New data helps illuminate why these banks—and Wells Fargo in particular—prefer forced arbitration to class action lawsuits. We already knew that consumers obtain relief regarding their claims in just 9 percent of disputes, while arbitrators grant companies relief in 93 percent of their claims.2 But not only do companies win the overwhelming majority of claims when consumers are forced into arbitration—they win big.

Some crucial background helps illustrate the stakes. In July 2017, the Consumer Financial Protection Bureau (CFPB) issued a final rule to restore consumers’ ability to join together in class action lawsuits against financial institutions.3 Based on five years of careful study, the final rule stems from a congressional directive instructing the agency to study forced arbitration and restrict or ban the practice if it harms consumers.

In recent weeks, members of Congress have introduced legislation to repeal the CFPB rule and take away consumers’ newly restored right to band together in court.4 Opponents of the rule have suggested that the bureau’s own findings show consumers on average receive greater relief in arbitration ($5,389) than class action lawsuits ($32).5 As we have previously shown, this is enormously misleading.6 While the average consumer who wins a claim in arbitration recovers $5,389, this is not even close to a typical consumer outcome. Because consumers win so rarely, the average consumer ends up paying financial institutions in arbitration—a whopping $7,725.

A recent report released by the nonprofit Level Playing Field hones in on Wells Fargo’s use of arbitration in consumer claims.7 Compiling publicly reported data from the American Arbitration Association (AAA) and JAMS (initially named Judicial Arbitration and Mediation Services, Inc.), the report found that just 250 consumers arbitrated claims with Wells Fargo between 2009 and the first half of 2017.8 This number is surprisingly small, since this period spans the prime years of the bank’s fraudulent account scandal.

But we can take this data a step further by looking at Wells Fargo’s overall gains and losses in arbitration. As one might suspect based on the CFPB data, Wells Fargo indeed won more money in arbitration between 2009 and the first half of 2017 than it paid out to consumers, despite creating 3.5 million fraudulent accounts during that same period.9

What is even more troubling is that forced arbitration seems to be significantly more lucrative for Wells Fargo than for other financial institutions. In arbitration with Wells Fargo, the average consumer is ordered to pay the bank nearly $11,000. We calculated a mean of $10,826 awarded to the bank across all claims in the Level Playing Field report.

No wonder Wells Fargo prefers forced arbitration to class action lawsuits, which return at least $440 million, after deducting all attorneys’ fees and court costs, to 6.8 million consumers in an average year.10 Banning consumer class actions lets financial institutions keep hundreds of millions of dollars that would otherwise go back to harmed consumers—and Wells Fargo seems to have harmed huge numbers of consumers.

Opponents of the CFPB’s arbitration rule argue that allowing consumers to join together in court will increase consumer costs and decrease available credit. Most recently, the Office of the Comptroller of the Currency (OCC) claimed that restoring consumers’ right to join together in court could cause interest rates on credit cards to rise as much as 25 percent.11

However, examining the OCC’s study, it appears the agency merely duplicated the conclusion reached by the CFPB and based its 25 percent estimate solely on results it admits are “statistically insignificant at the 95 percent (and 90 percent) confidence level.”12 In its 2015 study, the CFPB considered this same data and accurately assessed that there was no “statistically significant evidence of an increase in prices among those companies that dropped their arbitration clauses.”13

Perhaps more importantly, claims that the arbitration rule will increase consumer and credit costs are also contradicted by real-life experience. Consumers saw no increase in prices after Bank of America, JPMorgan Chase, Capital One, and HSBC dropped their arbitration clauses as a result of court-approved settlements, and mortgage rates did not increase after Congress banned forced arbitration in the mortgage market.14 Of course, many would argue that banks like Wells Fargo should bear any increase in cost associated with making consumers whole for egregious misconduct.

Once again, the numbers are clear: class actions return hundreds of millions in relief to consumers, while forced arbitration pays off big for lawbreakers like Wells Fargo.

Endnotes

1. Michael Hiltzik, “No Surprise: Wells Fargo Is Leveraging Its Arbitration Clause to Win an Advantageous Scandal Settlement,” Los Angeles Times, March 31, 2017; Associated Press, “Wells Fargo Wants Court to Toss Overdraft Lawsuits and Let It Use Arbitration,” Los Angeles Times, August 24, 2017.

2. Heidi Shierholz, Correcting the Record: Consumers Fare Better under Class Actions Than Arbitration, Economic Policy Institute, August 1, 2017.

3. Consumer Financial Protection Bureau, “New Protections against Mandatory Arbitration,” web page accessed July 31, 2017.

4. Sylvan Lane, “GOP Lawmakers Introduce Measures to Repeal Consumer Bureau Arbitration Rule,” The Hill, July 20, 2017.

5. U.S. Senate Committee on Banking, Housing, and Urban Affairs, “Senators File Resolution Disapproving of CFPB Arbitration Rule” [majority press release], July 20, 2017.

6. Heidi Shierholz, Correcting the Record: Consumers Fare Better under Class Actions Than Arbitration, Economic Policy Institute, August 1, 2017.

7. Level Playing Field, Wells Fargo and Forced Consumer Arbitration: September 2017 Update.

8. To my knowledge, AAA and JAMS are the only firms that routinely provide arbitration services to Wells Fargo. In arbitration agreements, Wells Fargo typically designates AAA as the arbitration firm to arbitrate any consumer dispute.

9. Stacy Cowley, “Now Wells Fargo up to 3.5 Million Fraudulent Accounts,” San Francisco Chronicle, August 31, 2017.

10. Heidi Shierholz, Correcting the Record: Consumers Fare Better under Class Actions Than Arbitration, Economic Policy Institute, August 1, 2017.

11. Renae Merle, “U.S. Chamber of Commerce Suing to Block Rule Allowing Consumers to Sue Their Banks,” Washington Post, September 29, 2017.

12. Office of the Comptroller of the Currency (OCC), OCC Review: Probable Cost to Consumers Resulting from the Consumer Finance Protection Bureau’s Final Rule on Arbitration Agreements, U.S. Department of the Treasury, September 20, 2017.

13. Consumer Financial Protection Bureau, Arbitration Study: Report to Congress, pursuant to Dodd–Frank Wall Street Reform and Consumer Protection Act § 1028(a), March 2015.

14. Adam J. Levitin, “Mandatory Arbitration Offers Bargain-Basement Justice,” American Banker BankThink (blog), May 13, 2014.

Introduction and key findings

Over the last three-and-a-half decades, rising inequality has been a defining feature of the American economy. The way rising inequality has directly affected most Americans is through sluggish hourly wage growth in recent decades, despite an expanding and increasingly productive economy. For example, had all workers’ wages risen in line with productivity, as they did in the three decades following World War II, an American earning around $50,000 today would instead be making close to $75,000. A hugely disproportionate share of economic gains from rising productivity is going to the top 1 percent and to corporate profits, instead of to ordinary workers—who are more productive and educated than ever. This rising inequality is largely the result of big corporations and the wealthy rewriting the rules of the economy to stack the deck in their favor. This has prevented the benefits of productivity growth from “trickling down” to reach most households.

Unfortunately, although inflation-adjusted wages grew across the board in 2015 (due to a sharp dip in inflation), the trend of rising wage inequality continued unabated last year. This paper begins by detailing the most up-to-date hourly wage trends through 2015 and then examines the continued growth in inequality that began in the late 1970s. This rising inequality is confirmed by the latest wage data, analyzed across the wage distribution and education categories, including striking differences by race and gender.

Key findings include:

• While real hourly wages (i.e., wages adjusted for inflation) grew across the board in 2015, this is largely due to a sharp dip in inflation; growth in nominal wages (i.e., wages unadjusted for inflation) has not accelerated. This dip in inflation is unlikely to be a durable source of future real wage gains.
  • Nominal wage growth of 2.2 percent remains below a level where workers would reap the benefits of economic growth.
  • There is no evidence of substantial acceleration of wages that would signal that the Federal Reserve Board should worry about incipient inflation and raise interest rates in an effort to slow the economy.

• Real hourly wage growth in 2015 was fastest at the top of the wage distribution, illustrating that wage inequality continued its 35-year rise last year.
  • The gap between the middle and bottom has remained stable since 2000.
  • The gap between the top and everyone else has grown.

• Looking at men and women separately, from 2014 to 2015, the strongest wage growth was at the top of the men’s wage distribution and at the bottom of the women’s wage distribution.
  • For men, wages at the 95th and 90th percentiles grew by 9.9 percent and 6.2 percent, respectively, compared with only 2.6 percent at the median.
  • While significant gender wage gaps remain across the wage distribution, there has been significant narrowing of the gap at the bottom and middle of the wage distribution since 2000.

• In 2015, faster low-wage wage growth occurred in states that increased their minimum wage.
  • The 10th percentile women’s wage grew 5.2 percent in states with legislated minimum-wage increases, compared with only 3.1 percent growth in states without any minimum-wage increase.

• Racial and ethnic wage gaps are on the rise at the top of the wage distribution.
  • Between 2014 and 2015, white wage growth was at least as fast as black wage growth at all deciles, and was far stronger at the top than at the middle or bottom.
  • Wage growth at the bottom of the Hispanic wage distribution since 2000 greatly exceeded that of low-wage black workers, who actually experienced losses over that period.
  • Regardless of race or ethnicity, within-group inequality has risen since 2000, with stronger growth at the top than at the bottom.

• Over 2000–2015, wage growth was faster among the more educated, but not fast enough to explain growing wage inequality.
  • For both men and women, those with less than a college degree had lower wages in 2015 than in 2007.
  • While there has been a slow narrowing of gender wage gaps for those with less than a college degree since 2000, gender wage gaps continued to grow among those with an advanced degree.

Nominal and real wage growth in 2015 illustrates just how far we remain from a full recovery

Between 2014 and 2015, average hourly nominal wages for private-sector workers grew 2.2 percent (according to the Bureau of Labor Statistics Current Employment Statistics), in line with the 2.0–2.2 percent trend over the previous six years. At that slow rate, this crucial measure illustrates just how far the economy remains from a full recovery, let alone full employment. The weakened labor market of the last seven years has put enormous downward pressure on wages. Employers still don’t have to offer substantial wage increases to attract or retain the workers they want, even as the recovery has forged ahead in recent years. In the last few months, nominal wage growth appears to be picking up slightly, but remains significantly below levels consistent with the Federal Reserve’s 2.0 percent inflation target and likely trends in potential productivity (for more on this, see EPI’s Nominal Wage Tracker).

At the same time that private-sector average hourly nominal wages grew 2.2 percent between 2014 and 2015, overall inflation fell from 1.6 percent between 2013 and 2014 to only 0.1 percent between 2014 and 2015. This is shown in Table 1, which displays a variety of measures of nominal wage growth, inflation, and real wage growth. (For this discussion, private-sector average nominal wage growth is used, shown in the first row; however, other data sets show similar wage trends.)

Table 1

Taken together, these nominal wage growth and inflation trends mean that real (inflation-adjusted) average wages grew 2.1 percent between 2014 and 2015. From a historical perspective, this is a relatively healthy rate of real wage growth. For example, real hourly wage growth for the bottom 90 percent of workers averaged less than 0.5 percent annually since 1979 (Bivens et al. 2014).

However, relying on low inflation to increase living standards is a poor long-term strategy for two major reasons. One, low inflation has been driven by large declines in commodity prices (particularly gas and oil), which are historically volatile and likely to stabilize or even rise in the next year. And, two, low permanent inflation would surely slow nominal wage growth (Bivens 2016). To see the role of commodity prices, compare overall inflation at 0.1 percent to core inflation (which removes more volatile energy and food prices) at 1.8 percent, shown in the sixth and seventh rows of Table 1. Using core inflation, most of the real wage gains disappear. To be clear, the gains to real living standards in 2015 were genuine—cheaper gas, oil, and other forms of energy all make paychecks stretch further. But gains that rely on near-zero inflation are not likely to be durable going forward.

Figure A switches to the wage series (based on the Current Population Survey Outgoing Rotation Group, or CPS-ORG) that allows measurement of wage growth at different points in the wage distribution and by demographic characteristics. The green line shows that median hourly nominal wages for all workers age 18–64 grew 1.8 percent between 2014 and 2015. As with the Current Employment Statistics series discussed above, median hourly nominal wage growth has been relatively low and steady. Using overall inflation, real wages grew by 1.7 percent in the last year. But, using the more stable core inflation measure, real wage growth disappears. It is clear that the short-term gain in inflation-adjusted wages is driven by (likely) short-term drops in commodity prices that have a positive short-term effect on living standards, but are likely not sustainable. In fact, higher inflation is a better strategy for long-term growth in living standards for a number of reasons. For example, in an economy where households are still laden with debt, higher inflation helps erode the real burden of the debt more quickly, which also spurs recovery (Bivens 2016).

Figure A

The benefits of non-zero inflation are recognized by the Federal Reserve, which targets 2.0 percent price inflation. The Federal Reserve relies on nominal wage growth as its essential measure to determine whether there are pressures from accelerating wage growth that might push inflation above its target. So far, the data are clear on this question. While private-sector hourly wages saw a slight rise throughout 2015 (EPI 2016a), any signs of acceleration in real wages between 2014 and 2015 are driven by the sharp dip in inflation in 2015, and not by nominal wage acceleration. The wage and price data over the past year reinforce that the Fed acted too soon by raising rates in December, and that real wage gains in 2015 are not evidence of an economy that has achieved genuine full employment.

Real hourly wage growth was fastest at the top of the wage distribution

Wage inequality has risen since the late 1970s—a trend that remained unchanged in 2015. The rise in wage inequality over the last three-and-a-half decades largely stems from intentional policy choices that have eroded ordinary workers’ leverage to secure higher pay (Bivens et al. 2014). These policy choices—made on behalf of those with the most economic power—include allowing the minimum wage to stagnate, eroding workers’ rights to bargain collectively, and prioritizing low inflation over low unemployment. Policies such as these have resulted in hourly pay for the vast majority of American workers stagnating despite growing economy-wide productivity, with economic gains highly concentrated at the top.

The latest data from 2015 reveal the continuation of trends experienced through the Great Recession and much of the last three-and-a-half decades: slower growth for most and faster growth for those at the top. Table 2 includes data from 2000 (the previous business cycle peak), 2007 (the most recent business cycle peak), and the two most recent years of data (2014 and 2015). In the full business cycle from 2000 to 2007, growth was relatively slow overall and relatively unequal. The gains at the 90th and 95th percentiles were higher than at the middle or bottom of the wage distribution. In fact, the middle and the bottom grew at practically the same rate over 2000–2007 as they did in the years since 2007. The ratio of wages at the 50th and 10th percentile of the wage distribution (i.e., the 50/10 wage gap, or the gap between the middle and the bottom) has remained fairly constant between 2000 and 2015, while the gaps between the 95th percentile and the 50th (the top and the middle), and the 95th and the 10th (the top and the bottom), have grown.

Table 2

Figure B illustrates the trends in wages for select deciles (and the 95th percentile), showing the cumulative change in real hourly wages between 2007 and 2015. Although the increases from 2014 to 2015 are primarily due to low inflation and not accelerating nominal wage growth (as previously discussed), the inequality story is clear. The lines clearly demonstrate that those with the highest wages have had the fastest wage growth in recent years. This trend continued quite clearly between 2014 and 2015. By 2015, the 95/10 ratio grew to 6.3 from 5.9 in 2007 and 5.5 in 2000. Similar trends are found in the 95/50 wage ratio, with those at the top pulling away.

While the data discussed here clearly show a continuation of increasing wage inequality between 2014 and 2015, the CPS-ORG does not allow analysis of wage trends within the top 5 percent of the wage distribution. Using Social Security wage data through 2014, it can be shown that the top 1 percent grew 149.4 percent, while the bottom 90 percent grew only 16.7 percent since 1979 (Mishel and Kimball 2015).

Figure B

Strongest wage growth was at the top of the men’s wage distribution and at the bottom of the women’s wage distribution

Analyzing wages at different points in the wage distribution over time can mask different outcomes between men and women. Table 3 replicates the analysis of wage deciles for men and women separately, with a comparison of gender wage disparities over 2000–2015. Figures C and D accompany this table, illustrating the cumulative percent change over 2007–2015 in real hourly wages of men and women at key wage percentiles. Long-term trends suggest that low- and middle-wage men have fared comparably poorly, and that wage gaps between the top and the middle (the 95/50 ratio) and the top and the bottom (the 95/10 ratio) have increased more for men than for women.

Table 3

Figure C, which depicts men’s wage trends, shows the striking pulling away of the 95th and 90th percentiles since 2007, as the bottom 70 percent of male workers’ wages stagnated or fell. In fact, between 2000 and 2015, wages for the bottom 60 percent of male workers were flat or declined. In the last year, growth among men was far faster at the top, with wages at the 95th and 90th percentiles rising by 9.9 percent and 6.2 percent, respectively, compared with only 2.6 percent at the median. This indicates a continued concentration of wage gains at the top.

Figure C

In general, women’s wages are lower and less unequal than men’s. However, there appears to be some slow but consistent closing of the gender wage gap for all but the highest earners. Over 2000–2015, the gender wage gap at the median fell, with median women’s wages rising from 78.0 percent to 83.3 percent of median male wages. Unfortunately, a nontrivial reason for this narrowing has been falling men’s wages. In fact, 40 percent of the closing of the gender wage gap since 1979 has been because of falling men’s wages (Gould and Davis 2015). This is clearly not a desirable strategy for achieving gender wage parity or economic security for either men or women. The gender wage gap is greatest at the 95th percentile and has not been closing recently.

Because of far lower wages at the top of the women’s distribution, the ratio between the highest and lowest wage earners is 5.6 for women, compared with 7.1 for men. As shown in Figure D, the spread of the women’s wage distribution has grown since 2007 (i.e., inequality has increased). Between 2014 and 2015, however, wage gains were more broadly shared. In fact, growth since 2014 was greatest among the lowest-wage women (though the 90th percentile was a close second).

Figure D

Wage growth at the bottom was faster in states that increased their minimum wage

Upon further investigation, the significantly higher increase in the 10th percentile women’s wage, compared with most of the rest of the women’s wage distribution, is likely related to state-level increases in the minimum wage. Women’s wages are lower than men’s at the bottom decile ($8.57 versus $9.29) and therefore may be more likely to be impacted by changes in the wage floor. Throughout 2015, though primarily at the start of the year, the minimum wage was increased in 23 states and the District of Columbia.1 Workers in states that increased their minimum wage in 2015 account for about 40 percent (39.8 percent) of the overall U.S. workforce.

