalities—embedded in public policies and private families—that remained highly resistant to market forces and attitudinal change. Economists and policymakers alike underestimated the inertial momentum of gender inequality or “hysteresis,” defined as the dependence of the state or a system on its history—in this case, the bargaining power of employers relative to workers.

“Equal pay” dictates were, from the very outset, subject to the caveat “for equal work.” Yet laws giving employers the right to restrict information about wages paid made it difficult for women to assess the fairness of their working conditions. High levels of occupational segregation were—and remain today—the largest immediate cause of gender inequality in earnings. Efforts to improve the relative pay of primarily female jobs—rather than merely female workers—met powerful criticism based on the simplistic neoclassical assumption that occupational pay was largely determined by occupational “value added” or productivity. Finally, the supply of women’s labor to the market was treated merely as the result of individual choices, with little attention to the constraints imposed by a traditionally male-oriented organization of work, school, housework, and child care.

Information asymmetries

No one has better dramatized the importance of information about what fellow workers earn than Lilly Ledbetter, a supervisor at Goodyear Tire & Rubber who was about to retire when she received an anonymous note revealing that she had, for some time, been paid considerably less than men with less seniority in the same job. She filed suit against her employer only to find that her window of opportunity had closed as a result of a technical statute of limitations in the relevant legislation. While Ledbetter never received any legal redress, Congress passed the Lilly Ledbetter Fair Pay Act in 2009, ensuring that future charges of discrimination would not be foreclosed in the same way. Had she not received and acted on that anonymous note, many workers today would remain unprotected.

Employers are often able to use their control over information to lower women’s wages relative to men’s. While the Fair Labor Standards Act of 1938 guarantees the right of some workers to discuss their salaries, not all workers are covered, and others remain unaware of their rights. As of 2014, about half of all employees in the U.S. were either contractually forbidden or strongly discouraged by their employers from revealing their salaries (DOL 2014). Highly paid workers may themselves be reluctant to reveal what they earn to those they believe earn less. A number of states, however, have explicitly outlawed pay secrecy, a step that has the added effect of helping change workplace norms. One statistical analysis taking advantage of differences in the timing of these laws in different states finds that such laws had positive effects on the earnings of women with higher education levels and reduced the gender wage gap (Kim 2015). Yet long-standing efforts to pass a federal law against pay secrecy have been consistently stymied by conservative opposition.

On the other side of the information street, employers in most states have the right to ask job applicants what they earned in their previous jobs, information that allows them to accurately assess individual fallback positions and minimize their salary offers. Of course, applicants can refuse to answer the question, but doing so may well jeopardize their chances of landing the job. In 2016, Massachusetts became the first state to ban such salary history questions and as of 2019 a total of 14 states plus Puerto Rico had implemented such bans (McGregor 2019). Statistical analysis of the consequences based on comparison of employer behavior across states and over time (a “difference in differences” design) shows that employers required to observe these rules posted wages more often and increased pay for job changes by about 5% overall, with an 8% increase for women and 13% for African Americans (Bessen, Denk, and Meng 2020).

Information asymmetries help reproduce existing wage inequalities but are not the only factor explaining resistance to change. Simple inertia comes into play: One study of the California state civil service finds that underpayment of female-dominated jobs in the 1930s carried over to the early 1990s (Kim 1999). A large sociological literature demonstrates that women’s entrance into previously male-dominated occupations tends to lower the average occupational wage (speeding men’s tendency to exit these occupations as they tip female; see, for instance, England, Allison, and Wu 2007). This process, often described as “devaluation,” is at least partially driven by the fact that women start out in lower-paying occupations, which lowers their bargaining power (or, in more technical terms, their “reservation wage”). Employers don’t need to ask a specific salary question to know that the average female worker earns less than her average male counterpart, and to adjust wage offers accordingly.

Defining ‘equal value’

The principle of equal pay for equal work always implied something larger—equal pay for work of equal value. However, the concept of “equal value” is underdeveloped in neoclassical economics, which typically equates price and value. By this reasoning, whatever a worker is paid represents his or her value added: The market promises just deserts (Folbre 2016). Gregory Mankiw’s assertion that, “The rich earn higher incomes because they contribute more to society than others do” (Mankiw 2013, 29), and the corollary that men earn more than women because men contribute more to society, is based on circular reasoning that ignores the ways bargaining power conditions the relationship between what workers produce and what they earn. Early efforts to articulate principles of comparable worth and pay equity emphasized this point, demonstrating that female-dominated occupations pay less than male-dominated occupations and, further, these differences were often embedded in company-level personnel policies and internal pay scales that were heavily insulated from market forces (Treiman and Hartmann 1981). As sociologists put it, female-dominated jobs were devalued by employers and, more broadly, by society as a whole (Feldberg 1984; England 1992). Cultural norms can affect perceptions of relative value in ways that reduce women’s bargaining power.