Figure E displays in green the states with legislated minimum-wage increases in 2015, while those in blue had automatic increases resulting from indexing the minimum wage to inflation. While the range of indexed increases ($0.12 to $1.25) is larger than legislated increases ($0.25 to $1.00), the average (unweighted) increases were in states with legislated increases ($0.77 versus $0.28).

Figure E

A comparison of 10th percentile wage growth between states grouped by whether they had a minimum-wage increase and by the type of increase yields highly suggestive results. As shown in Figure F, when looking at 10th percentile wages by gender, growth in states without minimum-wage increases was slower than in states with any kind of minimum-wage increases. Among states with any minimum-wage increase, the growth in the 10th percentile wage was faster in states with legislated increases (which are, on average, higher than indexed increases). This holds true for both men and women at the 10th percentile. For example, the 10th percentile women’s wage grew 5.2 percent in states with legislated minimum-wage increases, compared with only 3.1 percent growth in states without any minimum-wage increase. This is a clear indication that strong labor standards can improve outcomes even when workers generally have severely reduced bargaining power.

Figure F

Racial wage gaps on the rise at the top of the distribution

Table 4 examines wage deciles (and the 95th percentile wage) for white non-Hispanic, black non-Hispanic, and Hispanic workers between 2000 and 2015. Between 2014 and 2015, white wage growth was at least as fast as black wage growth at all deciles, and was far stronger at the top than the middle or bottom. Median white wages are now at the same level they were eight years ago, and have only grown 0.3 percent annually since 2000. In 2015, wages for the bottom 60 percent of black workers are still lower than in 2000.

Table 4

Between 2014 and 2015, Hispanic wage growth was more broadly based. In fact, Hispanic wage growth was faster in the bottom half than the top half of the Hispanic wage distribution. The faster wage growth at the 10th and 20th percentiles from 2000 to 2015 for Hispanic workers greatly exceeded that of similarly low-earning black workers, who actually experienced losses over that period. The differential was so great that the lowest-earning Hispanic workers now have higher wages than the lowest-earning black workers.

The bottom half of Table 3 displays wage disparities, comparing black and Hispanic wages as a share of white wages at each decile of their respective wage distributions. Compared with white workers, black workers have been losing ground, with increasing racial wage gaps across the entire distribution. At the median in 2000, black wages were 79.4 percent of white wages. By 2015, they were only 74.8 percent of white wages. Conversely, Hispanic workers have been slowly closing the gap with white workers at the lower end of the wage distribution. Regardless of race or ethnicity, within-group inequality has risen since 2000, with stronger growth at the top than at the middle or bottom.

Wage growth faster among the more educated, but not fast enough to explain growing wage inequality

Table 5 presents the most recent data on average hourly wages by education for all workers and by gender. Because of the significant drop in inflation, average hourly wages grew significantly across the board in 2015. The strongest growth was among those with a college degree, followed by those with less than a high school degree, at 4.2 and 3.9 percent, respectively. Wages for those with a high school diploma and no further education increased as well, but to a slightly lesser degree, at 2.4 percent. Because of this differential trend, the college wage premium—the regression-adjusted log-wage difference between the wages of college- and high-school-educated workers—grew, returning to its 2013 level. Even with this recovery in 2015, the college premium has increased only modestly over the last 15 years, whereas the growth in wage inequality between percentiles (i.e., the 95/50 wage gap) has been far greater (as shown in Table 2).

Table 5

A particularly prevalent story explains wage inequality as a simple consequence of growing employer demand for skills and education—often thought to be driven by advances in technology. According to this explanation, because there is a shortage of skilled or college-educated workers, the wage gap between workers with and without a college degree is widening. This is sometimes referred to as a “skill-biased technological change” explanation of wage inequality (since it is based on technology leading to the need for more skills). However, despite its great popularity and intuitive appeal, this story about recent wage trends being driven more and more by a race between education and technology does not fit the facts well, especially since the mid-1990s (Mishel, Shierholz, and Schmitt 2013). Even among college graduates, there has been a significant pulling away of the very top. The 50th percentile wage among those with a bachelor’s degree has fallen by 3.0 percent since 2000, while the 95th percentile wage of those with a bachelor’s degree rose 40.6 percent over the same period. The story is not one of a growing differential of wages between college and high school graduates, but increasingly one of growing wage inequality overall and within various education groups.

Figures G and H display the cumulative percent change in real hourly wages by education for men and women, respectively. For both men and women, those with less than a college degree had lower wages in 2015 than in 2007 (the last business cycle peak). Between 2014 and 2015, across all education groups (except for those with less than a high school degree), wage growth was at least as strong among men as women, sometimes to a striking degree. Among the college educated, men’s wages grew more than twice as fast as women’s. While there has been a slow narrowing of gender wage gaps for those with less than a college degree between 2000 and 2015, gender wage gaps continued to grow among those with an advanced degree.

Figure G

Figure H

Conclusion

Overall, 2015 saw overall real wage gains driven by a dip in inflation. It also saw a pronounced increase in wage inequality. Larger gains in real wages were clearly concentrated at the top of wage distributions, with the exception of 10th percentile gains associated with state-level minimum-wage increases. Growth was faster for male workers and white workers, particularly at the top of the wage distribution, which continued to exacerbate racial, ethnic, and gender wage gaps. Wages for those with additional schooling remain higher than those with less, though the college premium has remained fairly stable over the last few years.

Real wage growth in 2015 was driven by falling commodity prices and not by accelerating nominal wage growth. This is important, as the Federal Reserve regularly analyzes wage trends in search of inflationary pressures. If the Federal Reserve makes decisions based on the data, it is clear that the United States is not in a period of inflation-spurring wage growth. And, for the vast majority of workers to experience stronger and more durable wage growth, the Federal Reserve needs to let the economy continue to recover. In lieu of stronger bargaining power elsewhere, full employment is one of the remaining levers that will allow most workers to see wage increases, as employers will have to pay more to attract and retain the workers they need. That lever is most important for workers at the bottom of the wage distribution (Gould, Davis, and Kimball 2015). To see strong broad-based wage growth outside of the tightest of labor markets, policymakers could strengthen labor standards, such as by raising the minimum wage, expanding eligibility for overtime pay, and protecting and strengthening workers’ right to bargain collectively for higher wages and benefits. For more policies that will raise wages, see EPI’s Agenda to Raise America’s Pay.

Acknowledgments

The author thanks EPI research assistant Teresa Kroeger and EPI data programmer Jin Dai for their valuable contributions to this study.

Endnote

1. Minimum wage increases that occurred in New York and West Virginia on December 31, 2014, are counted as 2015 minimum-wage increases.

References

Bivens, Josh. 2016. “Should We Care About Slow Nominal Wage Growth When Price Inflation Is Slow? YES.” Working Economics (Economic Policy Institute blog), February 5.

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, Briefing Paper No. 378.

Bureau of Labor Statistics (U.S. Department of Labor) Current Employment Statistics program. Various years. Employment, Hours and Earnings—National [database].

Bureau of Labor Statistics (U.S. Department of Labor) National Compensation Survey. Various years. Employment Cost Index [database].

Current Population Survey Outgoing Rotation Group microdata. Various years. Survey con­ducted by the Bureau of the Census for the Bureau of Labor Statistics [machine-readable microdata file]. Washington, D.C.: U.S. Census Bureau.

Current Population Survey public data series. Various years. Aggregate data from basic monthly CPS microdata are available from the Bureau of Labor Statistics through three pri­mary channels: as Historical ‘A’ Tables released with the BLS Employment Situation Summary, through the Labor Force Statistics Including the National Unemployment Rate database, and through series re­ports.

Economic Policy Institute (EPI). 2016a. “Nominal Wage Tracker.” Economic Policy Institute website, accessed February 5.

Economic Policy Institute (EPI). 2016b. “The Agenda to Raise America’s Pay.” Economic Policy Institute website, accessed February 24.

Economic Policy Institute (EPI). 2016c. “Minimum Wage Tracker.” Economic Policy Institute website, accessed March 1.

Gould, Elise, and Alyssa Davis. 2015. Closing the Pay Gap and Beyond: A Comprehensive Strategy for Improving Economic Security for Women and Families. Economic Policy Institute, Briefing Paper No. 412.

Gould, Elise, Alyssa Davis, and Will Kimball. 2015. Broad-Based Wage Growth is a Key Tool in the Fight Against Poverty. Economic Policy Institute, Briefing Paper No. 339.

Mishel, Lawrence, and Will Kimball. 2015. “Wages for Top Earners Soared in 2014: Fly Top 0.1 Percent, Fly.” Working Economics (Economic Policy Institute blog), November 3.

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–Center for Economic and Policy Research Working Paper.

Summary

The Trans-Pacific Partnership (TPP) agreement between the United States and 11 other Pacific Rim countries lacks an absolutely key component to keep it from doing potential damage to the U.S. economy. The missing piece of this trade and investment deal is a set of restrictions and/or enforceable penalties against member countries that engage in currency manipulation. Currency manipulation is one of the key driving forces behind the high and rapidly rising U.S. trade deficit with the 11 other members of the TPP. In 2015, the U.S. deficit with TPP countries translated into 2 million U.S. jobs lost, more than half (1.1 million) of which were in manufacturing. Without such provisions against currency manipulation, the TPP could well follow other trade agreements and leave even greater U.S. trade deficits in its wake.

Currency manipulation occurs when a country artificially depresses the value of its currency. Currency manipulation acts like a subsidy to the exports of the manipulating country, and a tax on U.S. exports to every country where U.S. exports compete with the currency manipulator’s exports. In this way, currency manipulation increases U.S. imports, suppresses U.S. exports, and inflates U.S. trade deficits. As past EPI research has shown, currency-manipulation-fueled trade deficits have reduced U.S. gross domestic product (GDP), eliminated millions of U.S. jobs, driven down U.S. wages, and propelled the outsourcing of U.S. jobs to currency manipulators.

Many members of the proposed TPP, including Malaysia, Singapore, and Japan, are known currency manipulators. Others, namely Vietnam, appear to be following the lead of currency manipulators by, for example, acquiring excess foreign exchange reserves to depress the value of their currency. Currency manipulation explains a substantial share of the large, persistent U.S. trade deficit with the 11 other TPP countries that has not only cost millions of U.S. jobs but also increased income inequality and put downward pressure on American wages. We can’t afford a trade agreement that not only allows but would intensify these harmful trends:

• The $177.9 billion U.S. goods trade deficit with the 11 other TPP countries reduced U.S. GDP by $284.6 billion (1.6 percent) and eliminated 2 million jobs in 2015.
• The 2 million jobs lost due to the U.S. goods trade deficit with TPP member countries in 2015 included 418,900 direct jobs in commodity and manufacturing industries that competed with unfairly traded goods from TPP member countries.
• The currency-manipulation-fueled trade deficit with TPP countries in 2015 was also responsible for the loss of 847,200 indirect jobs in supplier industries, and an additional 759,700 “respending” jobs. These lost respending jobs are jobs that—in a U.S. economy still suffering from low demand—would have been supported by the wages of workers who would have had jobs were trade with the TPP member countries balanced.
• The U.S. trade deficit with TPP member countries in 2015 cost 1,057,200 manufacturing jobs (52.2 percent of the jobs lost due to the U.S. trade deficit with TPP member countries). Within manufacturing, by far the largest losses occurred in motor vehicles and parts, which lost 738,300 jobs (36.4 percent of total jobs lost). Other manufacturing industries with large losses include apparel (181,900 jobs lost or displaced, equal to 9 percent of total jobs lost) and computer and electronic parts (163,900 jobs, or 8.1 percent).
• The U.S. trade deficit with TPP member countries was also responsible for significant job losses outside of manufacturing in 2015. Industries that lost jobs include health care and social assistance (204,200 jobs, 10.1 percent); retail trade (142,800 jobs, 7 percent); accommodation and food services (101,800 jobs, 5 percent); finance and insurance (42,700 jobs, 2.1 percent); agricultural industries (41,600 jobs, 2.1 percent), and education services (37,300 jobs, 1.8 percent).

• Each of the 50 states and the District of Columbia lost jobs due to the U.S. trade deficit with TPP member countries in 2015. Net job losses were greatest in California, which lost 227,500 jobs (constituting 1.38 percent of total state employment). Michigan experienced the greatest jobs lost as a share of state employment (5.12 percent).
• In the 10 hardest-hit states (jobs lost as a share of all state jobs), the share of jobs lost due to the U.S. trade deficit with the TPP countries in 2015 ranged from 1.83 percent to 5.12 percent of total state employment.
• Seven of the 10 states with the highest job losses (as a share of total employment) are in the Midwest or Southeast, in states where manufacturing (especially of motor vehicles and parts) predominates: Michigan (214,600 jobs lost, equal to 5.12 percent), Indiana (103,800 jobs, 3.54 percent), Kentucky (53,700 jobs, 2.92 percent), Alabama (46,000 jobs, 2.32 percent), Tennessee (61,000 jobs, 2.19 percent), Ohio (112,500 jobs, 2.16 percent), and Mississippi (22,000 jobs, 1.86 percent). Other hard-hit states in the top 10 were Oklahoma (35,300 jobs, 2.10 percent), Wyoming (6,800 jobs, 2.34 percent), and Alaska (6,300 jobs, 1.83 percent), all of which have been hard hit by the collapse of the oil industry and related sectors.
• The U.S. trade deficit with TPP member countries in 2015 produced net job losses in all but two U.S. congressional districts. The 11th Congressional District in Michigan was the hardest-hit district in the country, ranked in terms of jobs eliminated as a share of total district employment, losing 26,200 jobs (7.66 percent of total employment). In the 20 congressional districts with the largest shares of jobs lost, net losses ranged from 11,400 to 26,200 jobs, and jobs lost as a share of overall employment ranged from 3.89 percent to 7.66 percent. Michigan had 10 districts in the top 20 job-losing districts, followed by Indiana (five districts); California (two districts); and Ohio, Alabama, and Tennessee (one district each).

These stark figures highlight how much damage the U.S. economy and American workers have already suffered from growing trade deficits with TPP member countries.

And we have seen this picture before—it’s similar to the economic impact that followed the North American Free Trade Agreement (NAFTA), but this time the stakes are higher and the costs more severe (Scott 2013). Prior to NAFTA, the United States sustained balanced trade with Mexico (Scott 2011). The U.S. trade deficit with Mexico took off only after NAFTA was adopted, further demonstrating the degree to which these unfair trade and investment agreements negatively affect the U.S. economy. With the TPP countries, the United States is already starting behind with a trade deficit of $177.9 billion. As a result, the TPP trade and investment deal is likely to be significantly more costly to the U.S. economy.

In this context the United States should insist that currency manipulation be directly addressed in the core of the TPP agreement. Member governments of the TPP should also agree to rebalance trade and currency markets, including by divesting excess foreign assets in their portfolios, before any trade and investment agreement takes effect. They should also forswear the use of currency manipulation in the future, and submit to strong, binding currency disciplines in the event these commitments are violated.

Background: Currency manipulation, trade, wages, and job loss

A considerable body of trade policy research has established connections between currency manipulation, trade deficits, job losses, and wages. These connections are heightened in an era of incomplete recovery from the Great Recession. This section provides a broad overview of the connections and introduces proposed new approaches for intervening when currency management unfairly threatens U.S. jobs and wages.

The effect of exchange rates on imports and exports

Exchange rates measure the value of a country’s currency relative to other currencies (Nelson 2013). The nominal exchange rate is simply the rate at which one currency can be exchanged for another. Exchange rates are used to calculate the value of foreign goods, services, and assets in terms of U.S. dollars. Thus, consumers and businesses in the United States use exchange rates to compute the cost of Japanese and Korean cars in terms of U.S. dollars. In the same way, consumers and businesses in Japan and South Korea use exchange rates to calculate the cost of U.S. cars in Japanese yen or Korean dollars.

Exchange rates are determined by the relative supply and demand for currencies in foreign exchange (FX) markets. The current constellation of trade imbalances is primarily the result of governments that use intentional policies, especially official purchases of foreign assets (public financial flows), to influence exchange rates (Gagnon 2013). This is the basic tool of currency manipulators. They purchase foreign assets such as U.S. treasuries to increase demand for the U.S. dollar, which increases the value of the dollar relative to their own currency.

The price of all of a country’s exports and imports are strongly influenced by the exchange rate; it is one of the most fundamental prices in the economy. Therefore, changes in the exchange rate can have a large impact on the level of imports and exports, and on the trade balance. When a country’s exchange rate declines, relative to other currencies (a depreciation or devaluation), its exports become cheaper in foreign markets, and imports from other countries become more expensive. Over time, devaluation will increase the level of exports and reduce the level of imports. (More on how currency manipulation affects employment levels can be found in Scott 2014c).

How trade deficits affect jobs

In turn, the levels of exports and imports have an effect on employment. Each $1 billion in U.S. exports supports some American jobs. However, each $1 billion in U.S. imports displaces the American workers who would have been employed making these products in the United States. The net employment effect of trade depends on the changes in the trade balance. An improving trade balance will, all else equal, support job creation, while growing trade deficits will result in growing net U.S. job displacement.

How trade deficits affect wages

For example EPI research has shown that growing U.S. trade deficits with China pushed American workers out of good jobs with excellent wages, primarily in manufacturing industries, into lower-paying jobs in nontraded industries, or into unemployment. Growing trade deficits with China between 2001 and 2011 resulted in the net loss of at least $13,505 per displaced worker in 2011 alone. For all displaced workers, using education group averages, net wage losses totaled $37 billion (Scott 2013).

Direct trade, job, and wage losses are just the tip of the iceberg when it comes to the cost of trade deficits, and globalization more broadly, for American workers. Using standard models to benchmark the cost of globalization for American workers without a college degree, Bivens (2013) estimated that in 2011, trade with low-wage countries lowered wages by 5.5 percent—roughly $1,800 for a full-time, full-year worker without a college degree. These losses were experienced by all American workers without a college degree, who make up about two-thirds of the labor force or roughly 100 million U.S. workers.1

Adding the TPP to a low-demand economy would aggravate chronic trade deficits

The United States has run chronic trade deficits for well over a decade. Since 2002, these deficits have been overwhelmingly driven by the conscious policy choices made by several of our major trading partners to manage the value of their currency for competitive advantage in U.S. and global markets. (Gagnon 2013; Bayoumi, Gagnon, and Saborowski 2014; Bergsten and Gagnon 2012; Krugman 2009; Scott 2014c). They buy dollar-denominated assets to boost the value of the dollar and depress the value of their own currencies.