Economists were uniformly skeptical of this argument, insisting that the forces of supply and demand in labor markets align wages and productivity in the long run, even though short-run shocks can disrupt this relationship (as when skilled workers are thrown out of work by automation). Neoclassical reasoning often explained differences in pay between male and female jobs as a result of female preferences for certain types of work (Killingsworth 1987). Economists pointed to differences on the supply side (workers’ choices) rather than on the demand side (employers’ choices) (Holzer 1995). Influential critics of comparable worth insisted that it would interfere unnecessarily with market forces, and they carried the day: Revised job evaluation standards had little impact outside of a few states (Washington, Minnesota, and Iowa), where they increased women’s earnings in state government employment (Sorensen 1994).

More recently, however, a new variant of comparable worth reasoning has emerged, emphasizing that many female-dominated services have a public good dimension—their social value exceeds their private value. The work of caring for others—whether in industries such as health, education, and social services or in occupations such as child care, elder care, teaching, and nursing—creates value that is difficult to capture through the market because it has positive spillover effects and involves emotional engagement, teamwork, and person-specific skills. The effective collaboration of “consumers” themselves (students, patients, clients) affects the value of the work performed (Folbre 2012, 2017).

A growing body of empirical work documents pay penalties for both women and men in many types of care work, controlling for individual levels of education and experience and also unobserved individual differences (England, Buddig, and Folbre 2002; Hirsch and Manzella 2015; Barron and West 2013; Pietrykowski 2017; Hodges, Budig, and England 2018; Folbre and Smith 2020). While a large percentage of care work is either provided or financed through the public sector, these effects hold even in private employment. The direct effects of occupation and industry are modest—with penalties ranging from about 2% to about 18% of earnings. However, since women are disproportionately concentrated in care industries and occupations, these pay penalties lower their earnings relative to those of men on top of the negative effects of gender per se. Furthermore, as noted earlier, women’s lower levels of labor force experience and hours worked (standard control variables) are themselves a reflection of institutional pressures for women to specialize in unpaid family care.

In some ways, research on care penalties builds on the neoclassical principle that private and social value can diverge, as is the case with environmental spillovers or externalities. Yet many neoclassical economists are reluctant to acknowledge the implications of this divergence. Women’s tendency to devote more hours to unpaid care work than do men is interpreted as a result of feminine preferences rather than as the result of institutional pressure to ensure a generous supply of female effort to activities such as family care that cannot be rewarded by market forces. Similarly, the professional commitments of essential care workers in the early stages of the COVID-19 pandemic are typically interpreted as moral heroism, worthy of applause, but not extra hazard pay. The obvious disjuncture between the pay these workers received and the increased social value of the services they provided goes largely unremarked (Folbre, Gautham, and Smith 2020).

Attention to the social benefits of care work extends the concept of comparable worth beyond the gender dimension per se to emphasize disjunctures between private and social value that have proved particularly disempowering for women. It also highlights links between the interests of providers and consumers of care services, since higher pay and improved working conditions often contribute to lower turnover and higher-quality services (Folbre 2006). Likewise, increased public funding of care services can benefit both workers and citizens. For instance, a political alliance between unions and disability rights activists in the late 1990s led to higher wages and benefits for home care workers in California’s Bay Area, prompting institutional changes in a significant number of other states (Howes 2005; Boris and Klein 2015). Likewise, teachers’ unions have been in the forefront of efforts to develop the concept of “bargaining for the common good” (McCartin 2016).

A recent wave of successful teacher strikes in predominantly Republican states drives this point home. In spring 2018, walkouts and related political actions in three traditionally red states—West Virginia, Oklahoma, and Arizona—won stunning victories, including significant salary boosts and increased budget commitments to public education (Blanc 2019). Local activists took the lead, with considerable union support, and spillover effects were felt in many other locales. Teachers had good reason to mobilize—their pay, low to begin with (especially by international standards) has stagnated for more than 10 years (Hanushek, Piopiunik, and Wiederhold 2019; Allegretto and Mishel 2019). Turnover rates are extremely high among new teachers, a result of poor working conditions, poor pay and benefits, and overemphasis on testing and accountability measures (Garcia and Weiss 2019).

Women constitute a growing majority of public school teachers, with their representation reaching 76% in 2015–2016 compared with 67% in 1980–1981 (Ingersoll et al. 2018). And women have proved adept at developing a bargaining-for-the-common-good strategy, building on their experience in a typically undervalued and underpaid feminine profession (Bhattacharya 2018). The red state actions built on the experience of the Chicago Teachers Union in 2012 in its actions against Mayor Rahm Emanuel (McCartin 2016). The teachers reached out to parents and local community members, arguing that better pay for teachers, linked to more generous funding for public education, would contribute to better outcomes for students. Evidence suggests that many parents affected by these actions, far from feeling aggrieved, became more sympathetic to—and interested in—workplace activism (Hertel-Fernandez, Naidu, and Reich 2019).

Teachers and staff worked together successfully, often relying on social media as a way of conducting strategic discussions as well as explaining their actions. Eric Blanc notes that, “One of the strikers’ secrets to success was that they consistently raised political demands—for examples, massively increased school funding—that lay outside the restricted bounds of normal collective bargaining” (Blanc 2019, 78). When the Oklahoma legislature offered salary concessions in an attempt to avert school closings, employees walked anyway, putting priority on the overall funding issue. Parents had good reason to support them.

Paid work/family work constraints

Plenty of media attention has been devoted to “work/family conflicts,” but the phrase itself betrays the assumption that family responsibilities don’t entail work. They do. Data from the American Time Use Survey, conducted annually since 2003, show that, on average, people devote about as much time to unpaid as to paid work, and the temporal constraints imposed by the care of dependents involve both active care and supervisory or on-call responsibilities (BLS 2019b). Responsibility for the care of family, friends, and neighbors weighs more heavily on women than men, not because women necessarily prefer this arrangement but because men often have sufficient bargaining power to minimize demands on their time. Women labeled as uncaring are typically stigmatized. Employers use this social norm to justify lower pay offers to women.

Gender, however, doesn’t tell the whole story. Employers who prefer “ideal workers” unencumbered by care responsibilities often have sufficient bargaining power to indulge this preference. Workers who would prefer “ideal employers,” on the other hand, are seldom able to find them. Many legal complaints of “family responsibility discrimination” have been filed by men (Williams et al. 2012). Private provision of family-friendly benefits has expanded in recent years, but firms often limit these to upper-echelon workers, leaving others unprotected (Gerstel and Clawson 2014). Public subsidies for child care are limited to means-tested assistance for the poor and modest tax benefits for others. Heather Boushey puts it this way on the cover of her book Finding Time (2016):

…business and government treat the most important things in life―health, children, elders―as matters for workers to care about entirely on their own time and dime. That might have worked in the past, but only thanks to a hidden subsidy: the American Wife, a behind-the-scenes, stay-at-home fixer of what economists call market failures. When women left the home―out of desire and necessity―the old system fell apart. Families and the larger economy have yet to recover.

The 1993 Family and Medical Leave Act, which guarantees rights to unpaid leave, covers only about half of U.S. workers, and many of those covered can’t afford to take it. No federal protections for paid sick leave from work have been put into place, despite evidence of threats to public health that became particularly visible in the early stages of the coronavirus pandemic. Employer opposition to paid sick days, paid parental leave, paid vacation, overtime, and scheduling regulation remains fierce.

Still, over the past two decades, efforts to regulate—through initiatives such as paid family leave and sick leave—the ability of employers to impose inflexible hours of work on women and men have enjoyed some success on the state level, increasing awareness of the advantages of the more family-friendly Nordic model (Maume 2016). California, New Jersey, Rhode Island, and New York have implemented paid family leave policies, and Massachusetts, Connecticut, Oregon, and Washington are committed to doing so in the near future. These gains are harbingers of future possibilities.

Conclusion

Social institutions solidify differences in collective bargaining power, making institutional change difficult to achieve. Emphasis on collective bargaining power does not imply that men and women are engaged in a zero-sum game where gains and losses cancel out. Processes of implicit and explicit negotiation often speed adaptation to new economic conditions and improve the terms of cooperation. However, long-run gains are often less visible—and less certain—than short-run losses. Furthermore, disempowered groups are diverse, making it difficult for them to act in concert. By contrast, powerful elites are fairly homogeneous, enjoying overlaps of many forms of privilege they are reluctant to relinquish. Employers have significant incentives to continue to pay women less than men.

The disadvantages women continue to experience in the labor market cannot be blamed on their own choices. Nor can they simply be attributed to inherent trade-offs between paid work and family care. As Claudia Goldin and Lawrence Katz (2016) observe, employers offer an hourly pay premium to workers willing to work exceptionally long hours, a factor that puts many women at a disadvantage. But while such long hours may be profitable for employers, they are not more productive for the economy as a whole. Indeed, the premium for long hours sends the message that time devoted to family care is a costly preference and an expensive luxury, rather than an important economic contribution.

Choices to collaborate with others to challenge unfair social institutions are among the most important choices people make. Such commitments, like investments in new technology, can be risky, but they also yield rich rewards for everyone. Like other disempowered groups, women are often able to overcome their differences, find allies, and bargain for change. Feminist efforts have never proved entirely successful, and their relative gains have been unevenly distributed. Nonetheless, the history of hard-won but cumulative successes challenges mainstream economic thinking and validates the rallying power of appeals to social well-being rather than private profit.

Endnotes

1. A particularly important exception is Jack Hirshleifer’s classic essay, “The Dark Side of the Force” (2001), which summarizes many points explored in more detail by conflict theorists such as Garfinkel and Skaperdas (2000).

References

Alderman, Naomi. 2017. The Power. Boston: Little Brown.

Allegretto, Sylvia, and Lawrence Mishel. 2019. The Teacher Weekly Wage Penalty Hit 21.4 Percent in 2018, a Record High. Washington, D.C.: Economic Policy Institute.

Amott, Teresa, and Julie Matthaei. 1999. Race, Gender and Work: A Multi-Cultural Economic History of Women in the United States. Boston: South End Press.

Andersen, Steffen, Seda Ertac, Uri Gneezy, John A. List, and Sandra Maximiano. 2013. “Gender, Competitiveness, and Socialization at a Young Age: Evidence from a Matrilineal and a Patriarchal Society.” Review of Economics and Statistics 95, no. 4:1438–43.