More than 20 countries, led by China, have, together, been spending about $1 trillion per year buying foreign assets in order to artificially suppress the value of their currencies (Scott 2014c). Several of this group—including Malaysia, Singapore, and Japan—are currently members of the TPP and several others—including South Korea, Taiwan, and China—have expressed interest in joining. In addition, Vietnam, which is part of the proposed TPP, has been accumulating foreign-exchange reserves over the past decade. Vietnam has seen its current account surplus, the broadest measure of its trade surplus with the world, rise to an estimated 4.9 percent of GDP in 2014 (IMF 2015 and 2016); in short it is behaving like the other currency manipulators in the TPP.

As Bivens (2016) notes, the threat posed by allowing currency manipulation to go unchecked is heightened in the current context of a U.S. economy not fully recovered from the Great Recession. Despite efforts by the Federal Reserve to bring the economy back to full employment, the U.S. economy has been stuck well below potential for more than eight years. Worse yet, there is widespread evidence that the shortfall in demand that has delayed a full recovery from the Great Recession could last for years to come (Bivens 2016 citing Krugman 2013; Summers 2014).

Economic history shows that such prolonged downturns are quite possible in advanced economies: Japan has been stuck below potential output for decades, and Western Europe is experiencing a double-dip recession because it failed to adequately boost aggregate demand. In the United States, fiscal policy has been notably contractionary since 2011, and the Fed has just raised short-term interest rates for the first time in more than a decade. The Fed’s mistaken rate increase in the face of chronic demand shortfalls means that we are now going in the wrong direction on both fiscal and monetary policy. Thus, a prolonged period of policy-induced, chronic demand shortfall or “secular stagnation” now seems likely in the United States and much of the developed world. For these reasons, more sensible exchange rate policies are needed now more than ever. (Bivens 2016)

Given that the economy is not at full employment and that there is no automatic mechanism that can return it there quickly due to our fiscal and monetary choices, trade flows can have a powerful influence on aggregate demand. Thus, ending the currency manipulation that has thrown U.S. trade flows out of whack is a crucially important goal for macroeconomic stabilization in coming years.

Responses to currency manipulation are gaining traction

Several policy alternatives for ending currency manipulation have already been proposed in Congress, including the Ryan-Murphy Currency Reform and Fair Trade Act (H.R. 2378), which would “clearly define currency manipulation as an illegal subsidy and authorize the Commerce Department to address currency manipulation in countervailing duty (CVD) complaints” (Scott 2014c, 16). In 2010, the House of Representatives approved the Ryan-Murphy act, but the Senate failed to pass a complementary measure (S.1027) in the 111th Congress (OpenCongress.org 2009). In 2011, the Senate was successful in passing the Currency Exchange Rate Oversight Reform Act (S.1619). Together, these bills would address currency manipulation by imposing tariffs on countries with undervalued currency.

Similar legislation was introduced in both houses of Congress in 2013, including the Currency Reform and Fair Trade Act (H.R. 1276) in the House and the Currency Exchange Rate Oversight Reform Act (S. 1114) in the Senate. Both bills have gained considerable bipartisan support in both houses of Congress and would produce the economic and political pressure needed to hold currency manipulators accountable.

In addition to legislative action, taxing or offsetting the acquisition of foreign assets and foreign exchange by currency manipulators is an effective policy tool for stopping currency manipulation. In the case of China, the world’s biggest currency manipulator and possible future member of the TPP, Gagnon and Hufbauer (2011) suggest that “the U.S. government should tax the income (the interest payments) earned on Chinese holding of U.S. financial assets.” This form of taxation is especially potent considering that China was holding $3.4 trillion in foreign exchange assets as of the end of December 2015 (Bloomberg 2016), about two-thirds of which are made up of U.S. Treasury bonds and other U.S. government assets.

Finally, trade agreements such as the TPP have the capacity to address currency manipulation by setting important precedents for international trade and financial regulations. Bipartisan majorities in both houses of Congress recognized the opportunity to make progress on currency manipulation through the Trans-Pacific Partnership. In June 2013, more than half of the U.S. House of Representatives signed a letter to President Obama urging that the TPP agreement include “currency disciplines” that would “bolster our ongoing efforts to respond to these trade-distorting policies” (Congress of the United States 2013). In September 2013, 60 senators signed a similar letter (United States Senate 2013) calling for “strong and enforceable foreign currency manipulation disciplines” in the “TPP and all future trade agreements.” Despite these clearly expressed desires on the part of majorities in both houses of Congress, enforceable currency disciplines where not included in the core of the proposed TPP agreement. The only progress made on addressing currency manipulation was a side pact among finance officials from the 12 countries that included promises to avoid “unfair currency practices and refrain from competitive devaluation,” and to provide a range of data on foreign-exchange holdings. But that agreement will not be subject to the TPP’s enforcement mechanisms. By not pushing to include penalties against currency manipulators in the core of the agreement, the United States has missed an opportunity to establish fair trade standards, protect American workers, and address the high and rapidly growing trade deficit.

In addition to including high standards designed to prohibit currency manipulation by TPP member countries in the future, more should have been done to eliminate existing currency manipulation as a precondition for membership, as noted by Scott (2015). Members of the TPP should have also agreed to rebalance trade and currency markets, including through divestiture of excess foreign assets in government portfolios, before any trade and investment agreement takes effect. They should also have forsworn the use of currency manipulation in the future, while also agreeing to submit to strong, binding currency disciplines in the event that these commitments are violated.

To quantify that missed opportunity, this report adds to the research on the costs of currency manipulation. Bergsten and Gagnon (2012) have estimated that currency manipulation by more than 20 countries had increased global trade (current account) surpluses of intervening countries by between $400 billion and $800 billion per year. They have also estimated that the “largest loser is the United States, whose trade and current account deficits have been $200 to $500 billion per year larger as a result” (Bergsten and Gagnon 2012, 2). Building on this research, Scott (2014c) found that eliminating currency manipulation could create between 1.0 million and 5.8 million U.S. jobs.

What role could rewriting the terms of the TPP to end currency manipulation by TPP members play? Consider that Bergsten and Gagnon’s list of currency manipulators includes several important members of the TPP (Japan, Malaysia, and Singapore), and several countries that have expressed interest in joining the agreement at some point (China, South Korea, and Taiwan). In addition, although TPP member Vietnam is a low-income country (and was thus excluded from Bergsten and Gagnon’s list), it is following the lead of other currency manipulators and acquiring excess foreign exchange reserves and achieving a large trade surplus.2 While some have argued that China and other countries are not presently manipulating their currencies, nothing in the agreement would prevent these countries from engaging in massive interventions again in the future, thereby nullifying any potential benefits of tariff and nontariff trade barrier reductions and other provisions included in the TPP.3

Also, there are significant risks that currency manipulation by China and other TPP neighbors would increase pressure on many of the TPP countries to either initiate or increase the degree to which they engage in currency manipulation, and thereby nullify the benefits of the TPP to the United States. Artificial reductions in the values of the currencies of our TPP trading partners would increase U.S. trade deficits and job losses, and reduce GDP growth.

Estimating the impact of currency manipulation by TPP member countries on the United States

Currency manipulation is the most important cause of the large and growing U.S. goods trade deficit with the group of countries in the Trans-Pacific Partnership. Coupled with the fact that the United States is the largest and most reliable trading partner for many of the TPP countries, this is a recipe for U.S. pain at others’ gain. But for the subsidies provided by currency manipulation, Japanese automakers, for example, would have found it difficult or impossible to achieve their dominance in wide segments of the U.S. market. And currency manipulation has made it difficult or impossible for U.S. firms to penetrate the markets of currency manipulators for many products, due to the effective tax imposed on U.S. products by currency manipulation.

As shown in Table 1, the U.S. goods trade deficit with TPP member countries reached $177.9 billion in 2015. Using a simple macroeconomic model developed by Bivens (2014), we estimate the effects of this trade deficit on U.S. GDP and employment, including respending effects. (The approach is also based, in part, on the models developed in Scott 2014b.) The macroeconomic model estimates the amount of labor (i.e., number of jobs) required to produce a given volume of exports and the labor displaced when a given volume of imports is substituted for domestic output. Within that model, we use an input-output (IO) model to determine the distribution of jobs supported by exports and the jobs eliminated by imports in the U.S. economy. (See the appendix methodology section below for further details on the model structure and data sources used in this study). By providing estimates of the direct and indirect labor requirements of producing output in a given domestic industry, the model tells us each industry’s share of the overall jobs lost.4 The IO employment model is based on the Bureau of Labor Statistics’ (BLS) employment requirements matrix (ERM), which includes 195 U.S. industries, 77 of which are in the manufacturing sector (see the appendix for details on model structure and data sources). This paper assumes that currency manipulation is the primary cause of the U.S. goods trade deficit with Japan, Malaysia, Singapore, and Vietnam; deficits with these four countries make up the majority of the U.S. trade deficit with the TPP countries overall.

Table 1

Jobs eliminated by the U.S. goods trade deficit with TPP member countries directly decrease total employment in trade-related industries, especially those in manufacturing. The IO model also estimates the number of “indirect” jobs supported or eliminated in supplier industries, including those in manufacturing, and in related service sectors. Finally, wages that would have been earned by the jobs people would have held had trade with the TPP member countries been balanced would have supported additional rounds of “respending,” which would have a multiplier effect on output (GDP) and employment.

A breakdown of the jobs eliminated by the U.S. trade deficit with the TPP countries

Using data on U.S. imports from and exports to the 11 other TPP member countries in 2015, coupled with the models developed in this paper, we estimate the total impact of TPP trade on U.S. GDP and the total number of jobs lost. The $177.9 billion U.S. goods trade deficit with the 11 other TPP countries reduced U.S. GDP by $284.6 billion (1.6 percent) in 2015, as shown in Table 1. This analysis includes both the direct effect of the trade deficit on U.S. GDP (-$177.9 billion), and the multiplier or respending effect ($106.7 billion or 60 percent, not shown in Table 1).

The U.S. trade deficit with the 11 other TPP countries eliminated 2 million jobs, as shown in Table 2, which reports the number of direct, indirect, and respending jobs lost (aggregated over all industries). The trade deficit between the United States and the 11 other TPP member countries in 2015 directly eliminated 418,900 jobs. In addition to the direct jobs lost, the U.S. trade deficit with the TPP country group eliminated an additional 847,200 indirect jobs in supplier industries, including jobs in manufacturing, commodity, and service industries. Finally, wages lost because of direct and indirect job cuts from the trade deficits with the TPP member countries would have supported an additional 759,700 respending jobs. The direct, indirect, and respending jobs displaced by the U.S. trade deficit with TPP member countries totals 2,025,800 jobs lost.

Table 2

Job losses and gains by industry

U.S. imports from and exports to the 11 other TPP member countries in 2015 were used to estimate the distribution of net jobs (direct, indirect, and respending) eliminated by the U.S. trade deficit with the TPP member countries by industry for the 45 unique industries (plus eight aggregate sectors) in the U.S. Census Bureau sector plan (U.S. Census Bureau 2009). Our analysis compares jobs lost or gained with 2011 employment data as a baseline to estimate jobs lost or gained as a share of industry employment (U.S. Census Bureau 2013).

Table 3 provides a snapshot of the U.S. goods trade balance with the TPP countries in 2015. The United States had trade surpluses with the TPP countries in some industries and deficits in most others. Most of the surpluses occurred in the category of manufactured products referred to as “industrial supplies,” in which the United States had an overall surplus of $46.7 billion in 2015. The largest surpluses occurred in chemicals ($25.3 billion), petroleum and coal products ($19.1 billion, predominantly refined petroleum products), and plastics and rubber products ($5.9 billion).5 On the other hand, the United States was a net importer of crude oil and gas (-$58.0 billion) from the TPP in 2015. Thus, the United States has become a specialist in the production of basic chemical products and refined petroleum products that are used in other countries to make final products (for example, toys and tires) that are then re-exported back to the United States.

Completing the picture of manufacturing trade with the TPP, the United States has a small trade deficit in nondurable goods of -$16.2 billion, which includes trade surpluses in textiles ($3.3 billion) and food products ($1.8 billion), and a sizeable deficit in apparel, -$13.3 billion. The surplus in textiles reflects, in part, the NAFTA rules of origin, which favor fabric that originates within North America in NAFTA apparel trade. Rules of origin have been substantially weakened in many sectors in the TPP, and it is unclear if the United States will retain its net trade advantage in textile products if the TPP is approved and implemented.

By far the vast majority of the U.S. trade deficit with the TPP countries is in the durable goods industries (-$151.3 billion, 85.0 percent of the total TPP deficit). This deficit, in turn, is dominated by the trade deficit in motor vehicles and parts (-$118.7 billion), computer and electronic parts (-$27.8 billion), and primary metals (-$13.0 billion). Large trade deficits in these sectors explain a large share of the jobs lost due to trade with the TPP countries, as shown in Table 4. It is important to note that durable goods industries such as motor vehicles, computer and electronic parts (including communications, audio, and video equipment), and primary metals industries (including basic steel and steel products) provide large numbers of good jobs with high wages and excellent benefits, especially for workers without a college education. These are the sectors that have been hardest hit by the TPP trade deficit, as shown below.

Overall, the U.S. trade deficit with the 11 other TPP members eliminated 1,057,200 jobs in manufacturing (52.2 percent of jobs lost across all industries), by far the largest number of jobs lost in any major industry, as shown in Table 4. Within manufacturing, the largest losses occurred in motor vehicles and parts, which lost 738,300 jobs (36.4 percent of total jobs lost). Other manufacturing industries with large losses include apparel (181,900 jobs, 9 percent) and computer and electronic parts (163,900 jobs, 8.1 percent). Trade with TPP member countries did contribute to employment in a few manufacturing industries including chemicals (105,400 jobs created); machinery (66,900 jobs); fabricated metal products (55,700 jobs); plastics and rubber products (40,200 jobs); printed matter and related products (21,800 jobs); and petroleum and coal products (20,900 jobs).

In the case of petroleum and coal products, chemicals, plastics, and rubber, while high-wage jobs were created in these industries, the products derived from petroleum and natural gas are also associated with the generation of large amounts of toxic byproducts which have resulted in increased air and water pollution that is most concentrated at domestic production sites. Over the last 10 years, the United States has, in effect, imported pollution and exported chemical products for the production of manufactured goods in other countries. These developments are a byproduct of the rapid development of oil and gas fracking in the United States, which has dramatically increased the supply and reduced the prices of natural gas and related petroleum byproducts.

The U.S.–TPP trade deficit was also responsible for significant job losses outside of manufacturing, in agricultural industries (41,600 jobs); mining (182,800 jobs); utilities (8,400 jobs); wholesale trade (26,700 jobs); retail trade (142,800 jobs); transportation and warehousing (17,900 jobs); information (19,000 jobs); finance and insurance (42,700 jobs); real estate and rental and leasing (16,500 jobs); professional, scientific, and technical services (10,700 jobs); administrative and support and waste management and remediation services (6,900 jobs); education services (37,300 jobs); health care and social services (204,200 jobs); arts, entertainment, and recreation (23,000 jobs); accommodation and food services (101,800 jobs); other services (except public administration) (70,700 jobs); and public administration (15,700 jobs). These jobs losses reflect the combined effects of both indirect jobs loss and respending effects, which reduced the demand for services.

Job losses and gains by state and congressional district

Estimates of job losses by industry form the foundation for the estimates of job losses and gains by state and congressional district. Estimates of employment by state and congressional district for each of the 45 unique industries in the model were obtained from the U.S. Census Bureau (2013). These were used to estimate employment shares by state and congressional district for each industry. These shares were used to estimate total jobs lost or gained per district, with 2011 employment used as the baseline for estimating jobs lost as a share of total state or district employment. Thus, states and congressional districts that have high shares of employment in industries with a large exposure to trade with the TPP member countries (such as motor vehicles and equipment, apparel, or computer and electric parts) were the biggest losers from the trade deficit between the United States and TPP member countries in 2015.

The U.S. trade deficit with the 11 other TPP member countries in 2015 produced net job losses in all 50 states and the District of Columbia. Jobs lost by state, ranked by jobs lost as a share of total state employment, are reported in Table 5a. (Table 5b ranks the states by net jobs displaced and Table 5c lists them alphabetically). Michigan lost the most jobs as a share of total state employment, with 214,600 jobs lost (5.12 percent of total state employment in 2011). Seven of the 10 states with the highest job loss shares are in the Midwest or Southeast census regions, all states where manufacturing predominates. After Michigan they include Indiana (103,800 jobs or 3.54 percent ), Kentucky (53,700 jobs, or 2.92 percent), Alabama (46,000 jobs, 2.32 percent), Tennessee (61,000 job, 2.19 percent), Ohio (112,500 jobs, 2.16 percent), and Mississippi (22,000 jobs, 1.86 percent). Rounding out the top 10 states losing the largest shares of jobs were Wyoming (6,800 jobs, 2.34 percent), Oklahoma (35,300 jobs, 2.1 percent), and Alaska (6,300 jobs, 1.83 percent). The distribution of job losses in the 50 states and the District of Columbia is shown in the map in Figure A. In the online version of this report, the map is clickable, and contains additional data on job losses due to the U.S. trade deficit with the 11 other TPP member countries in 2015.

Figure A

This study also estimates trade-related employment changes by congressional district for the 114th Congress (elected in 2014), using congressional district boundaries from the 2010 Census. The distribution of job losses in the 435 congressional districts and in the District of Columbia is shown in the map in Figure B. In the online version of this report, the map is clickable, and contains additional data on job losses due to the U.S. trade deficit with TPP countries.

Figure B

Our analysis compares jobs lost with 2011 employment data as a baseline to estimate job losses as a share of district employment. The data show that the U.S. trade deficit with the TPP member countries resulted in net job losses in all but two U.S. congressional districts, and displaced 26,200 jobs in a single U.S. congressional district (Michigan’s 11th Congressional District, located in Detroit’s northwest suburbs in parts of Wayne and Oakland counties). The 20 congressional districts with the largest shares of jobs lost are shown in Table 6. Each of the top 20 districts lost between 11,400 and 26,200 jobs. Job losses as a share of district employment among the top 20 U.S. congressional districts ranged from 3.89 percent to 7.66 percent (for the 11th Congressional District in Michigan). Of the states with top 20 job-losing districts, the hardest-hit state was Michigan (with 10 districts in the top 20, followed by Indiana (five districts); California (two districts); and Ohio, Alabama, and Tennessee (one district each). Complete lists of jobs lost or gained by congressional district for all 435 congressional districts and for the District of Columbia are included in Table 7. The table provides net jobs affected and jobs affected as a share of total district employment. The only two congressional districts that experienced net job gains as a result of trade with the 11 other TPP member countries are the 14th Congressional District in Texas (100 jobs gained) and the 4th Congressional District in Kansas (200 jobs gained). Table 8 displays this information alphabetically by congressional district.