Arrow, Kenneth. 1974. “The Theory of Discrimination.” In Orley Ashenfelter and Albert Rees, eds., Discrimination in Labor Markets. Princeton, N.J.: Princeton University Press, pp. 1–33.

Badgett, M. V. Lee, and Nancy Folbre. 1999. “Assigning Care: Gender Norms and Economic Outcomes.” International Labour Review 138, no. 3: 311–26.

Badgett, M. V. Lee, and Nancy Folbre. 2003. “Job Gendering: Occupational Choice and the Marriage Market.” Industrial Relations: A Journal of Economy and Society 42, no. 2: 270–98.

Badgett, M.V., Brad Sears, Holning Lau, and Deborah Ho. 2009. “Bias in the Workplace: Consistent Evidence of Sexual Orientation and Gender Identity Discrimination 1998-2008.”Chicago-Kent Law Review 84, no. 2, Rev. 559.

Bailey, Martha J., Brad Hershbein, and Amalia R. Miller. 2012. “The Opt-In Revolution? Contraception and the Gender Gap in Wages.” American Economic Journal: Applied Economics 4, no. 3: 225–54.

Barron, David N., and Elizabeth West. 2013. “The Financial Costs of Caring in the British Labour Market: Is There a Wage Penalty for Workers in Caring Occupations?” British Journal of Industrial Relations 51, no. 1: 104–23.

Baum, Bernadette. 2019. “Workplace Sexual Harassment in the ‘Me Too’ Era: The Unforeseen Consequences of Confidential Settlement Agreements.” Journal of Business and Behavioral Sciences 31, no. 1: 4–24.

Bayer, Patrick, and Kerwin Kofi Charles. 2018. “Divergent Paths: A New Perspective on Earnings Differences Between Black and White Men Since 1940.” Quarterly Journal of Economics 133, no. 3: 1459-1501.

Becker, Gary. 1957. The Economics of Discrimination. Chicago: University of Chicago Press.

Becker, Gary. 1991. A Treatise on the Family. Enlarged edition. Cambridge, Mass.: Harvard University Press.

Bessen, James, Erich Denk, and Chen Meng. 2020. Perpetuating Inequality: What Salary History Bans Reveal About Wages. Boston University School of Law, Public Law and Legal Theory Paper no. 20-19.

Bertrand, Marianne, Emir Kamenica, and Jessica Pan. 2015. “Gender Identity and Relative Income Within Households.” Quarterly Journal of Economics 130, no. 2: 571–614.

Biddle, J. E., and D. S. Hamermesh. 2013. “Wage Discrimination Over the Business Cycle.” IZA Journal of Labor Policy 2, no. 7.

Bittman, Michael, Paula England, Liana Sayer, Nancy Folbre, and George Matheson. 2003. “When Does Gender Trump Money? Bargaining and Time in Household Work.” American Journal of Sociology 109, no. 1: 186–214.

Bhattacharya, Tithi. 2018. “Women Are Leading the Wave of Strikes in America. Here’s Why.” The Guardian, April 10.

Bjorvatn, Kjetil, Ranveig Falch, and Ulrikke Hernæs. 2016. “Gender, Context, and Competition: Experimental Evidence from Rural and Urban Uganda.” Journal of Behavioral and Experimental Economics 61: 31–37.

Blanc, Eric. 2019. Red State Revolt. New York: Verso.

Blau, Francine D., and Lawrence M. Kahn. 2017. “The Gender Wage Gap: Extent, Trends, and Explanations.” Journal of Economic Literature 55, no. 3: 789–865.

BLS (Bureau of Labor Statistics). 2019a. Employer Costs for Employee Compensation.

BLS (Bureau of Labor Statistics). 2019b. “Table A-1. Time Spent in Detailed Primary Activities and Percent of the Civilian Population Engaging in Each Activity, Average Per Day by Sex, 2018 Annual Averages.” American Time Use Survey Tables.

Boeckmann, Irene, Joya Misra, and Michelle J. Budig. 2015. “Cultural and Institutional Factors Shaping Mothers’ Employment and Working Hours in Postindustrial Countries.” Social Forces 93, no. 4: 1301–33.

Bonacich, Edna. 1972. “A Theory of Ethnic Antagonism: The Split Labor Market.” American Sociological Review 37, no. 5: 547–59.

Boris, Eileen, and Jennifer Klein. 2015. Caring for America: Home Health Workers in the Shadow of the Welfare State. New York: Oxford University Press.

Bornstein, Stephanie, Joan C. Williams, and Genevieve R. Painter. 2012. “Discrimination Against Mothers Is the Strongest Form of Workplace Gender Discrimination: Lessons from U.S. Caregiver Discrimination Law.” International Journal of Comparative Labour Law and Industrial Relations 28, no. 1: 45–62.

Boserup, Ester. 1970. Women’s Role in Economic Development. London: George Allen and Unwin, 1970.

Boushey, Heather. 2016. Finding Time. Cambridge, Mass.: Harvard University Press.

Bowles, Hannah Riley, Linda Babcock, and Lei Lai. 2007. “Social Incentives for Gender Differences in the Propensity to Initiate Negotiations: Sometimes It Does Hurt to Ask.” Organizational Behavior and Human Decision Processes 103, no. 1: 84–103.

Budig, Michelle J., and Paula England. 2001. “The Wage Penalty for Motherhood.” American Sociological Review 66, no. 2: 204–25.

Budig, Michelle J., and Joya Misra. 2011. “How Care‐Work Employment Shapes Earnings in Cross‐National Perspective,” International Labour Review 149, no. 4: 441–60.

Budig, Michelle J., Joya Misra, and Irene Boeckmann. 2012. “The Motherhood Penalty in Cross-National Perspective: The Importance of Work–Family Policies and Cultural Attitudes.” Social Politics 19, no. 2: 163–93.

Carrillo, Dani, Kristen Harknett, Allison Logan, Sigrid Luhr, and Daniel Schneider. 2017. “Instability of Work and Care: How Work Schedules Shape Child-Care Arrangements for Parents Working in the Service Sector.” Social Service Review 91, no. 3: 422–55.

Cassar, Alessandra, Feven Wordofa, and Y. Jane Zhang. 2016. “Competing for the Benefit of Offspring Eliminates the Gender Gap in Competitiveness.” Proceedings of the National Academy of Sciences 113, no. 19: 5201–05.

Cha, Youngjoo, and Kim A. Weeden. 2014. “Overwork and the Slow Convergence in the Gender Gap in Wages.” American Sociological Review 79, no. 3: 457–84.

Charles, Maria, and David B. Grusky. 2005. Occupational Ghettos. Stanford, Calif.: Stanford University Press.

Chetty, Raj, Nathaniel Hendren, Maggie R. Jones, and Sonya R. Porter. 2019. “Race and Economic Opportunity in the United States: An Intergenerational Perspective.” Working Paper No. 24441. Cambridge, Mass.: National Bureau of Economic Research. https://www.nber.org/papers/w24441.pdf.

Cobble, Dorothy Sue. 2005. The Other Women’s Movement: Workplace Justice and Social Rights in Modern America. Princeton, N.J.: Princeton University Press.

Correll, Shelley J., Stephen Benard, and In Paik. 2007. “Getting a Job: Is there a Motherhood Penalty?” American Journal of Sociology 112, no. 5: 1297–338.

Croson, Rachel, and Uri Gneezy. 2009. “Gender Differences in Preferences.” Journal of Economic Literature 47, no. 2: 448–74.

Darity, William Jr. 2005. “Stratification Economics: The Role of Intergroup Inequality.” Journal of Economics and Finance 29, no. 2: 144–53.

DOL (Department of Labor). 2014. “Pay Secrecy.” Women’s Bureau Fact Sheet. Washington, D.C.: DOL.

Economic Policy Institute (EPI). 2020. Gender wage gap data from State of Working America Data Library. Last updated February 20, 2020.

England, Paula. 1992. Comparable Worth: Theories and Evidence. Piscataway, N.J.: Transaction Publishers.

England, Paula, Paul Allison, and Yuxiao Wu. 2007. “Does Bad Pay Cause Occupations to Feminize, Does Feminization Reduce Pay, and How Can We Tell with Longitudinal Data?” Social Science Research 36, no. 3: 1237–56.

England, Paula, Michelle Budig, and Nancy Folbre. 2002. “Wages of Virtue: The Relative Pay of Care Work.” Social Problems 49, no. 4: 455–73.

Feldberg, R.L. 1984. “Comparable Worth: Toward Theory and Practice in the United States.” Signs: Journal of Women in Culture and Society 10, no. 2: 311–28.

Folbre, Nancy. 2006. “Demanding Quality: Worker/Consumer Coalitions and ‘High Road’ Strategies in the Care Sector.” Politics and Society 34, no. 1 (March): 1–21.

Folbre, Nancy. 2012. “Should Women Care Less? Intrinsic Motivation and Gender Inequality.” British Journal of Industrial Relations 50, no. 4: 597–619.

Folbre, Nancy. 2016. “Just Deserts? Earnings Inequality and Bargaining Power in the U.S. Economy.” Working Paper. Washington, D.C.: Washington Center for Equitable Growth.

Folbre, Nancy. 2017. “The Care Penalty and Gender Inequality.” In S. Averett, L. Argys, and S. Hoffman, eds., The Oxford Handbook of Women and Economy, pp. 1–28. New York: Oxford University Press.

Folbre, Nancy. 2021 (forthcoming). The Rise and Decline of Patriarchal Systems: An Intersectional Political Economy. New York: Verso.

Folbre, Nancy, Leila Gautham, and Kristin Smith. 2020. “The Covid-19 Care Penalty.” Manuscript under review at Feminist Economics, available from first author at Department of Economics, University of Massachusetts Amherst.

Folbre, Nancy, and Kristin Smith. 2020. “Preferences, Bargaining, and Pay in the U.S. Care Sector.” Manuscript under review at Social Problems, available from first author at Department of Economics, University of Massachusetts Amherst.

Fortin, Nicole M. 2008. “The Gender Wage Gap Among Young Adults in the United States: The Importance of Money Versus People.” Journal of Human Resources 43, no. 4: 884–918.