Other problems with the TPP

Many researchers have raised concerns over the negative impacts of the Trans-Pacific Partnership. This paper does not include an exhaustive review but cites as an example Capaldo, Izurieta, and Sundaram (2016), who noted that studies claiming that the TPP would have a positive impact on the U.S. and global economy are based on unrealistic assumptions, including no change in the U.S. trade balance with the TPP countries and full employment.

For example, Capaldo, Izurieta, and Sundaram disprove the claim by Petri and Plummer (2016) that the TPP would increase real, annual income in the United States by $131 billion per year, or 0.5 percent of GDP. In fact, after incorporating more realistic assumptions into their model, Capaldo, Izurieta, and Sundaram estimate that the TPP would reduce economic growth in the United States by 0.54 percent after 10 years.6 They also find that though all 12 member countries would incur job losses from the TPP, the United States would be hardest hit, with 448,000 job losses. These job losses are the product of changes in the structure of trade, with the United States producing more capital-intensive goods and fewer labor-intensive goods. This changing structure, as policies to raise profits in some industries (pharmaceuticals, software, and other intellectual property) “push labor shares lower, redistributing income from labor to capital in all countries,” would increase income inequality across the member countries (Capaldo, Izurieta, and Sundaram 2016, 2).

It is important to note that Capaldo, Izurieta, and Sundaram maintain the assumption of stable trade balance among the TPP countries, for consistency with Petri and Plummer’s model. Thus, the estimate of 448,000 jobs lost is a lower bound on likely outcomes. In reality, the TPP would likely result in growing trade deficits and job losses for the United States for the reasons shown here. This would increase the downward pressure on wages in the United States as more good jobs in manufacturing are destroyed.

Conclusion

The failure to include provisions to stop currency manipulation alone casts the Trans-Pacific Partnership as a fatally flawed trade and investment deal. U.S. trade deficits with the 11 other members of the proposed agreement eliminated 2 million U.S. jobs in 2015, and reduced U.S. GDP by nearly $300 billion (1.6 percent). Even if the trade balance with the TPP remains stable, as assumed by the most optimistic proponents of the agreement, growing imports of labor-intensive products would over the next decade eliminate more than 400,000 U.S. jobs, reduce U.S. GDP by an additional one-half percent, and lead to growing income inequality in the United States and other members of the proposed agreement. Under a more likely scenario, the TPP would do that and more—fueling increased outsourcing, growing trade deficits, and even greater downward pressure on the incomes of working Americans.

Currency manipulation is the most important cause of large and growing U.S. trade deficits with the TPP countries. Majorities of both houses of Congress demanded that President Obama include in the core of the TPP “currency disciplines” that could be enforced with trade sanctions. But the president refused to even discuss those issues in the TPP negotiations. Currency manipulation by members of the TPP and by neighboring countries such as China, South Korea, and Taiwan (who all may soon be invited to join the deal) would likely nullify any benefits the U.S. might achieve from the TPP, which will reinforce the negative consequences of the deal for working families and manufacturing communities in the United States and other member countries. Congress should reject this agreement. The president can and should have done better for American workers, communities, and domestic businesses based in the United States.

About the authors

Robert E. Scott is director of trade and manufacturing policy research at the Economic Policy Institute. He joined EPI as an international economist in 1996. Before that, he was an assistant professor with the College of Business and Management of the University of Maryland at College Park. His areas of research include international economics and trade agreements and their impacts on working people in the United States and other countries, the economic impacts of foreign investment, and the macroeconomic effects of trade and capital flows. He has a Ph.D. in economics from the University of California-Berkeley.

Elizabeth Glass is a trade and manufacturing policy research assistant at the Economic Policy Institute. She provides research support on a variety of trade-related issues including currency manipulation, industrial policy, and employment. Prior to joining EPI in 2015, she worked with a number of international development organizations on international education policy and economic development research. She holds an M.A. in international economic development from The New School.

Acknowledgments

The authors thank Josh Bivens for comments, Jessica Schieder and Will Kimball for research assistance, Lora Engdahl for editing, and Chris Frazier and Chris Roof for layout and interactive graphics.

Appendix: Methodology

This analysis uses a simple macroeconomic model developed by Bivens (2014) to estimate the effects of the U.S. goods trade deficit with the TPP countries in 2015 on U.S. GDP and employment, including respending effects (based, in part, on the models developed in Scott 2014b). It then uses an input-output model based on Bureau of Labor Statistics (BLS) data to allocate jobs displaced by the U.S.-TPP trade deficit (derived from the macroeconomic model) to industries, states, and congressional districts. This combined macroeconomic/IO model uses data from 2015 to estimate the impacts of the trade deficit in that year. This appendix identifies the specific data sources and comparisons used.

The macroeconomic model

The effect of the U.S.-TPP trade deficit on GDP and jobs is determined by economic multipliers surveyed by Bivens (2014). As he notes, “the most pressing economic challenge for the U.S. economy remains the depressed labor market” (Bivens 2014, 1). The share of prime-age adults (age 25–54) remains barely above the level at the official end of the recession in 2009, and well below the peaks of the last two business cycles. In this economic environment, changes in spending for domestic goods have large multiplier effects on the economy. Bivens estimates that in the current economic environment, increases in infrastructure spending have a large, macroeconomic multiplier impact on the domestic economy through the wages earned and spent by workers employed by such spending. According to Bivens, that infrastructure spending is associated with a multiplier effect of 1.6 on the domestic economy (Bivens 2014, Table 5 at 21). This paper assumes that changes in trade flows also have a multiplier effect of 1.6, and that reductions in domestic spending caused by the U.S.-TPP trade deficit impact the economy in a way that is symmetric with increases in spending associated with increased infrastructure investment (that is, the multiplier works the same way for both increases and decreases in domestic spending).

The overall number of jobs eliminated by this reduction in output (GDP) is estimated from a simple rule of thumb also developed by Bivens (2014, Table 5 at 21), based on historical relationships between output and employment in which each 1 percent increase in GDP supported would increase total employment by 0.9 percent (approximately 1.3 million jobs in the economy in 2015). Likewise, an identical reduction in GDP would eliminate the same number of jobs in the U.S. economy.

This study examines the impacts of total trade in goods with the 11 other TPP countries (that is, total exports and general imports). The trade deficit (and job loss estimates) would be even larger if we had separated exports produced domestically from foreign exports—which are goods produced in other countries, exported to the United States, and then reexported from the United States, as we have done in earlier studies of trade and employment.7 The use of total exports in this study yields more conservative estimates of trade-related job displacement. New data collection efforts by the Customs Bureau and U.S. Census Bureau may be required to specifically identify the import sources (by country) of U.S. foreign exports. Further research is required on the origin of foreign exports to more accurately assess the impacts of trade on domestic product (and domestic exports only). We examine trade in total exports and general imports in order to develop the most conservative estimates of the U.S. trade deficit and jobs lost due to trade with the TPP countries.8

The trade and jobs model

This section describes the IO model that is used to allocates jobs lost due to trade to individual industries, and the census data which are then used to allocated those losses to states and Congressional districts. The trade and employment analyses by industry in this report are based on a detailed, industry-based study of the relationships between changes in trade flows and employment for each of approximately 195 individual industries of the U.S. economy. For the state and congressional district analysis, these are specially grouped into 45 custom sectors using the North American Industry Classification System (NAICS) with data obtained from the U.S. Census Bureau (2013). Trade data for this analysis were obtained from the U.S. International Trade Commission (USITC 2016).

The number of jobs supported by $1 million of exports or imports for each of 195 different U.S. industries is estimated using a labor requirements model derived from an input-output table developed by the BLS–EP (2014a).9 This model includes both the direct effects of changes in output (for example, the number of jobs supported by $1 million in auto assembly) and the indirect effects on industries that supply goods (for example, goods used in the manufacture of cars). So, in the auto industry for example, the indirect impacts include jobs in auto parts, steel, and rubber, as well as service industries that provide inputs to the motor vehicle manufacturing companies, such as accounting, finance, and computer programming. This model estimates the labor content of trade using empirical estimates of labor content and goods flows between U.S. industries in a given base year (an input-output table for the year 2010 was used in this study) that were developed by the U.S. Department of Commerce and the BLS–EP. It is not a statistical survey of actual jobs gained or lost in individual companies, or the opening or closing of particular production facilities (Bronfenbrenner and Luce 2004 is one of the few studies based on news reports of individual plant closings).

Nominal trade data used in this analysis were converted to constant 2005 dollars using industry-specific deflators (see next section for further details). This was necessary because the labor requirements table was estimated using price levels in that year. Data on real trade flows were converted to constant 2005 dollars using industry-specific price deflators from the BLS–EP (2014b). These price deflators were updated using Bureau of Labor Statistics producer price indexes (industry and commodity data; BLS 2016b). Use of constant 2005 dollars was required for consistency with the other BLS models used in this study.

The IO model is used to estimate the distribution of jobs displaced by trade, and by the loss of wages and respending, as explained below.

Estimation and data sources

Data requirements

Step 1. U.S.-TPP trade data are obtained from the U.S. International Trade Commission DataWeb (USITC 2016) in four-digit, three-digit, and two-digit NAICS formats. Consumption imports and domestic exports are downloaded for each year.

Step 2. To conform to the BLS Employment Requirements tables (BLS–EP 2014a), trade data must be converted into the BLS industry classifications system. For NAICS-based data, there are 195 BLS industries. The data are then mapped from NAICS industries onto their respective BLS sectors. The trade data, which are in current dollars, are deflated into real 2005 dollars using published price deflators from the BLS–EP (2014b) and the Bureau of Labor Statistics (2016b).

Step 3. A 1×195 vector of data for total personal consumer expenditures (PCE) in 2005 dollars for 2010 was extracted from historical input-output data assembled by the BLS-EP (2015). These data were used to estimate total employment supported by PCE expenditures (using the job-equivalents analysis described below). The results were used to estimate the share of respending jobs supported in each of 195 BLS industries.

Step 4. Real domestic employment requirements tables are downloaded from the BLS–EP (2014a). These matrices are input-output industry-by-industry tables that show the employment requirements for $1 million in outputs in 2005 dollars. So, for industry i the aij entry is the employment indirectly supported in industry i by final sales in industry j and where i=j, the employment directly supported.

Analysis

Step 1. Job equivalents. BLS trade data are compiled into matrices. Let [T₂₀₁₅] be the 195×2 matrix made up of a column of imports and a column of exports for 2015. To estimate the vector of jobs displaced by trade, perform the following matrix operations:

[J₂₀₁₅]=[T₂₀₁₅]×[E₂₀₁₀]

[J₂₀₁₅] is a 195×2 matrix of jobs displaced (eliminated) by imports and jobs supported by exports for each of 195 industries in 2015. This matrix is used to create vectors of net jobs displaced by imports from and jobs supported by exports to the TPP countries, as described above. The total number of direct and indirect jobs displaced by trade is estimated using the macroeconomic model described above.

The employment estimates for retail trade, wholesale trade, and advertising were set to zero for this analysis.10 We assume that goods must be sold and advertised whether they are produced in the United States or imported for consumption.

Similarly, for respending (multiplier) analysis, let [PCE₂₀₁₀] be the 195×1 matrix of total U.S. personal consumer expenditures by industry in 2010 (in real 2005 dollars). To estimate the distribution of jobs supported by respending, perform the following matrix operations:

[J_PCE2010]=[PCE₂₀₁₀]×[E₂₀₁₀]

Direct and indirect jobs. In order to estimate the direct jobs, the diagonal vector was extracted from the employment requirements matrix [E₂₀₁₀]. This vector was multiplied by the trade vector to estimate direct trade-related jobs (e.g., [J_DIRECT2015]₎ for both imports and exports. Indirect jobs just equal total jobs less direct (e.g., [J_INDIRECT2015] =[J₂₀₁₅] – [J_DIRECT2015]).

Step 2. Combining macroeconomic and IO jobs analyses. The IO jobs estimates in vectors [J₂₀₁₅] and [J_PCE2015] are converted into share vectors, representing the share of total jobs supported in each of 195 industries by reductions in trade deficits and related respending in the domestic economy. The shares in each vector sum to 1. Share vectors are used to allocate jobs gained by industry. The sum of direct and indirect jobs gained (Table 2) in each scenario is multiplied by the trade jobs share vector derived from [J₂₀₁₅], and the respending (also Table 2) jobs is multiplied by the respending jobs share vector derived from [J_PCE2015]. The results yield estimates of jobs gained or lost by industry in the total economy as a result of the U.S.-TPP trade deficit, which are reported in Table 4.

Step 3. State-by-state analysis. For states, employment-by-industry data are obtained from the Census Bureau’s American Community Survey (U.S. Census Bureau 2013) data for 2011 and are mapped into 45 unique census industries and eight aggregated total and subtotals for a total of 53 sectors.11 We look at jobs displaced in 2015, so from this point, macroeconomic jobs estimates are derived from the vectors [J₂₀₁₅] and [J_PCE2010]. In order to work with 45 sectors, we group the 195 BLS industries into a new matrix, defined as [Jnew₂₀₁₅], a 45×1 matrix of job gains and losses. Define [St₂₀₁₁] as the 45×51 matrix of state employment shares (with the addition of the District of Columbia) of employment in each industry. Calculate:

[Stj₂₀₁₅]=[St₂₀₁₁]_T [Jnew₂₀₁₅]

where [Stj₂₀₁₅] is the 45×51 matrix of job displacement/support by state by industry. To get state total job displacement, we add up the subsectors in each state.

Step 4. Congressional district analysis. Employment by congressional district, by industry, and by state is obtained from the ACS data for 2011, which for the first time use geographic codings that match the boundaries of the 113th Congress (elected in 2012) and the 114th Congress (elected in 2014). In order to calculate job gains or losses in each congressional district, we use each column in [Stj₂₀₁₃], which represent individual state job-gain and loss-by-industry estimates, and define them as [Stj₀₁], [Stj₀₂], [Stj_i]…[Stj₅₁], with i representing the state number and each matrix being 45×1.

Each state has Y congressional districts, so [Cd_i] is defined as the 45xY matrix of congressional district employment shares for each state i. Congressional district shares are calculated thus:

[Cdj₀₁]=[Stj₀₁]_T [Cd₀₁]

[Cdj_i]=[Stj_i]_T [Cd_i]

where [Cdj_i] is defined as the 45xY job gains and losses in state i by congressional district by industry.

To get total job displacement by congressional district, we add up the subsectors in each congressional district in each state.

Endnotes

1. See Scott (2016) for further background on the impacts of trade on U.S. wages.

2. Bergsten and Gagnon (2012) established four criteria for identifying currency manipulators based on the levels of foreign-exchange reserves relative to imports of goods and services, on foreign-exchange reserves growing faster than GDP, on having a current account surplus, and on having GDP in excess of $3,000 per capita. The first three requirements had to demonstrate persistence (e.g., a current account surplus continuously between 2001 and 2011). Vietnam would be excluded under the fourth criterion alone, as its per capita GDP was less than $2,200 in 2015 (IMF 2015). In addition, Vietnam’s trade and foreign-exchange accumulation have only developed in the past few years. But the size of its current account surplus, and its likely role as a low-wage export platform in the TPP, suggest that its currency is, and may continue to be, undervalued due to manipulation.

3. See Scott (2016). China continues to accumulate massive reserves in its SWFs. Based on these data and recent changes in prices, relative productivity growth rates, and trade balances, we believe that the RMB is still substantially undervalued.

4. The Economic Policy Institute and other research entities have examined the job impacts of trade in recent years by netting the job opportunities lost to imports against those gained through exports. This report follows that approach, using standard input-output models and data to estimate the jobs displaced by trade. Many reports by economists in the public and private sectors have used this “all-but-identical” methodology to estimate jobs gained or displaced by trade, including Groshen, Hobijn, and McConnell (2005) of the Federal Reserve Bank of New York, and Bailey and Lawrence (2004) in the Brookings Papers on Economic Activity. The U.S. Department of Commerce recently published estimates of the jobs supported by U.S. exports (Johnson and Raumussen 2013) using input-output and “employment requirements” tables from the Bureau of Labor Statistics Employment Projections program (BLS-EP 2014a), the same source used to develop job displacement estimates in this report.

5. This classification is not used by the Census or in the North American Industrial Classification System (NAICS, see Census: http://www.census.gov/eos/www/naics/). Within the NAICS system, manufacturing consists of the two-digit industries in the ranges of 31, 32, and 33. Those sectors we refer to as industrial supplies are all NAICS industries in the 32 classification. Most of these industries are classified as nondurable goods in other classifications of industrial output. However, they are qualitatively different from other nondurable goods such as textiles and apparel, so we treat them separately here.

6. Capaldo et al. (2016) maintain the assumption of balanced trade, but include a greater number of industries than in Petri and Plummer (2016) and thereby examine changes in the structure of trade, with the U.S. producing more capital-intensive goods and fewer labor-intensive goods, which results in growing unemployment and other impacts discussed here.

7. See, for example, Scott (2014a). Foreign exports have become an especially large proportion of U.S. trade with Mexico after NAFTA, as shown in Scott (2011).

8. The United States had total domestic exports to the TPP countries of $565.8 billion in 2015, and consumption imports of $836.6 billion for a net export trade deficit of $270.8 billion, 52.2 percent larger than the overall trade deficit reported in Table 1. If this more narrowly defined trade deficit had been used, the GDP and jobs lost due to TPP trade would have been proportionately larger. However, this measure would likely include some imports that were reexported to other TPP countries, and thus would have overstated the actual deficit with the TPP. Until we can precisely identify the source of foreign exports (by country and industry or product code) we will be unable to more accurately estimate net domestic trade flows with the TPP or other trade partners.

9. The model includes 195 NAICS industries. The trade data include only goods trade. Goods trade data are available for 85 commodity-based industries, plus software, waste and scrap, used or secondhand merchandise, and goods traded under special classification provisions (e.g., goods imported from and returned to Canada; small, unclassified shipments). Trade in scrap, used, and secondhand goods has no impact on employment in the BLS model. Some special classification provision goods are assigned to miscellaneous manufacturing.

10. The respending analysis does include some impacts on employment in wholesale and retail trade, and in advertising. Thus, the net jobs analysis presented in Table 4 (which includes all direct, indirect, and respending jobs supported or displaced by the trade deficit) does include some net jobs displaced in these industries.

11. The Census Bureau uses its own table of definitions of industries. These are similar to NAICS-based industry definitions, but at a somewhat higher level of aggregation. For this study, we developed a crosswalk from NAICS to Census industries, and used population estimates from the ACS for each cell in this matrix.