Fortin, N.M., B. Bell, and M. Böhm. 2017. “Top Earnings Inequality and the Gender Pay Gap: Canada, Sweden, and the United Kingdom.” Labour Economics 47: 107–23.

Fuchs, Victor. 1988. Women’s Quest for Economic Equality. Cambridge, Mass.: Harvard University Press.

Garcia, Emma, and Elaine Weiss. 2019. The Teacher Shortage Is Real, Large, And Growing, And Worse Than We Thought. Economic Policy Institute, March.

Garcia, Sandra E. 2017. “The Woman Who Created #MeToo Long Before Hashtags,” New York Times, October 20.

Garfinkel, Michelle, and Stergios Skaperdas. 2000. “Contract or War? The Consequences of Broader View of Self-Interest in Economics.” American Economist 44, no. 1, 5–16.

Geddes, Rick, and Dean Lueck. 2002. “The Gains from Self-Ownership and the Expansion of Women’s Rights.” American Economic Review 92, no. 4: 1079–92.

Gerstel, Naomi, and Dan Clawson. 2014. “Class Advantage and the Gender Divide: Flexibility on the Job and at Home” American Journal of Sociology 120, no. 2: 395–431.

Gneezy, Uri, Kenneth L. Leonard, and John A. List. 2009. “Gender Differences in Competition: Evidence from a Matrilineal and a Patriarchal Society.” Econometrica 77, no. 5: 1637–64.

Goff, Isabel Boni-Le, and Nicky Le Feuvre. 2017. “Professions from a Gendered Perspective.” Oxford Research Encyclopedia of Business and Management. New York: Oxford University Press.

Golden, Lonnie. 1996. “The Economics of Worktime Length, Adjustment, and Flexibility: A Synthesis of Contributions from Competing Models of the Labor Market.” Review of Social Economy 54, no. 1: 1–45.

Golden, Lonnie. 2020. “Part-Time Pay Penalties Persisting.” Paper presented at the meetings of Allied Social Science/Labor and Employment Relations Association, San Diego, Calif., January 4.

Goldin, Claudia. 1990. Understanding the Gender Gap: An Economic History of American Women. New York: Oxford University Press.

Goldin, Claudia. 2014. “A Grand Gender Convergence: Its Last Chapter.” American Economic Review 104, no. 4: 1091–119.

Goldin, Claudia, and Lawrence F. Katz. 2016. “A Most Egalitarian Profession: Pharmacy and the Evolution of a Family-Friendly Occupation.” Journal of Labor Economics 34, no. 3: 705–46.

Goldin, Claudia, and Cecilia Rouse. 2000. “Orchestrating Impartiality: The Impact of ‘Blind’ Auditions on Female Musicians.” American Economic Review 90, no. 4: 715–41.

Gorsuch, Marina Mileo. 2019. “Gender, Sexual Orientation, and Behavioral Norms in the Labor Market.” Industrial & Labor Relations Review 72, no. 4: 927–54.

Green, Stuart P. 2020. “The Looming Question in Rape Law After Harvey Weinstein.” Wall Street Journal, January 3.

Greenman, E. and Y. Xie. 2008. “Double Jeopardy? The Interaction of Gender and Race on Earnings in the United States.” Social Forces 86, no. 3: 1217–44.

Greenwood, Jeremy, Ananth Seshadri, and Mehmet Yorukoglu. 2005. “Engines of Liberation.” Review of Economic Studies 72, no. 1: 109–33.

Gupta, Sarita, Stephen Lerner, and Joseph A. McMartin. 2019. “Why the Labor Movement Has Failed and How to Fix It.” Boston Review, June 6.

Hakim, Catherine. 2000. Work-Lifestyle Choices in the 21st Century: Preference Theory. New York: Oxford University Press.

Hanushek, E.A., M. Piopiunik, and S. Wiederhold. 2019. “Do Smarter Teachers Make Smarter Students? International Evidence on Teacher Cognitive Skills and Student Performance.” Education Next 19, no. 2: 56–65.

Harkness, Susan, and Jane Waldfogel. 2003. “The Family Gap in Pay: Evidence from Seven Industrialized Countries.” Research in Labor Economics 22, no. 1: 369–414.

Hertel-Fernandez, Alexander, Suresh Naidu, and Adam Reich. 2019. “Do Teacher Strikes Make Parents Pro- or Anti-Labor? The Effects of Labor Unrest on Mass Attitudes.” Working Paper. Washington Center for Equitable Growth.

Hill, Herbert. 1996. “The Problem of Race in American Labor History.” Reviews in American History 24, no. 2: 189–208.

Hirsch, Barry T., and Julia Manzella. 2015. ″Who Cares—And Does it Matter? Measuring the Wage Penalty for Caring Work.” In Solomon W. Polachek, Konstantinos Tatsiramos, and Klaus F. Zimmermann, eds., Gender Convergence in the Labor Market, 213–275. Somerville, Mass.: Emerald Insight Publishing Limited.

Hirshleifer, Jack. 2001. The Dark Side of the Force. Economic Foundations of Conflict Theory. New York: Cambridge University Press.

Hodges, Melissa, Michelle Budig, and Paula England. 2018. “Wages of Nurturant and Reproductive Care Workers: Individual and Job Characteristics, Occupational Closure, and Wage-Equalizing Institutions.” Social Problems 66, no. 2: 294–319.

Hoff, Karla, and James Walsh. Forthcoming. “The Third Function of Law Is to Transform Cultural Categories.” In Kaushik Basu and Robert Hockett, eds., Law, Economics, and Conflict. Ithaca, N.Y.: Cornell University Press.

Holzer, Harry J. 1995. “Comparable Worth: Is it a Worthy Policy? By Elaine Sorensen” Journal of Policy Analysis and Management 14, no. 2: 356–59.

Howes, Candace. 2005. “Living Wages and Retention of Homecare Workers in San Francisco.” Industrial Relations 44, no. 1: 139–63.

ILO (International Labour Organization). 2018. Care Work and Care Jobs for the Future of Decent Work. Geneva: ILO.

Ingersoll, R., E. Merrill, D. Stuckey, and G. Collins. 2018. Seven Trends: The Transformation of the Teaching Force. Philadelphia.: Consortium for Policy Research in Education, University of Pennsylvania.

Institute for Women’s Policy Research (IWPR). 2019. “Women’s Median Earnings as a Percent of Men’s Median Earnings, 1960-2018 (Full-Time, Year-Round Workers) with Project for Pay Equity in 2059.” Quick Figures #Q076.

Katznelson, Ira. 2006. When Affirmative Action Was White: An Untold History of Racial Inequality in Twentieth-Century America. New York: W.W. Norton.

Kijakasi, Kilolo. 2019. “Bold, Equitable Policy Solutions Are Needed to Close the Racial and Gender Wealth Gaps.” Testimony before the Subcommittee on Diversity and Inclusion, Financial Services Committee, United States House of Representatives, September 24.

Killingsworth, M.R. 1987. “Heterogeneous Preferences, Compensating Wage Differentials, and Comparable Worth.” Quarterly Journal of Economics 102, no. 4: 727–42.

Kim, Marlene. 1999. “Inertia and Discrimination in the California State Civil Service.” Industrial Relations: A Journal of Economy and Society 38, no. 1: 46–68.

Kim, Marlene. 2015. “Pay Secrecy and the Gender Wage Gap in the United States.” Industrial Relations: A Journal of Economy and Society 54, no. 4: 648–67.

Knowles, R. 2013. “The Intertwined Fates of Affirmative Action and the Military.” Loyola University Chicago Law Review 45: 1027.

Kurtulus, Fidan Ana. 2016. “The Impact of Affirmative Action on the Employment of Minorities and Women: A Longitudinal Analysis Using Three Decades of EEO‐1 Filings.” Journal of Policy Analysis and Management 35, no. 1: 34–66.

Lambert, Susan J. 2008. “Passing the Buck: Labor Flexibility Practices That Transfer Risk onto Hourly Workers.” Human Relations 61, no. 9: 1203–27.

Laurison, D. and S. Friedman. 2016. “The Class Pay Gap in Higher Professional and Managerial Occupations.” American Sociological Review 81, no. 4: 668–95.

Lepore, Jill. 2018. These Truths: A History of the United States. New York: W.W. Norton.

Levinson, M. 2019. “Mandatory Arbitration: How the Current System Perpetuates Sexual Harassment Cultures in the Workplace.” Santa Clara Law Review 59: 485–524.

Lott, Yvonne, and Christina Klenner. 2018. “Are the Ideal Worker and Ideal Parent Norms About to Change? The Acceptance of Part-Time and Parental Leave at German Workplaces.” Community, Work & Family 21, no. 5: 564–80.

Mankiw, N. Gregory. 2013. “Defending the One Percent” Journal of Economic Perspectives 27, no. 3: 21–34.

Maume, David J. 2016. “Can Men Make Time for Family? Paid Work, Care Work, Work-Family Reconciliation Policies, and Gender Equality.” Social Currents 3, no. 1: 43–63.

McCall, Leslie. 2001. Complex Inequality: Gender, Class, and Race in the New Economy. London: Psychology Press.

McCartin, Joseph A. 2016. “Bargaining for the Common Good.” Dissent, 63, no. 2: 128–35.

McClintock, Elizabeth Aura. 2020. “Occupational Sex Composition and Marriage: The Romantic Cost of Gender-Atypical Job.” Journal of Marriage and Family 82, no. 3: 911–33.

McGregor, Jena. 2019. “More States Are Banning Questions About Salary History in Job Interviews. What to Say if You’re Asked About It Anyway.” Washington Post, August 15.

McGuire, John Thomas. 2008. “The Most Unjust Piece of Legislation: Section 213 of the Economy Act of 1932 and Feminism During the New Deal.” Journal of Policy History 20, no. 4: 516–41.

McLaughlin, H., C. Uggen, and A. Blackstone. 2017. “The Economic and Career Effects of Sexual Harassment on Working Women.” Gender & Society 31, no. 3: 333–58.

Miller, Conrad. 2017. “The Persistent Effect of Temporary Affirmative Action.” American Economic Journal: Applied Economics 9, no. 3: 152–90.