References

Bailey, Martin N., and Robert Z. Lawrence. 2004. “What Happened to the Great U.S. Jobs Machine? The Role of Trade and Electronic Offshoring.” Brookings Papers on Economic Activity, vol. 35, no. 2, pp. 211–284.

Bayoumi, Tamim, Joseph Gagnon, and Christian Saborowski (BGS). 2014. Official Flows, Capital Mobility and Global Imbalances. Peterson Institute for International Economics, Working Paper WP 14-8.

Bergsten, C. Fred, and Joseph E. Gagnon. 2012. Currency Manipulation, the US Economy, and the Global Economic Order. Peterson Institute for International Economics, Policy Brief 12-25.

Bivens, Josh. 2013. Using Standard Models to Benchmark the Costs of Globalization for American Workers without a College Degree. Economic Policy Institute, Briefing Paper #354.

Bivens, Josh. 2014. The Short- and Long-Term Impact of Infrastructure Investments on Employment and Economic Activity in the U.S. Economy. Economic Policy Institute, Briefing Paper #374.

Bivens, Josh. 2016. The Trans-Pacific Partnership is Unlikely to be a Good Deal for American Workers. Economic Policy Institute, Briefing Paper #397.

Bloomberg. 2016. “China Monthly Foreign Exchange Reserves.” BloombergBusiness, January 31.

Bronfenbrenner, Kate, and Stephanie Luce. 2004. The Changing Nature of Corporate Global Restructuring: The Impact of Production Shifts on Jobs in the U.S., China, and Around the Globe. Commissioned research paper for the U.S. Trade Deficit Review Commission.

Bureau of Labor Statistics (BLS). 2016a. “Current Employment and Establishment Statistics.”

Bureau of Labor Statistics (BLS). 2016b. “Producer Price Indexes: Industry and Commodity Data” [Excel files].

Bureau of Labor Statistics, Employment Projections program (BLS–EP). 2014a. “Special Purpose Files—Employment Requirements; Chain-Weighted (2005 dollars) Real Domestic Employment Requirements Table for 2010” [Excel sheet, converted to Stata data file].

Bureau of Labor Statistics, Employment Projections program (BLS–EP). 2014b. “Special Purpose Files—Real Final Demand Aggregate” [REAL_FDAGG-2010 DAT File, converted to Excel sheet and PCE VECTOR converted to Stata data file].

Bureau of Labor Statistics, Employment Projections program (BLS–EP). 2015. “Special Purpose Files—Industry Output and Employment – Industry Output.” [CSV File, converted to Excel sheet and Stata data file].

Capaldo, Jeronim, Alex Izurieta, and Jomo Kwame Sundaram. Trading Down: Unemployment, Inequality and Other Risks of the Trans-Pacific Partnership Agreement. No. 16-01. GDAE, Tufts University, 2016.

Congress of the United States. 2013. “June 6 2013 Congress of the United States Letter to President Barack Obama.” June 6.

Gagnon, Joseph, and Gary Hufbauer. 2011. “Taxing China’s Assets: How to Increase U.S. Employment Without Launching a Trade War.” Foreign Affairs.com.

Gagnon, Joseph E. 2013. The Elephant Hiding in the Room: Currency Intervention and Trade Balances. Peterson Institute for International Economics, Working Paper 13-2.

Groshen, Erica L., Bart Hobijn, and Margaret M. McConnell. 2005. “U.S. Jobs Gained and Lost Through Trade: A Net Measure.” Current Issues in Economics and Finance, vol. 11, no. 8, pp. 1–7.

International Monetary Fund (IMF). 2015. “World Economic Outlook Database: October 2015 Edition.”

International Monetary Fund (IMF). 2016. International Financial Statistics: Database and Browser (CD ROM). Washington, D.C.: International Monetary Fund. January.

Johnson, Martin, and Chris Rasmussen. 2013. “Jobs Supported by Exports 2012: An Update.” International Trade Administration, Department of Commerce.

Krugman, Paul. 2009. “Chinese New Year.” The New York Times, December 31.

Krugman, Paul. 2013. “Secular Stagnation, Coalmines and Larry Summers.” The Conscience of a Liberal (New York Times blog), November 16.

Nelson, Rebecca M. 2013. Current Debates over Exchange Rates: Overview and Issues for Congress. Congressional Research Service Report R43242.

OpenCongress.org. 2009. S. 1027 – Currency Reform for Fair Trade Act of 2009.

Petri, Peter A., and Michael G. Plummer. 2016. “The Economic Effects of the Trans-Pacific Partnership: New Estimates.” Peterson Institute for International Economics, Working Paper 16-2.

Scott, Robert E. 2011. Heading South: U.S.-Mexico Trade and Job Displacement after NAFTA. Economic Policy Institute, Briefing Paper #308.

Scott, Robert E. 2013. Trading away the Manufacturing Advantage: China Trade Drives Down U.S. Wages and Benefits and Eliminates Good Jobs for U.S. Workers. Economic Policy Institute, Briefing Paper #367.

Scott, Robert E. 2014a. China Trade, Outsourcing and Jobs: Growing U.S. Trade Deficit with China Cost 3.2 Million Jobs between 2001 and 2013, with Job Losses in Every State.” Economic Policy Institute, Briefing Paper #385.

Scott, Robert E. 2014b. “The Effects of NAFTA on US Trade, Jobs and Investment, 1993–2013.” Review of Keynesian Economics, vol. 2, no. 4, Winter 2014, pp. 429–441.

Scott, Robert E. 2014c. Stop Currency Manipulation and Create Millions of Jobs: With Gains across States and Congressional Districts. Economic Policy Institute, Briefing Paper #372.

Scott, Robert E. 2015. Currency Manipulation and the 896,600 U.S. Jobs Lost Due to the U.S.-Japan Trade Deficit. Economic Policy Institute, Briefing Paper No. 387.

Scott, Robert E. 2016. Trans-Pacific Partnership Agreement: Currency Manipulation, Trade, Wages, and Job Loss. Economic Policy Institute.

Summers, Lawrence. 2014. “US Economic Prospects: Secular Stagnation, Hysteresis, and the Zero Lower Bound.” Business Economics, vol. 49, no. 2, 65–73.

United States Senate. 2013. Letter to Secretary Lew and Ambassador Froman. September 23.

U.S. International Trade Commission (USITC). 2016. “USITC Interactive Tariff and Trade DataWeb” [Excel files].

U.S. Census Bureau. 2009. “2007 Census Industrial Classification.”

U.S. Census Bureau. 2013. “American Community Survey: Special Tabulation Over 45 Industries, Covering 435 Congressional Districts and the District of Columbia (113th Congress Census Boundaries), Plus State and US Totals Based on ACS 2011 1-year file” [spreadsheets received March 6].

Table 3

Table 4

Table 5a

Table 5b

Table 5c

Table 6

Table 7

Table 8

Introduction and key findings

Millions of Americans rely upon public assistance programs to help meet their basic needs. These programs provide a vital lifeline for individuals and families struggling to get by. Indeed, given rising costs of necessities such as child care, housing, and health care, many families’ ability to achieve a modest but adequate standard of living requires resources earned on the job and assistance from government programs.1

However, for many workers in certain sectors, wages are so low that even those who work full time must rely heavily on government assistance to make ends meet. This suggests that low pay by many employers—facilitated by weakened or inadequate labor standards, such as a low minimum wage and outdated overtime regulations—is placing unwarranted demands on public resources. As corporations achieve extraordinarily high profit levels and executive pay reaches new heights, it is appropriate to question whether employers are effectively passing off a portion of their societal responsibilities on to taxpayers.

This report examines the utilization of public assistance among low-wage workers and their families. After a brief review of previous research, it presents data on program participation and transfer income receipt by working individuals’ annual hours of work, hourly wage level, major industry of employment, and state. Then it examines how higher wages among workers at various wage levels affect utilization rates and benefit dollars received. Finally, it discusses policies that would raise wages and the effect these policies would have on public assistance utilization and overall program spending. It concludes that higher hourly wages for low- and middle-wage workers, achievable through a variety of labor-market policies, would unambiguously generate savings in government safety-net and income-support programs—savings that could be used to strengthen and expand anti-poverty programs or make critical public investments to boost productivity and grow the economy.

Key findings include:

• Most recipients of public assistance work or have a family member who works.
  • Among families or individuals receiving public assistance, the majority (66.6 percent) work or are in working families (families in which at least one adult in the household works). This number grows to 71.6 percent when focusing on non-elderly recipient families and individuals (those under age 65).
  • About 69.2 percent of all public assistance benefits received by non-elderly families or individuals go to those who work.
  • Nearly half (46.9 percent) of all working recipients of public assistance work full time (at least 1,990 hours per year).
• Working recipients of public assistance are concentrated at the bottom of the wage scale and in low-paying industries.
  • Roughly 60 percent of all workers in the bottom decile of wage earners (those paid less than $7.42 per hour) receive some form of government-provided assistance, either directly or through a family member. Similarly, over half (52.6 percent) of workers in the second decile of wage earners (those paid between $7.42 and $9.91 per hour) receive public assistance.
  • Workers in the arts, entertainment, recreation, accommodation, food services, and retail trade industries are disproportionately represented among public assistance recipients. Workers in these industries receive even more disproportionate shares of program benefits, underscoring the particularly low wages in these industries.
• Raising wages for low-wage workers (defined as those in the bottom three wage deciles, who earn up to $12.16 per hour) would unambiguously reduce net spending on public assistance, particularly among workers likely to be affected by a federal minimum-wage increase.
  • Among workers in the bottom three wage deciles, every $1 increase in hourly wages reduces the likelihood of receiving means-tested public assistance by 3.1 percentage points. This means that the number of workers receiving public assistance could be reduced by 1 million people with a wage increase of just $1.17 an hour, on average, among the lowest-paid 30 percent of workers. These workers would see higher incomes, even as they no longer received public assistance.
  • For every $1 that wages rise among workers in the bottom three wage deciles, spending on government assistance programs falls by roughly $5.2 billion. This estimate is conservative, as it does not include the value of Medicaid benefits.
  • Raising the federal minimum wage to $12 per hour by 2020 would reduce means-tested public assistance spending by $17 billion annually. These savings could fund a variety of improvements to government anti-poverty tools, such as expanding the Earned Income Tax Credit (EITC) to childless adults, or provide funding for new education initiatives, such as improving access to preschool for children from low- and moderate-income families.

Background

The failure of regular, full-time employment to provide adequate levels of income is one of the many damaging consequences of the long-term stagnation of American workers’ pay. Over the past generation, despite significant increases in labor productivity, inflation-adjusted hourly pay for the vast majority of American workers has either stagnated or declined. This is the result of deliberate policy choices that have reduced workers’ ability to negotiate higher pay, and that have allowed capital owners and corporate managers to extract an increasing share of the income generated by American workers (Bivens et al. 2014). As a result, living standards for the typical American household have been largely unchanged since the late 1970s, with households working increased hours yet taking home little more in inflation-adjusted pay.

Persistent wage stagnation has also hamstrung efforts to reduce poverty, even as key elements of the tax-and-transfer system have expanded in recent decades. Since the early 1990s, policymakers have reformed and expanded some poverty alleviation programs, yet they have increasingly tied these programs to work. Unfortunately, at the same time, policy decisions made on behalf of those with the most income, power, and wealth have limited workers’ ability to attain higher pay (Bivens et al. 2014). As a result of these policy choices—which are discussed in Mishel and Eisenbrey (2015)—expansions to post-tax aid programs have largely only offset the decline in pre-tax earnings (Gould, Davis, and Kimball 2015).

By preventing stronger wage growth, these policies have increased reliance on public assistance programs. Consider the policy decision to neglect the federal minimum wage: Infrequent and inadequate adjustments have left the minimum wage significantly below its inflation-adjusted peak in 1968. At that time, a full-time minimum-wage salary could keep a family of three out of poverty. Today, a full-time worker paid the federal minimum wage is not paid enough to keep a family of two out of poverty. Consequently, these families often depend on public assistance programs to make ends meet.

Review of previous literature

There is relatively little research into how increases in hourly wages among low-wage workers, typically through increases in minimum wages, affect government spending and participation in public assistance programs. While the minimum wage is one of the most studied topics in economics, this research has focused primarily on how minimum-wage policies affect employment; see Kuehn (2014), Schmitt (2013), or Belman and Wolfson (2014) for a summary. In recent years, some researchers have looked more closely at how increases in the minimum wage have affected family incomes and poverty rates. Dube (2013) provides an extensive survey of this literature and conducts his own analysis of Current Population Survey data. Consistent with most past research, he finds that increases in minimum wages significantly reduce poverty rates and increase family incomes, particularly for low-income families. He does not describe specific effects on income from non-wage sources, such as public assistance programs.

More recently, the Congressional Budget Office (CBO) examined the potential effects of a federal minimum-wage increase to $10.10 (CBO 2014). While there is some debate over the CBO’s estimates of employment effects—see Bernstein (2014) or Shierholz and Cooper (2014)—CBO estimated that raising the federal minimum wage to $10.10 would increase family incomes of workers below the federal poverty line by $5 billion and lift incomes of workers between one and three times the poverty line by $12 billion. CBO predicts that such an increase would lift 900,000 people above the federal poverty line.

CBO notes that these increased earnings would result in higher tax revenue and reduced spending on certain means-tested federal assistance programs, although they do not detail the predicted effects upon individual programs. They also predict that the government would face some additional direct costs from increased wages to a small number of government employees, and possibly a small increase in purchasing costs of certain goods and services if producers raised prices in response to the wage hike. CBO predicts that federal expenses would initially go down, but could later increase if the higher minimum wage has a significant negative effect on employment. On net, they conclude that “it is unclear whether the effect for the coming decade as a whole would be a small increase or a small decrease in budget deficits.”2 It is important to note that the CBO’s ambiguity on this point is driven by their atypically high estimates of the probability of significant employment loss stemming from such an increase. If employment loss is insignificant (as most research on a minimum-wage increase of this magnitude indicates), the budget savings would surely dominate.

Other researchers have attempted to quantify the anticipated savings to government transfer programs from changes in federal wage standards. Giannarelli, Morton, and Wheaton (2007) used a microsimulation model of all U.S. tax, transfer, and health programs to estimate the effects of a package of labor and anti-poverty policies, including raising the federal minimum wage from $5.15—its level in 2007—to $7.25. They estimated that such an increase would decrease transfer costs or raise federal revenues by the equivalent of $2.5 billion in 2014 dollars. They also simulated raising the minimum wage to half the average wage of production, nonsupervisory employees—at that time, equal to $8.40—and estimated it would reduce transfer costs or raise revenues by $14.8 billion in 2014 dollars.

Zabin, Dube, and Jacobs (2004) examined utilization of California’s 10 major means-tested public assistance programs among working families. They found that more than half of the state’s spending on public assistance goes to “working families,” defined as families where at least one family member worked at least 45 weeks out of the year. Using a microsimulation approach, they estimated that if the state raised its minimum wage from $6.75—where it stood at that time—to $8 per hour (an 18.5 percent increase), state public assistance payments would fall by $2.7 billion ($3.3 billion in 2014 dollars).

More recently, a number of researchers have examined how low wages and the minimum wage interact with utilization of specific transfer programs or utilization in particular states or industries. West and Reich (2014a) look specifically at the effect of past minimum-wage increases on Supplemental Nutrition Assistance Program (SNAP) enrollments and expenditures. They use a regression framework that exploits state variation in minimum-wage levels over a 12-year period from 1990 to 2012 to measure how changing minimum-wage levels affected SNAP participation. They find that “a 10 percent increase in the minimum wage reduces SNAP enrollment by between 2.4 and 3.2 percent and reduces program expenditures by 1.9 percent.” Based on these findings, they conclude that an increase in the federal minimum wage to $10.10 would reduce SNAP enrollments by up to 3.8 million persons, and decrease program expenditures by nearly $4.6 billion. Because both the proposed minimum-wage level and the SNAP eligibility level are indexed for inflation, these savings would total $46 billion over the 10-year budget window.3

West and Reich (2014b) also consider how increasing the federal minimum wage would affect participation in Medicaid. In particular, they note that the availability of federal funding for states that expanded Medicaid through the Affordable Care Act (ACA) provides a unique opportunity for these states to generate budgetary savings by increasing the minimum wage. Under the traditional framework for Medicaid, states split Medicaid costs evenly with the federal government, but only individuals with very low levels of income could qualify. The ACA encourages states to expand eligibility for Medicaid to individuals with higher levels of income by initially providing 100 percent federal funding for these newly eligible participants, and 90 percent funding in later years. The authors explain that by raising minimum wages, the incomes of many Medicaid beneficiaries who previously qualified under traditional Medicaid will go up, pushing them into the range of expanded eligibility under the ACA. When this happens, states’ responsibility for costs for these participants will shift almost exclusively to the federal government. West and Reich estimate that raising the federal minimum wage to $10.10 would effectively shift $2.5 billion per year from state to federal balance sheets.

Allegretto et al. (2013) looked specifically at receipt of public assistance among workers in the fast food industry. They found high rates of take-up among front-line fast food workers, with more than half of the families of such workers utilizing public assistance, compared with 25 percent of the workforce as a whole. While the authors do not attempt to simulate any effect on assistance payments from raising wages in the industry, they estimate that taxpayers spend $7 billion annually in public assistance programs for families of these low-paid workers.

Finally, Sawhill and Karpilow (2014) discuss how increasing the federal minimum wage to $10.10 could more than offset the cost of expanding the EITC to childless adult workers. The authors calculate that in the interaction between an expanded EITC and higher wages resulting from an increase to the federal minimum wage, the net effect would be increased tax revenues or reduced public assistance outlays of $1 billion. While they describe, in detail, the fiscal and social benefits that a tandem enactment of these policies would have, they do not elaborate on changes in participation rates or program benefit outlays for individual public assistance programs.