Miller, Grant. 2008. “Women’s Suffrage, Political Responsiveness, and Child Survival in American History.” Quarterly Journal of Economics no. 3, 1287–1327.

Morsy, Leila, and Richard Rothstein. 2015. “Parents’ Non-Standard Work Schedules Make Adequate Childrearing Difficult: Reforming Labor Market Practices Can Improve Children’s Cognitive and Behavioral Outcomes.” Issue Brief #400. Washington: Economic Policy Institute.

Mueller, Gerrit, and Erik Plug. 2006. “Estimating the Effect of Personality on Male and Female Earnings.” Industrial & Labor Relations Review 60, no. 1: 3–22.

Murray-Close, Marta, and Misty L. Heggeness. 2018. “Manning Up and Womaning Down: How Husbands and Wives Report Their Earnings When She Earns More.” Working Paper 2018-20. Washington: U.S. Census Bureau Social, Economic, and Housing Statistics Division. https://doi.org/10.21034/iwp.28.

Nelson, Julie A. 2016. “Not-so-Strong Evidence for Gender Differences in Risk Taking.” Feminist Economics 22, no. 2: 114–42.

Neumark, David, Roy J. Bank, and Kyle D. Van Nort. 1996. “Sex Discrimination in Restaurant Hiring: An Audit Study.” Quarterly Journal of Economics 111, no. 3: 915–41.

Nunn, Nathan, Alberto F. Alesina, and Paola Giuliano. 2013. “On the Origins of Gender Roles: Women and the Plough.” Quarterly Journal of Economics 128, no. 2: 469–530.

Orentlicher, D. 2016. “Economic Inequality and College Admissions Policies.” Cornell Journal of Law and Public Policy 26: 101.

Pal, Ipshita, and Jane Waldfogel. 2016. “The Family Gap in Pay: New Evidence for 1967 to 2013.” RSF: The Russell Sage Foundation Journal of the Social Sciences 2, no. 4: 104–27.

Pietrykowski, Bruce. 2017. “The Return to Caring Skills: Gender, Class, and Occupational Wages in the U.S.” Feminist Economics 23, no. 4: 32–61.

Rauch, Jonathan. 2020. “Did the Civil Rights Movement Go Wrong?” New York Times, January 17.

Rich, Charlotte J. 2009. Transcending the New Woman: Multiethnic Narratives in the Progressive Era. Columbia: University of Missouri Press.

Rose, Stephen, and Heidi Hartmann. 2004. “The Long-Term Gender Gap.” Challenge 47, no. 5: 30–50.

Rosen, S. 1986. “The Theory of Equalizing Differences.” Handbook of Labor Economics 1: 641–92.

Schloesser, Pauline E. 2002. The Fair Sex: White Women and Racial Patriarchy in the Early American Republic. New York: New York University Press.

Schwab, Stewart. 1986. “Is Statistical Discrimination Efficient?” American Economic Review 76, no. 1: 228–34.

Sorensen, Elaine. 1994. Comparable Worth: Is It a Worthy Policy? Princeton, N.J.: Princeton University Press.

Stigler, George J., and Gary S. Becker. 1977. “De Gustibus Non Est Disputandum.” American Economic Review 67, no. 2: 76–90.

Tax, Meredith 2001. The Rising of the Women: Feminist Solidarity and Class Conflict, 1880-1917. Urbana: University of Illinois Press.

Thomas, Gillian. 2016. Because of Sex: One Law, Ten Cases, and Fifty Years that Changed American Women’s Lives at Work. New York: St. Martin’s Press.

Tilcsik, András. 2011. “Pride and Prejudice: Employment Discrimination Against Openly Gay Men in the United States.” American Journal of Sociology 117, no. 2: 586–626.

Treiman, Donald J., and Heidi I. Hartmann, eds. 1981. Women, Work and Wages: Equal Pay for Jobs of Equal Value. Washington: National Academy Press.

Wall Street Journal Editorial Board. 2019. “The ‘Wage Gap’ Commissars.” Wall Street Journal, May 27.

Webber, D. A. 2016. “Firm‐Level Monopsony and the Gender Pay Gap.” Industrial Relations 55, no. 2: 323–45.

Weil, David. 2017. The Fissured Workplace. Cambridge, Mass.: Harvard University Press.

Williams, Joan C., Robin Devaux, Patricija Petrac, and Lynn Feinberg. 2012. “Protecting Family Caregivers from Employment Discrimination.” American Association for Retired Persons, Insight on the Issues Report 68.

Williams, Rhonda M. 1987. “Capital, Competition, and Discrimination: A Reconsideration of Racial Earnings Inequality.” Review of Radical Political Economics 19, no. 2: 1–15.

Witteveen, D. and P. Attewell. 2017. “Family Background and Earnings Inequality Among College Graduates.” Social Forces 95, no. 4: 1539–76.

Yavorsky, Jill E., Lisa A. Keister, Yue Qian, and Michael Nau. 2019. “Women in the One Percent: Gender Dynamics in Top Income Positions.” American Sociological Review 84, no. 1: 54–81.