Incidence and value of government assistance benefits by annual hours of work, hourly wage decile, industry, and state

This report examines participation in eight federal and state means-tested programs for low-income families: EITC; the refundable portion of the Child Tax Credit (CTC); SNAP; the Low Income Home Energy Assistance Program (LIHEAP); the Supplemental Nutrition Program for Women, Infants, and Children (WIC); the Section 8 Housing Choice Voucher program; Medicaid; and the Temporary Assistance for Needy Families program (TANF) or equivalent state and local cash assistance programs.4

For all of these programs, eligibility is restricted to individuals with low total family incomes, often some percentage of the federal poverty line. Certain programs have additional requirements, such as the presence of young children in the family, income below some percentage of the median rental cost in the person’s region, or total family assets below a certain threshold. Most programs also are designed to “phase out” as family incomes rise—i.e., as a family’s income increases, benefits levels decrease at some proportional rate—such that higher labor earnings still result in a net increase in total (labor and non-labor) income. Medicaid eligibility, however, terminates above a specific income threshold.5

The EITC and CTC are slightly different. Qualifying beneficiaries of these wage-subsidizing tax credits will receive larger benefits as their wage income rises, up to a certain point. At that point, benefits plateau at a maximum amount for a set income range, and then begin to phase out beyond the maximum benefit range. Because of this structure, a low-income worker below the maximum benefit range could see her benefits increase as her wages went up, unless she experienced an income gain large enough to put her on the downslope of the phase-out range.6

Before analyzing the use of public assistance by working individuals, it is important to recognize that the majority of means-tested government income support goes to working families—i.e., families with at least one working family member within the household. As explained in Gould, Davis, and Kimball (2015), for better or worse, public assistance programs have become increasingly tied to work. Table 1 shows participation in means-tested public assistance programs among elderly and non-elderly families and individuals, as well as participation by families’ work status. The data show that 88.2 percent of families or individuals in non-family households who receive means-tested benefits are non-elderly families, and 93.4 percent of benefits (excluding Medicaid) go to non-elderly persons or families.7 Among all families and individuals receiving benefits, two-thirds (66.6 percent) are either working or in working families. Focusing on non-elderly families and persons, 35.5 percent participate in at least one public assistance program. Importantly, the data show that nearly three-quarters (71.6 percent) of these participants are either working or in working families. Similarly, 69.2 percent of all public assistance benefits going to non-elderly families go to working families. (Note that the total benefits amount reported in subsequent tables is larger than reported in Table 1 because the subsequent analysis of working recipients includes elderly workers and those in elderly families.)

Table 1

Utilization by annual hours of work

Inadequate levels of labor income can stem from both low hourly wages and a lack of adequate hours of work. Table 2 shows the receipt and value of benefits among workers and their families by annual hours of work. Workers are grouped by those working less than 1,000 hours per year, 1,000–1,499 hours per year, 1,500–1,989 hours per year, and 1,990 hours or more, with this final category constituting regular full-time employment. As the table shows, roughly two-thirds of all wage earners work annual hours that would constitute full-time, year-round employment. Among these workers, 21.9 percent receive some form of public assistance, either directly or through a family member. This is lower than the overall rate among all workers of 29.3 percent. In contrast, 46.6 percent of individuals working less than 1,000 hours per year receive some form of public assistance. Similarly, 42.9 percent of individuals working 1,000–1,499 hours per year (effectively “half time”) receive some public assistance, as do 35.5 percent of individuals working 1,500–1,989 hours per year.

Table 2

The distribution of beneficiaries also is informative. Individuals working less than 1,000 hours per year make up about 15 percent of all wage earners, although nearly one-fourth (23.7 percent) of workers who receive some form of public assistance, either directly or through a family member, fall into this group. Similarly, individuals working between 1,000 and 1,989 hours per year account for 22.0 percent of all workers, yet constitute 29.4 percent of all working public assistance beneficiaries. While it is not surprising that individuals with lower annual work hours are disproportionately represented among public assistance recipients, the share of working recipients who work full time is quite large, at 46.9 percent.

It is also illuminating to look at the distribution of beneficiaries across the various programs. Half of all working WIC and CTC beneficiaries work full time, as do 41.0 percent of working food stamp recipients and 38.0 percent of working Medicaid beneficiaries. The table also shows the distribution of program dollars by annual hours of work. Most of the benefit dollars are spread in roughly equal proportion to the share of beneficiaries in each work hour category, although EITC and CTC benefits predictably skew more heavily toward those working more hours.

It is worth noting that because of the structure of the data, with each individual record carrying the family’s total benefit information, it is possible that some of the work hour statistics reflect the work hours of family members who do not directly receive support from the applicable program. (See Appendix A for further detail.) The extent to which this biases the results is unclear. Nevertheless, the large shares of individuals working full time whose families participate in these programs underscores the question of whether these programs are serving more as temporary support during times of financial stress—the original intent of many of these programs—or whether they have become permanent wage subsidies for workers paid unlivable wages.

Utilization by hourly wage level

Figure A shows the share of all wage earners within each hourly wage decile whose families receive public assistance.8 As the figure shows, an estimated 29.3 percent of all wage earners receive benefits from at least one of the means-tested public assistance programs included in this study, either directly or through a family member in the household. Roughly 60 percent of all workers with hourly wages in the bottom decile (less than $7.42 per hour) receive benefits or have a family member receiving benefits—more than double the overall rate of receipt. Among workers in the second decile (whose hourly wages are between $7.42 and $9.91 per hour), just over half (52.6 percent) receive benefits either directly or through a family member. As expected, the rates of receipt decline steadily as hourly wages increase. (Appendix Table B2 shows all data on program utilization and benefit amounts by wage decile.)

Figure A

It is somewhat surprising that any workers with high hourly wages would show any receipt of public assistance. This is likely the result of the structure of the CPS-ASEC data, in which some households may contain both low-income/low-wage and higher-income/higher-wage tax units in the same family, such as a young adult living with his parents. In such cases, the data would show a higher-wage individual benefiting from the public assistance dollars of her low-wage family member when, in reality, the incomes of those individuals may be separate. In addition, some individuals who work only a small portion of the year may report relatively large wage income for those limited annual hours of work, resulting in an imputed hourly wage that is artificially high.

Nevertheless, even with these potential sources of error, the data still show that high-wage earners comprise a small share of working benefit recipients. As shown in Figure B, of all wage earners in families receiving public assistance, 77.8 percent have wages in the bottom half of the wage distribution. More than half (54.4 percent) have wages in the bottom 30 percent, and roughly 40 percent are in the bottom fifth of hourly wage earners.

Figure B

Figure C shows the total value of public assistance benefits received by workers, either directly or through a family member, by the worker’s hourly wage decile. Once again, the bulk of benefit dollars are going to workers with the lowest wages. Workers with wages in the bottom decile receive over $28 billion in benefits, while workers in the second decile—between the 10th and 20th percentiles—receive nearly $29 billion in benefits.9 The percentages below each dollar figure are each decile’s share of all benefit dollars received by workers and their families. Combining the bottom and second deciles, workers in the bottom fifth of the wage distribution—earning hourly wages of $9.91 or less—receive just less than half (46.4 percent) of all public assistance dollars accruing to workers and their families. Fully 86.8 percent of all public assistance benefits going to working families go to workers with hourly wages in the bottom half of the wage distribution.

Figure C

Utilization by industry

While there are workers who benefit from public assistance across all major industry groups, some industries have much greater numbers of workers receiving benefits and much higher rates of receipt. In Figure D, the green bar shows again that just over 29 percent of all workers have families receiving means-tested benefits. As the figure shows, workers in agriculture, forestry, fishing, and hunting have an exceptionally high rate of receipt: Half receive public assistance, either directly or through a family member. Workers in the arts, entertainment, recreation, accommodation, and food services industries have take-up rates of roughly 45 percent.10 Retail trade, construction, the armed forces, and other services except public administration all show rates of receipt above the overall average of 29.3 percent. Workers in public administration have the lowest rate of utilization, with only 16.1 percent of workers receiving benefits.

Figure D

Whereas Figure D showed rates of receipt within industries, Figure E shows the distribution of all wage earners receiving benefits across industries. The industry with the largest share of workers receiving benefits (either directly or through a family member) is educational, health, and social services, at 19.9 percent. However, this is also the largest industry category by far, containing roughly one-quarter (23.6 percent) of all wage earners in the sample—thus, their share of benefit recipients is actually disproportionately low. In contrast, retail trade combined with the arts, entertainment, recreation, accommodation, and food service industries contain almost one-third (30.5 percent) of all working public assistance recipients, despite accounting for just over one-fifth (21.9 percent) of the total workforce. (Appendix Table B3 contains detailed breakdowns of program participation and employment shares by industry.)

Figure E

Figure F shows the total value and share of public assistance dollars received by families of working individuals by the major industry employing each worker. Once again, the largest portion of benefit dollars goes to families of workers in education, health, and social services. Workers or families of workers in these industries receive nearly $25 billion in public assistance, or 20.3 percent of all benefit dollars received by workers in the sample. However, this is a smaller percentage than their 23.6 percent share of the workforce. The two major industry groups with disproportionately large shares of workers receiving benefits—retail trade, and the arts, entertainment, recreation, accommodation, and food service industries—also receive a disproportionately large 32.1 percent share of all benefit dollars going to workers (equal to roughly $39.5 billion). Note also that this share of benefit dollars exceeds these industries’ share of recipients (30.5 percent), an indication that wages and incomes for workers in these industries are particularly low.

Figure F

Utilization by state

Unsurprisingly, the states with the largest shares of the country’s program beneficiaries are the most populous states: California, Texas, New York, Florida, and Illinois. However, as depicted in Figure G, the states with the highest rates of public assistance usage are primarily those in the South and Southwest: New Mexico, Mississippi, Arkansas, Arizona, and Louisiana. Southern states, in particular, tend to have lower wages and higher poverty rates (Cooper 2015b; BLS 2014). Accordingly, they receive disproportionate shares of public assistance dollars. For example, as shown in Appendix Table B4, Georgia accounts for 3.0 percent of U.S. workers, yet the state’s workers receive 3.6 percent of all public assistance dollars going to workers. Similarly, Louisiana has 1.3 percent of the U.S. workforce and receives 1.9 percent of public assistance dollars going to workers. Mississippi contains 0.8 percent of the total working population, yet receives 1.4 percent of benefit dollars going to workers. These differences are not enormous, due in part to the fact that hourly wages have converged across the states over the past four decades.11 Nevertheless, they do underscore the relationship between low hourly pay and greater reliance on public assistance. Moreover, because these figures only describe benefit dollars going to workers, they almost certainly understate the disproportionate share of all public assistance going to working and nonworking families in these states, as many of them also have lower overall rates of employment than those states with lower rates of public assistance receipt among workers.12

Figure G

How higher wages would affect public assistance programs

This section examines how changes in hourly wages affect the participation rates and value of benefits received from means-tested public assistance programs. The analysis uses linear regression to isolate how receipt of public assistance by workers at one hourly wage compares with receipt by similar workers at different hourly wages. Because most programs have income restrictions, as hourly wages rise, program participation and costs generally decline. However, because some refundable tax credits—such as the Earned Income Tax Credit—are designed so that benefits increase, up to a point, as labor income rises, it is not theoretically obvious that increasing wages among all low-income workers would necessarily lead to a net reduction in overall public assistance spending.

Effects on the likelihood of receiving benefits

Table 3 summarizes the results of this analysis.13 Each column indicates the wage range these results describe; each row indicates the relevant public assistance program. The upper block of results shows how a $1 increase in hourly wages among workers in each wage range affects the likelihood that they (or their families) participate in the specified program.

Table 3

The results in the first column, which describe workers in the bottom two deciles (with hourly wages up to $9.91), indicate that a $1 increase in hourly wages for a worker in this range reduces the likelihood that she receives any public assistance by 3.8 percentage points. This means that if this entire group of workers received, on average, a $1 per hour raise, we would expect the share of this group needing to rely on public assistance to decline by 3.8 percentage points. With 15.5 million workers in this wage range, that translates to roughly 600,000 fewer people receiving benefits.

For each individual program, including the EITC, increasing hourly wages is expected to reduce the rate of program participation. Among those in the bottom two deciles, the magnitude of these effects ranges from a reduction of around 0.1 percentage points in the share receiving benefits from WIC or TANF to a reduction of 4.1 percentage points in the share receiving the EITC.14 For SNAP, a $1 increase in hourly wages reduces the participation rate by 2.0 percentage points; for Medicaid, 1.3 percentage points; for LIHEAP, 0.4 percentage points; and for housing assistance, 0.3 percentage points. It is not surprising that the effects for TANF, WIC, housing assistance, and Medicaid are substantially smaller than the effect for the EITC—and to a lesser extent SNAP—simply because the share of workers participating in each of these programs is considerably smaller. As expected, the largest impact for workers in this range is on the EITC, since the EITC has the highest rate of program participation for this group (about 60 percent) and overall (roughly 29 percent).

The second column describes a broader range of workers—those earning wages in the bottom three deciles, or up to $12.16 per hour. As the sample of workers expands to include those earning higher wages, the overall rate of program participation declines because, on average, higher wages lead to higher family incomes and reduced program eligibility. This, in turn, leads to an average effect that is smaller for this larger group than for the first two wage deciles. Yet because the pool of workers has expanded, a $1 average increase in wages for this larger group would still lead to a larger reduction in the number of people receiving public assistance. For every $1 increase in hourly wages for the roughly 27.5 million workers earning up to $12.16 an hour, the share relying on public assistance is predicted to decline by 3.1 percentage points, or approximately 850,000 people. An equivalent way to interpret these results is that the number of low-wage workers relying on public assistance could be reduced by 1 million people with a wage increase of $1.17 an hour, on average, among the lowest-paid 30 percent of workers.

Similar calculations can be made for the third column, which expands the sample of wage earners to include those in the fourth decile, who earn up to $14.72 per hour. Using the results for this group, a $1 average increase in hourly wages would reduce the share receiving means-tested benefits by about 2.5 percentage points. When multiplied by the 40.7 million workers in this range, this means just over 1 million fewer workers would receive benefits.

The fourth and fifth columns in the table are slightly different. Rather than describing effects at hourly wages up from the bottom of the distribution, they describe effects limited to the third and fourth, and fourth and fifth, deciles of the wage distribution, respectively. In limiting the pool of workers to these groups, the results generally show how at higher hourly wage rates, the anticipated reduction in program utilization from raising wages becomes smaller, simply because a smaller share of workers in these wage ranges receive benefits. The one noticeable exception, however, is participation in the refundable Child Tax Credit. It is likely that the higher observed effects for this group result from the unique way the CTC is structured. Although the Child Tax Credit is available for families earning high annual incomes, it appears that in these hourly wage ranges, workers’ tax liabilities, on average, are becoming large enough to more significantly reduce their likelihood of receiving the refundable portion of the credit. In other words, though these workers may still be taking the Child Tax Credit against the taxes they owe, at these wages they may start to owe more in taxes than the maximum value of the credit.15

The results in column four (workers with hourly wages between $12.16 and $14.72) and column five (wages between $14.72 and $17.62) are similar. As expected, the decline in the likelihood of receiving any benefits is smaller for the higher wage group because a smaller share of this group receives benefits. Among individual programs, the expected reduction in benefit receipt is noticeably smaller for SNAP, meaning that higher wages for workers in this range of the wage distribution are not going to have the same magnitude of effect as they would for raising wages among lower-paid workers. Similar, albeit smaller, reductions in magnitude are evident for Medicaid and housing assistance. Among the other programs—the EITC, CTC, WIC, and TANF—the effects are not meaningfully different between workers in the third and fourth deciles (column 4) and those in the fourth and fifth deciles (column 5).

Effects on program spending

The lower block of results in Table 3 shows the predicted effect from a $1 increase in hourly wages on the annual value of benefits received by workers in each wage range. Again looking at the first column, the results indicate that for each additional dollar in hourly wages paid to workers earning up to $9.91 per hour, total benefit dollars received from all public assistance programs decline by $199 per worker annually, on average. This means that if the 15.5 million workers with wages in this range received, on average, a $1 increase in hourly pay, total means-tested government benefit expenditures would decline by about $3.1 billion annually.

For the bottom two deciles, the predicted change in benefit dollars is negative for all programs, although the effects are not statistically distinguishable from zero for the refundable Child Tax Credit (CTC) and TANF.16 The estimated effects are largest for SNAP, with each dollar increase in hourly wages resulting in a $93 decline in annual benefits per worker. Extrapolating this effect to the full population of workers in this wage range, a $1 average raise for these workers would reduce annual SNAP expenditures by $1.45 billion.17 The predicted per person effects for the other programs are an average annual decline of about $76 for the EITC, $14 for housing assistance, $8 for TANF, $4 for WIC, $2 for LIHEAP, and $2 for the CTC.

Note that the sample for these calculations includes workers who do not receive any benefits from these programs. Including these non-participating individuals—recorded as receiving zero benefits for the relevant program—makes the results considerably smaller than they would be if the analysis were restricted to only those receiving benefits from each program. This is why the values for CTC, LIHEAP, TANF, and WIC are so small; the share of all wage earners in the bottom two deciles receiving benefits from these programs is quite low, meaning that the average effect among all workers in this range (the majority of whom receive no benefits from these programs) is necessarily going to be small. Excluding the zero values from the sample would more accurately describe how wage increases affect individual benefit amounts among beneficiaries, but that is a different question from what is being explored here. This report’s focus is on the average effect raising wages would have on aggregate program spending, which requires accounting for those for whom there is no effect.

As with the results for program participation, when the group of workers in the sample expands to include those in the bottom three deciles (with hourly wages up to $12.16), the average effect on benefit amounts shrinks for most programs, with the exception of TANF, the EITC, and the CTC. As far as the tax credits are concerned, this result is consistent with the structure of these credits, wherein higher earned income results in a larger benefit amount, up to a maximum benefit, which is then gradually reduced after workers and their families reach higher income levels. The fact that expanding the pool of wage ranges in the sample increases the magnitude of the effect on EITC and CTC benefits indicates that even though some workers in this group may be getting larger EITC and CTC credits from higher wages, most have maxed out their benefits, and some are likely on the downslope of the benefit curve. The results indicate that for every $1 average increase in hourly wages for the 27.5 million workers earning up to $12.16 per hour, we should expect a decline in EITC expenditures of roughly $80 per person annually, or $2.2 billion overall. Such a pay raise for workers in the bottom three deciles would reduce total annual expenditures on all means-tested programs included in this study by $189 per worker, or roughly $5.2 billion overall.

The differences in the effect of higher wages on benefit amounts shown in the fourth and fifth columns—where wages are limited to the third, fourth, and fifth deciles—are similar to the differences in effects on program participation. Workers in these higher wage ranges are generally less likely to participate in most of these programs, thus there are more observations with a zero value. Consequently, the average effect for all workers in these wage ranges is smaller. The EITC and CTC are again the exceptions; workers in these higher wage ranges who are still receiving benefits from these programs are more likely to be on the downslope of the benefit curve—i.e., they are either receiving the maximum benefit amount, or have begun to see their benefits reduced due to their higher incomes. That effects for these programs among workers in the fourth and fifth wage deciles are only slightly larger than effects for workers in the third and fourth deciles reflects that benefits in these programs are reduced at a constant rate.

Still, even with these larger effects for the EITC and CTC, the aggregate effect on all means-tested programs is smaller among workers in these higher wage ranges than it is for lower-paid workers. A $1 average increase in hourly pay for workers in the fourth and fifth deciles (column five) is expected to reduce annual means-tested program expenditures by $155 per person, roughly three-fourths the magnitude of effect for workers in the bottom two deciles (column one).

Policy implications

Despite the relatively small average effects for individual programs, it is clear that broadly increasing wages among low-wage workers would lead to sizable declines in aggregate public assistance spending. For example, if a randomly selected 10 million workers with wages in the bottom four deciles—just less than a quarter of the 40.7 million workers earning up to $14.72 per hour—received a $1 increase in their hourly pay, 250,000 fewer workers would receive benefits, and annual public assistance outlays would fall by about $1.8 billion. If more workers received a pay raise or the wage increase were larger, the expected savings would be bigger. The exact mechanism used to spur higher wages—be it active policy interventions or market forces—could influence the particular subgroup of workers that would be affected, which could change the magnitude of the effects. Predictably, higher wages among the lowest-paid workers would produce the largest program savings. Yet even among more moderate-wage workers, lifting wages reduces the likelihood that a worker relies on public assistance, and is likely to free program dollars that could then be repurposed in any number of ways—including directing them to the neediest workers and families.

One obvious way to raise wages for low-wage workers is to increase the minimum wage at the federal and state levels. As explained in Cooper, Schmitt, and Mishel (2015), the federal minimum wage is far below historical levels according to every relevant benchmark. Raising the federal wage floor would provide a long-overdue boost to the pay of millions of working Americans and reduce the implicit subsidy that taxpayers are giving to employers who pay inadequately low wages. Conveniently, the 30th percentile wage cutoff discussed in this report aligns reasonably well with the minimum-wage level proposed in the Fair Minimum Wage Act of 2015, which would raise the federal minimum wage to $12 by 2020.18 As discussed previously, among workers earning less than $12.16 per hour, every $1 increase in hourly wages is predicted to reduce the share of workers relying on public assistance by 3.1 percentage points, and provide average annual savings to government programs of $190 per worker. Cooper (2015a) estimates that increasing the federal minimum wage to $12 by 2020 would directly raise the wages of 28 million workers, giving affected workers an average pay increase of $3.16 per hour. Multiplying this average raise by the total number of affected workers and the measured effects in the second column of Table 3 indicates that such a raise would reduce the number of Americans relying on public assistance by 2.7 million workers. The estimates on benefit amounts indicate that raising the minimum wage to $12 by 2020 would generate $17 billion in annual savings to means-tested government assistance programs. Again, this savings estimate is conservative because it does not include savings from Medicaid, nor does it account for workers earning wages above the $12 minimum wage who might also get a raise as employers adjusted overall pay scales.

Whatever the mechanism used to achieve higher wages, the potential for substantial savings to government assistance programs should motivate policymakers in both parties to prioritize raising wages for low-wage workers. In fiscal 2014, the seven programs studied in this report distributed roughly $190 billion in benefits to workers and families needing help to get by.19 Thus, annual savings of $17 billion, as predicted from a $12 minimum wage, would equal about 9 percent of total benefit outlays. Clearly, raising the minimum wage would not eliminate the need for these programs, nor would it dramatically change their combined overall cost. Yet $17 billion in annual savings is more than enough to significantly improve or expand upon existing assistance programs or fund new tools amplifying government’s capacity to improve people’s lives. For example, both the president and congressional leaders have proposed expanding the EITC to provide greater benefits to childless adult workers, who currently are largely excluded from the program.20 The president’s proposal is estimated to cost roughly $6 billion a year. If enacted in tandem with a minimum-wage increase, the cost would be somewhat higher because some workers would receive larger benefits from their higher wages, yet it would still be well below the savings in other programs resulting from the higher wage floor.21

There are numerous other ways such savings could strengthen the safety net, fund new priorities, or make job-creating investments. Savings of $17 billion are enough to double housing assistance benefits—either increasing benefits to current recipients or expanding eligibility. It is enough to triple TANF benefits or quadruple WIC or LIHEAP benefits. It is enough to fund new education proposals from the president’s budget, such as the “Preschool for All” initiative that would provide 4-year-olds in low- and moderate-income families with access to preschool, or the federal component of the president’s proposal for tuition-free community college (OMB 2015a; OMB 2015b). By raising the minimum wage to $12 by 2020, Congress could generate the funding required to support these worthwhile programs, and potentially still have savings left over.

Conclusion

Raising wages among low- and middle-wage workers would simultaneously lift incomes and reduce spending on public assistance programs. The government could then use these savings to bolster anti-poverty efforts or make new job-creating investments. Increasing the federal minimum wage is one simple way this could be achieved—though it is not the only way. As explained in EPI’s Agenda to Raise America’s Pay, we can raise wages by eliminating the lower subminimum wage for tipped workers, updating overtime protections, strengthening workers’ ability to organize and negotiate with employers collectively, improving enforcement of labor laws, providing undocumented immigrant workers a path to citizenship, and ensuring monetary policy prioritizes full employment. These policies would help undo decades of wage stagnation that have prevented greater improvement in living standards for the vast majority of American households. They would also bring greater balance to the roles that the private and public sectors play in improving American workers’ quality of life.

— This research is supported by contributions from The Nick and Leslie Hanauer Foundation.

About the author

David Cooper is an economic analyst with the Economic Policy Institute. He conducts national and state-level research on a variety of issues, including the minimum wage, employment and unemployment, poverty, and wage and income trends. He also provides support to the Economic Analysis and Research Network (EARN) on data-related inquiries and quantitative analyses. David has been interviewed and cited by numerous local and national media for his research on the minimum wage, poverty, and U.S. economic trends. He holds a Master of Public Policy degree from Georgetown University.

Appendix A: Methodology and data sources

This study uses three years of microdata from the Current Population Survey Annual Social and Economic Supplement (CPS-ASEC), describing respondents’ economic conditions in calendar years 2012–2014. The CPS is the monthly survey used by the Census Bureau to track a variety of labor market indicators, including the unemployment rate. The ASEC is a set of additional questions asked each March about respondents’ economic status in the preceding year. It is used to calculate the government’s official measures of family income, assess sources of income, and determine poverty rates.

In addition to the ASEC data, this paper also incorporates microdata from the Census Bureau’s Supplemental Poverty Measure (SPM) public-use research files. The SPM is an alternative poverty measure developed by the Census Bureau that takes a more holistic appraisal of both family income sources and family expenses. The CPS-ASEC files contain information on most earned and some unearned income. The SPM data adds information on several means-tested public assistance programs, such as income from the Earned Income Tax Credit (EITC), the refundable portion of the Child Tax Credit (CTC), the Low-Income Home Energy Assistance Program (LIHEAP), and the Supplementary Nutrition Program for Women, Infants, and Children (WIC).

While the CPS-ASEC and SPM data provide excellent information on annual incomes—including income from wages—the survey’s focus on annual income creates challenges for assessing how increases to hourly wages might affect public assistance levels. Low annual incomes could be the result of low hourly wages, inadequate annual hours of work, or some combination of the two. While a thorough examination of these influences is beyond the scope of this paper, this issue is discussed in Appendix C.

The hourly wage information used in this report is developed as follows: The CPS-ASEC survey asks respondents to describe their annual income and sources of income over the preceding year. Respondents also report the number of weeks they worked during that year, and the usual number of hours they worked in the weeks that they worked. With these three pieces of information, one can impute each individual’s implied hourly wage for the time they were working.

Admittedly, these implied hourly wages from the ASEC data are less robust than other sources of hourly wage information, such as the wage data from the Current Population Survey Outgoing Rotation Group. As Giannarelli, Morton, and Wheaton (2007) note, imputing hourly wages compounds measurement error from the three variables used in the imputation process, and can produce hourly wage values that fall below the statutory minimum wage of $7.25. However, as explained in Appendix C, these implied subminimum values, while certainly the product of error in many cases, may be indicative of troubling labor practices in others—and may even be accurate in others still. In any case, the ASEC data is one of the only public datasets with information on income from means-tested public assistance programs, and while not ideal, the imputed hourly wages provide adequate measures with which to assess utilization of these programs by relative levels of imputed hourly wages.

Another potential source of error comes from the fact that in the CPS data, receipt of income from public assistance is recorded as the total income received by all members of the family—meaning that each individual observation carries the family’s total income from public assistance. In order to assess the relationship between individual hourly wages and public benefit receipt, in this analysis, for families with multiple workers, income from public assistance programs is divided evenly among all workers. In a limited number of cases, this may create instances where the data report benefits received by individuals of the same family, when in actuality the individuals are separate tax units—such as adult siblings living together—where one of the respondents is not actually receiving benefits. Nevertheless, these are likely a small number of cases, and even in such cases, some portion of income may still be shared across tax units.

Finally, it is well known that the CPS-ASEC data significantly understate actual participation in, and income received from, public assistance programs. See Wheaton (n.d.) or Meyer, Mok, and Sullivan (2009) for details. Consequently, the CPS-ASEC data in this report have been adjusted to be consistent with administrative data that show actual program enrollment and expenditures for all programs except LIHEAP.22 The adjustments are made using the same method of reweighting recipient observations, non-recipient observations, and reported benefit amounts as in Zabin, Dube, and Jacobs (2004) and Allegretto et al. (2013). Still, because of changes in Medicaid resulting from the Affordable Care Act, data from the period in this study may not have good predictive value in estimating effects on the value of Medicaid benefits. For this reason, the fungible value of Medicaid benefits is not included in any calculations throughout this study. Consequently, these results understate the true total public costs of public assistance received by workers and their families.

The sample for the descriptive statistics and regression analysis consists of the three years of CPS-ASEC data from March 2013, 2014, and 2015, reflecting respondents’ economic conditions from 2012 to 2014. The sample is restricted to individuals age 16 and older who worked for some portion of the year, and for whom a valid hourly wage value can be determined.

The regression analysis uses two sets of regressions for each subsample of wage earners. One set of regressions has binary dependent variables indicating whether the respondent or a family member in the household reported participation in, or positive income from, any of the eight public assistance programs under study. In the second set, the dependent variables are the total dollar value—in level terms—of benefits received from each program, and the sum of all benefits from all programs. For both sets of regressions, the explanatory variable of interest is the coefficient on real hourly wages. Models were also estimated for total benefits, the EITC, and the CTC containing real hourly wages in quadratic form in order to capture likely changes in the direction of the effect. These results are discussed in the text, and a full listing of these results is available upon request. All models employ a standard set of demographic controls (age, age-squared, age-cubed, gender, race, citizenship, marital status, and metropolitan status), as well as controls for industry, major occupation group, state, family size, number of children in the family, the number of wage earners in the family, whether the individual worked part time at any point during the year, and the presence of any disabled persons in the household.

Appendix B: Additional tables and figures

Appendix Table B1

Appendix Table B2

Appendix Table B3

Appendix Table B4

Appendix Table B5

Appendix C: Discussion of imputed subminimum hourly wage values in the CPS-ASEC

For the vast majority of observations within the ASEC sample, imputing an hourly wage from the respondent’s reported annual wage income, weeks worked per year, and usual hours worked per week yields a plausible hourly wage value. However, for a small but not insignificant portion of the sample—9.9 percent—imputed wage values fall below the federally mandated minimum wage of $7.25 per hour. Some of these values fall below the wage floor almost certainly due to measurement error compounded by the imputation process. It is understandably difficult for some individuals to recount accurately their total wage income and total time in the workforce for a 12-month period three months removed—especially if they changed jobs, worked only a portion of the year, or worked inconsistent hours.

Appendix Figure CA shows the distribution of all imputed wage values below the federal minimum wage (hereafter referred to as “subminimum wage values”) and the distribution of all imputed wage values, both as shares of their respective totals, by respondents’ reported usual hours worked per week. The distribution shows that subminimum wage values disproportionately occur among individuals who worked less than 35 hours per week. Appendix Figures CB1 and CB2 show the distribution of subminimum wage values and total wage values by respondents’ reported weeks worked in the previous year. As expected, subminimum wage values disproportionately appear among part-year workers.

Appendix Figure CA

Appendix Figure CB1

Appendix Figure CB2

The Census Bureau takes steps to validate responses, and reduce potentially invalid data, but unfortunately, the low-wage labor market is prone to characteristics that can aggravate measurement error: Turnover is high, work is more likely to be seasonal or part time, and many low-wage jobs suffer from inconsistent hours (CBO 2006). This means that workers whose actual hourly wage was above yet close to the minimum wage may be particularly prone to err in their reporting of any of the three data points needed to impute hourly wages.

Given these challenges, some researchers have been wary of using these implied hourly wages (see Giannarelli, Morton, and Wheaton 2007). Yet rather than simply dismissing these questionable values outright, a more thorough examination suggests that perhaps not all these subminimum wage values are the result of measurement error. In fact, some may be indicative of significant gaps in legal protections, and troubling real-world labor practices.

For example, certain groups of workers are exempt from the minimum-wage provisions of the Fair Labor Standards Act (FLSA), such as farmworkers on small farms, employees of some seasonal and recreational establishments, fishermen, newspaper delivery workers, and anyone employed by a business with less than $500,000 in annual revenue that does not engage in interstate commerce (U.S. Department of Labor 2014). For workers that fall into these categories, their low imputed hourly wage values may be entirely correct and legal.

For others, however, implied hourly wages below the minimum wage may indicate greater incidence of wage theft. As explained in Meixell and Eisenbrey (2014), wage theft—the practice of employers not paying workers the full wages that they are owed—is a significant problem, particularly in low-wage jobs, that costs American workers hundreds of millions, if not billions, of dollars each year.23

Appendix Tables C1 and C2 show the occupations with the highest shares and highest incidence, respectively, of subminimum imputed wage values. Bolded occupations are those that appear in both tables. Many of the listed jobs represent the lowest-paying jobs in the economy, and indeed, several occupations that appear in this list may not be covered by the FLSA or other federal labor and employment laws, such as home health aides, personal and home care aides, and agricultural workers.24

Appendix Table C1

Appendix Table C2

Several other listed occupations are occupations where workers customarily receive tips as a portion of their wages: waiters and waitresses; counter attendants, cafeteria, food concession, and coffee shop workers; non-restaurant food servers; bartenders; bartender helpers; and hairdressers, hairstylists, and cosmetologists. Allegretto and Cooper (2014) explain how problems of wage theft and workers receiving wages below the prevailing minimum wage are particularly acute among tipped workers due to “tip credit” provisions in minimum-wage laws that allow employers of tipped workers to pay them a base wage as low as $2.13 per hour. While these employers are legally required to make up any shortfalls between the effective hourly rate earned by tipped workers from their tips and the prevailing minimum wage, enforcement of this requirement is highly problematic, and there is evidence of considerable abuse. Indeed, these data suggest that when workers in these occupations are asked to tally their total annual earnings, weeks of work, and usual hours, a high percentage report figures that imply hourly pay below the minimum wage.

Finally, immigrants comprise many of the workers in these occupations. In fact, of the 15 occupations most commonly held by immigrants, 10 appear in Appendix Table C2: cooks; housekeepers, maids, and butlers; nursing aides; janitors; truck, delivery, and tractor drivers; construction laborers; cashiers; gardeners and groundskeepers; retail sales clerks; and farm workers.25 Some of these immigrant workers may be undocumented, while others may be working on temporary work visas that restrict their ability to change jobs—both scenarios that can leave these workers powerless against exploitive employers. Such practices are damaging not just for the immigrants themselves, but for all other workers (immigrant and non-immigrant) in the same occupations. Any time an employer can exploit certain groups of workers and pay wages lower than would otherwise be possible, it places downward pressure on the wages of other workers with the same jobs. For this reason, the high prevalence of subminimum wages in occupations common to immigrants need not be instances of wage theft solely among immigrant workers.

Appendix Tables C3 and C4 examine the prevalence of subminimum-wage values by industry, showing the industries with the highest share and highest incidence, respectively, of wage values below the federal minimum. Once again, bolded industries are those that appear in both tables. As with the distribution by occupation, one can surmise plausible reasons for why workers in many of these industries may be underpaid for the total hours they worked in a given year, thus bringing their implied hourly wage below the federal minimum. Subminimum wages could be the result of unscrupulous employers or gaps in labor protections. For example, restaurants again appear prominently in both lists, which is expected given the high incidence of subminimum values for waiters and waitresses, cooks, and other food service occupations. Workers in recreational parks and camps (e.g., campgrounds) have the highest rate of imputed subminimum values, likely because many such facilities are seasonal and not covered by the FLSA. Similarly, some workers in animal production facilities, crop production facilities, and home health care services industries also lack FLSA protections. Workers in private households also show high incidence of subminimum imputed wage values, which could be the result of informal or “under the table” arrangements, which Shierholz (2013) notes leave workers particularly vulnerable to violations of labor standards. Finally, as with the top occupations with the highest incidence of subminimum-wage values, many of these industries are also disproportionately staffed by immigrants, some of whom may be vulnerable to exploitation due to their immigration status.

Appendix Table C3

Appendix Table C4

To be clear, these data cannot and should not be viewed as conclusive evidence of wage theft or labor abuses in any particular occupation or industry. However, the high prevalence of what should, in many cases, be impossibly low imputed wage values should raise questions about how workers are being paid and the hours they are expected to work in many of these jobs.

To the extent that these subminimum-wage values represent instances of wage theft or gaps in labor standards, better enforcement of labor law or expansions in coverage to workers either outright excluded from the FLSA (such as seasonal farm workers) or treated as a separate class of workers (such as tipped workers) would yield additional public savings.

Appendix Table C5 shows rates of receipt and the total value of benefits paid out to workers with imputed hourly wages below $7.25. Among these workers, 59.8 percent report receiving some form of public assistance, with total benefits from all programs totaling nearly $29 billion. Again, the program with the highest participation rate is the EITC—60.4 percent of workers in this wage range receive roughly $12.8 billion in benefits. Participation in SNAP is also common, with 26.2 percent of workers in this group receiving $12.1 billion in benefits. Participation rates for the other programs are 21.0 percent for Medicaid, 3.7 percent for WIC, 4.3 percent for LIHEAP, 4.4 percent for housing assistance, and 3.1 percent for TANF. The average total amount of benefits received among beneficiaries of any program is about $3,500, while the largest average benefit among the individual programs is for recipients of Section 8 housing vouchers, who receive an average benefit worth $3,900.

Appendix Table C5

The regression results for this group are presented in Appendix Table C6. As with the results for workers with wage values in the valid range, increasing wages for this group is predicted to reduce the share receiving public assistance—in this case, by 2.0 percentage points from every $1 increase in average hourly wages. For the EITC, participation declines significantly, by 1.8 percentage points for every $1 increase in wages; for SNAP, the decline is 1.2 percentage points, and for Medicaid, the decline is 0.7 percentage points. The coefficients values for the other programs are either statistically insignificant or not economically meaningful.

Appendix Table C6

In the regressions on the value of benefits, the results are mixed. The coefficient on the value of all benefits is positive, suggesting that benefits for workers in this group could increase in aggregate on average if their wages rose, although this result is not statistically significant. We can see, however, that this positive value is driven, as expected, by the EITC and CTC. The results for these programs show that for every additional dollar in hourly wages, aggregate EITC and CTC dollars for workers in this group would increase by an average of approximately $89 and $47 per worker, respectively. At the same time, the value of benefits from SNAP and housing assistance would decrease by an average of about $75 and $24 per worker, respectively. Values for the other programs are not statistically distinguishable from zero.

In light of the fact that these workers’ imputed hourly wages are subject to considerable error, it is not clear whether we can draw any strong conclusions from these results. Nevertheless, the regression results do suggest that in cases where reported subminimum values are the result of wage theft or gaps in FLSA coverage, workers may be entitled to greater income both from their employer and from the tax-and-transfer system.

Endnotes

1. Costs of basic necessities and the levels of income required to achieve a modest but adequate standard of living are discussed in detail in Gould, Cooke, and Kimball (2015).

2. CBO (2014, 3).

3. In an update to this report, West (2015) estimates that increasing the federal minimum wage to $12 by 2020 would reduce SNAP expenditures by $5.3 billion annually, or $53 billion over 10 years.

4. This report does not include any data on participation or costs in the National School Lunch Program, the only other large national means-tested public assistance program that does not specifically target the elderly or disabled.

5. This threshold is significantly higher in states that expanded Medicaid through the Affordable Care Act. In fact, because Medicaid costs for participants who gained eligibility from the Affordable Care Act’s expansion are covered almost entirely by the federal government, states that expanded Medicaid stand to achieve considerable state government savings in Medicaid if existing participants’ incomes are raised. West and Reich (2014b) describe these effects in detail.

6. For more information on the structure or function of the EITC or CTC, see Marr et al. (2015) or Center on Budget and Policy Priorities (2015).

7. This report focuses on utilization of means-tested public assistance; thus, it does not include Social Security or Medicaid, as they are programs available specifically and universally for retired or elderly people.

8. Wage decile information is presented in Appendix Table B1.

9. Note that throughout this report, Medicaid is included for all calculations of participation in assistance programs; however, the fungible value of Medicare benefits are not included in any estimates of program costs or value of benefits.

10. This finding is consistent with Shierholz (2014), which finds particularly high poverty rates among restaurant workers, and Allegretto et al. (2014), which finds that more than half of full-time fast food workers were enrolled in some public assistance program.

11. See Cooper, Schmitt, and Mishel (2015).

12. For example, from 2012 to 2014, Georgia’s employment-to-population ratio was 57.8 percent, Louisiana’s was 55.8 percent, and Mississippi’s was 51.5 percent. The national average for this period was 58.7 percent.

13. The results in Table 3 are all produced using a linear estimation model, meaning that the effects only describe average effects across the range of wages in each group. This provides suitable results for predicting overall changes in participation rates and benefit amounts. However, because the EITC and CTC have non-linear benefit structures, more precise effects at particular wage levels can be better estimated using nonlinear models. For this study, we did produce estimations with a quadratic form of the hourly wage variable. The results showed no meaningful difference in effects from the linear models at the average of each wage range; thus, we only report the linear results. However, we are happy to provide the quadratic regression results upon request.

14. The effects for WIC and TANF in this wage group are not statistically distinguishable from zero. This is likely because of the relatively small number of workers participating in these programs captured in this wage range, but it may also be that other factors—e.g., the recent birth of a child—are far more significant in determining participation in these programs.

15. The Child Tax Credit is structured so that workers may receive 15 percent of their earnings after the first $3,000 back as a credit to their tax liability, with a maximum credit of $1,000 per dependent child below age 17. They will only receive the refundable portion of the credit if their tax liability does not exceed the value of their credit. See Marr et al. (2015) for further detail.

16. Because the hourly wages in this study were constructed using reported annual wages and hours worked, it is impossible to isolate effects from changes in work hours from changes in wages. Workers at the lowest wages are also the most likely to work limited hours. TANF participation requirements vary across states, although income limits tend to be very low. Thus, it is likely that factors beyond hourly wage rates that limit workers’ annual income, such as a lack of adequate hours, may be confounding the results for this program. This is also true for the CTC; however, the program’s unique structure, in which workers’ first $3,000 in earnings are excluded from eligibility, may also play a role.

17. This finding is similar to that of West and Reich (2014a), who estimated that for every 10 percent increase in the minimum wage, SNAP expenditures fall by 1.9 percent. They estimate that increasing the minimum wage to $10.10 would reduce program outlays by nearly $4.6 billion. By the estimates in this report, if workers in the range likely to benefit from a $10.10 minimum wage received an average hourly wage increase of $1.61—as calculated in Cooper (2014)—it would reduce SNAP expenditures by $4.1 billion.

18. Adjusting the proposed $12 in 2020 minimum wage for expected inflation back to 2014 dollars would likely place it closer to the 25th percentile rather than the 30th percentile value of $12.16 measured in this report’s data. However, this does not include potential spillover effects above the $12 minimum. Moreover, these figures do not include the fungible value of Medicaid benefits and do not account for underreporting in LIHEAP participation, meaning that this estimate is still arguably conservative.

19. Author’s analysis of administrative data for each program. For the EITC and CTC, fiscal 2014 data were not available at the time of publication, so fiscal 2013 data were used as a substitute.

20. See Council of Economic Advisers et al. (2014) or Reeves and Venator (2014) for details on the proposal.

21. Sawhill and Karpilow (2014) estimate that the EITC expansion coupled with a federal minimum-wage increase to $10.10 would cost $10 billion annually, with savings in other means-tested programs equaling $11 billion. With the increasingly negative effects on aggregate EITC spending observed in this report, one can assume that adopting the EITC expansion with a higher minimum wage would result in even greater net savings.

22. Comprehensive administrative data for LIHEAP are not available; thus, no adjustment was made to reported LIHEAP participation rates and benefits amounts.

23. See also Greenhouse (2014).

24. These data are from 2010 to 2012. Home health care workers and aides were brought under the FLSA in 2014; however, farm workers on small farms remain exempt from the FLSA’s wage protections.

25. Author’s analysis of American Community Survey microdata, pooled sample 2007–2011. Data compiled by Ruggles et al. (2010).

References

Allegretto, Sylvia, and David Cooper. 2014. Twenty-Three Years and Still Waiting for Change: Why It’s Time to Give Tipped Workers the Regular Minimum Wage. Economic Policy Institute, Briefing Paper #379.

Allegretto, Sylvia, Marc Doussard, Dave Graham-Squire, Ken Jacobs, Dan Thompson, and Jeremy Thompson. 2013. Fast Food, Poverty Wages: The Public Cost of Low-Wage Jobs in the Fast Food Industry. U.C. Berkeley Center for Labor Research and Education.

Belman, Dale, and Paul J. Wolfson. 2014. What Does the Minimum Wage Do? W.E. Upjohn Institute for Employment Research.

Bernstein, Jared. 2014. “The Minimum Wage Increase and the CBO’s Job Loss Estimate.” On The Economy (Jared Bernstein’s blog), February 23.

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, Briefing Paper #378.

Bureau of Labor Statistics (BLS) Occupational Employment Statistics. 2014. “May 2014 State Occupational Employment and Wage Estimates.”

Center on Budget and Policy Priorities (CBPP). 2015. Chart Book: The Earned Income Tax Credit and Child Tax Credit.

Congressional Budget Office (CBO). 2006. Changes in Low-Wage Labor Markets Between 1979 and 2005.

Congressional Budget Office (CBO). 2014. The Effects of a Minimum-Wage Increase on Employment and Family Income.

Cooper, David. 2014. Raising the Federal Minimum Wage to $10.10 Would Save Safety Net Programs Billions and Help Ensure Businesses Are Doing Their Fair Share. Economic Policy Institute, Issue Brief #387.

Cooper, David. 2015a. Raising the Federal Minimum Wage to $12 by 2020 Would Lift Wages for 35 Million American Workers. Economic Policy Institute, Briefing Paper #405.

Cooper, David. 2015b. “In Virtually Every State, the Poverty Rate is Still Higher than Before the Recession.” Working Economics (Economic Policy Institute blog), September 18.

Cooper, David, John Schmitt, and Lawrence Mishel. 2015. We Can Afford a $12.00 Federal Minimum Wage in 2020. Economic Policy Institute, Briefing Paper #398.

Council of Economic Advisers, National Economic Council, U.S. Office of Management and Budget, and U.S. Treasury Department. 2014. The President’s Proposal to Expand the Earned Income Tax Credit.

Current Population Survey Annual Social and Economic Supplement microdata. Various years. Survey conducted by the Bureau of the Census for the Bureau of Labor Statistics [machine-readable microdata file]. Washington, D.C.: U.S. Census Bureau.

Current Population Survey Outgoing Rotation Group microdata. 2013. Survey conducted by the Bureau of the Census for the Bureau of Labor Statistics [machine-readable microdata file]. Washington, D.C.: U.S. Census Bureau.

Dube, Arindrajit. 2013. Minimum Wages and the Distribution of Family Income. University of Massachusetts Amherst, unpublished working paper.

Giannarelli, Linda, Joyce Morton, and Laura Wheaton. 2007. Estimating the Anti-Poverty Effects of Changes in Taxes and Benefits with the TRIM3 Microsimulation Model. Urban Institute Technical Report.

Gould, Elise, Tanyell Cooke, and Will Kimball. 2015. What Families Need to Get By: EPI’s 2015 Family Budget Calculator. Economic Policy Institute, Issue Brief #403.

Gould, Elise, Alyssa Davis, and Will Kimball. 2015. Broad-Based Wage Growth Is a Key Tool in the Fight Against Poverty. Economic Policy Institute, Briefing Paper #339.

Greenhouse, Steven. 2014. “More Workers Are Claiming ‘Wage Theft.’” The New York Times, August 31.

Kuehn, Daniel. 2014. The Importance of Study Design in the Minimum Wage Debate. Economic Policy Institute, Issue Brief #384.

Marr, Chuck, Chye-Ching Huang, Arloc Sherman, and Brandon Debot. 2015. EITC and Child Tax Credit Promote Work, Reduce Poverty, and Support Children’s Development, Research Finds. Center on Budget and Policy Priorities.

Meixell, Brady, and Ross Eisenbrey. 2014. An Epidemic of Wage Theft Is Costing Workers Hundreds of Millions of Dollars a Year. Economic Policy Institute, Issue Brief #385.

Meyer, Bruce D., Wallace K.C. Mok, and James X. Sullivan. 2009. The Under-Reporting of Transfers in Household Surveys: Its Nature and Consequences. NBER Working Paper No. 15191.

Mishel, Lawrence, and Ross Eisenbrey. 2015. How to Raise Wages: Policies that Work and Policies that Don’t. Economic Policy Institute, Briefing Paper #391.

Office of Management and Budget (OMB). 2015a. The President’s Budget for Fiscal Year 2016.

Office of Management and Budget (OMB). 2015b. Table S-9 in Summary Tables from The President’s Budget for Fiscal Year 2016.

Reeves, Richard, and Joanna Venator. 2014. Are Obama and Ryan Proposals for an EITC Expansion Pro- or Anti-Mobility? Brookings Institute Social Mobility Memo.

Ruggles, Steven, Katie Genadek, Ronald Goeken, Josiah Grover, and Matthew Sobek. 2010. Integrated Public Use Microdata Series: Version 5.0 [machine-readable database]. Minneapolis: University of Minnesota.

Sawhill, Isabel, and Quentin Karpilow. 2014. Raising the Minimum Wage and Redesigning the EITC. Brookings Center on Children and Families.

Schmitt, John. 2013. Why Does the Minimum Wage Have No Discernible Effect on Employment? Center for Economic and Policy Research.

Shierholz, Heidi. 2013. Low Wages and Scant Benefits Leave Many In-Home Workers Unable to Make Ends Meet. Economic Policy Institute, Briefing Paper #369.

Shierholz, Heidi. 2014. Low Wages and Scant Benefits Mean Many Restaurant Workers Can’t Make Ends Meet. Economic Policy Institute, Briefing Paper #383.

Shierholz, Heidi, and David Cooper. 2014. “CBO Report Shows Low-Wage Workers Would Be Better Off With a Minimum Wage of $10.10.” Working Economics (Economic Policy Institute blog), February 20.

U.S. Department of Labor. 2014. “Fair Labor Standards Act Advisor: Exemptions.” Accessed September 16, 2014.

West, Rachel. 2015. The Murray-Scott Minimum-Wage Bill: A Win-Win for Working Families and Taxpayers. Center for American Progress.

West, Rachel, and Michael Reich. 2014a. The Effects of Minimum Wage on SNAP Enrollments and Expenditures. Center for American Progress.

West, Rachel, and Michael Reich. 2014b. A Win-Win for Working Families and State Budgets. Center for American Progress.

Wheaton, Laura. n.d. “Under-Reporting of Means-Tested Transfer Programs in the CPS and SIPP.” The Urban Institute.

Young, Katherine, Robin Rudowitz, Saman, Rouhani, and Rachel Garfield. 2015. Medicaid Per Enrollee Spending: Variation Across States. The Kaiser Family Foundation.

Zabin, Carol, Arindrajit Dube, and Ken Jacobs. 2004. The Hidden Public Costs of Low-Wage Jobs in California. Center for Labor Research and Education, U.C. Berkeley.

Summary

The Essential Worker Immigration Coalition (EWIC)—a lobbying group representing the interests of employers—claims that it is “concerned with the shortage of both semi-skilled and unskilled (‘essential worker’) labor” and thus “supports policies that facilitate the employment of essential workers by U.S. companies that are unable to find American workers.” Representatives of other influential corporate lobbying groups, including the U.S. Chamber of Commerce and ImmigrationWorks USA, have made similar claims.

These groups are advocating to deregulate and expand the H-2B temporary foreign worker program, which permits U.S. employers to temporarily hire workers from abroad with H-2B nonimmigrant visas for lower- and semi-skilled occupations that are non-agricultural and seasonal in nature. And, claiming that “many American businesses could not function without the H-2B program” the Chamber and ImmigrationWorks USA want Congress to create a new and much larger program that would permit them to hire lower- and semiskilled guestworkers for year-round jobs.

While Congress debates whether to expand existing temporary foreign worker programs and whether to create a larger new program, it should note the lack of credible evidence that there are labor shortages in lesser-skilled jobs. This report does not attempt to establish whether labor shortages exist in H-2B occupations, but instead looks at employment growth, wages, and unemployment rates in the main occupations of H-2B workers. Following are the main findings of the report.

1. Despite above-average employment growth in some of the top H-2B occupations, the fact that wages have been stagnant or declining, combined with persistently high unemployment rates, suggests that there are no labor shortages at the national level in the top H-2B occupations.

Specifically, looking at the top 15 H-2B occupations (the occupations with the largest numbers of certifications) in fiscal year 2014 and assessing how they changed from 2004 to 2014, we see that:

• There was no significant wage growth; in fact, wages were stagnant or declining for workers in all of the top 15 H-2B occupations.
• Seven of the top 15 occupations experienced employment growth that exceeded the overall growth of 5.5 percent for all occupations, two experienced employment growth that was less than the overall growth for all occupations, and six contracted.
• In the three fastest-growing occupations of Nonfarm Animal Caretakers (up 99.5 percent), Coaches and Scouts (up 72.3 percent), and Cooks in Restaurants (up 44.3 percent), wages declined over the same ten-year period.
• Unemployment rates increased in all but one of the top 15 H-2B occupations, and all 15 occupations had very high average unemployment rates in 2013–2014 (the most recent data available). In 11 of the top 15 H-2B occupations, unemployment dropped from 2004–2005 to 2006–2007, but then rose significantly between 2006–2007 and 2013–2014. Such high unemployment rates suggest a loose labor market in the top 15 H-2B occupations.

2. While a change in prevailing wage rules in the middle of 2013 (known as the 2013 Interim Final Rule or IFR) may have helped fuel a slight increase in the wages that H-2B jobs were certified at by the U.S. Department of Labor, H-2B wage rules continue to allow hourly wage rates that are far lower than state and national averages in the overwhelming majority of cases.

As a result, employers save multiple dollars an hour per worker by hiring a lower-paid H-2B worker instead of a U.S. worker earning the local average wage for the occupation. This suggests that despite the rule change, the wage rates employers are required to pay H-2B workers are not high enough to attract U.S. workers and thus not high enough to ensure compliance with the Immigration and Nationality Act’s statutory requirement that an H-2B worker not be hired unless “unemployed persons capable of performing such service or labor cannot be found in this country.” (The 2013 rule requires employers to pay the local average wage unless a collective bargaining agreement is applicable, or unless the U.S. Department of Labor approves the use of a wage survey that it did not conduct.)

Specifically, the report calculates the difference between the H-2B wage certified by the U.S. Department of Labor and the average national or state wage from DOL’s Occupational Employment Statistics (OES) wage survey data. In most cases this difference represents how much employers can save on their wage bill, on average, by hiring an H-2B worker instead of a U.S. worker earning the local average wage for the occupation. This report finds that for the top 15 H-2B occupations in fiscal years 2012, 2013, and 2014:

• Except for six instances out of 45, nationwide, on average, H-2Bs were certified at a wage that was below the national OES average wage.
• In the top H-2B occupation, of Landscaping and Groundskeeping Workers, employers saved on average between $2.59 and $3.37 per hour by hiring an H-2B worker instead of a worker earning the national average wage for the occupation.
• In the second-largest H-2B occupation, Forest and Conservation Workers, employers saved on average between $3.27 and $3.80 per hour by hiring an H-2B worker instead of a worker earning the national average wage for the occupation.
• Comparing the average hourly certified H-2B wage for an occupation in a particular state with the average hourly wage for the occupation in the state yields similar results, with the share of all top 15 H-2B certifications that were in occupations where the average certified H-2B wage exceeded the state OES average rising but still low: from 1.2 percent in fiscal 2012 to 3.5 percent in fiscal 2013, to 9 percent in fiscal 2014.

3. Soon after the 2013 Interim Final Rule came into effect, H-2B employers shifted en masse to the use of private wage surveys—and evidence revealed in federal litigation clearly suggests that the shift to the use of private wage surveys was a systematic response by H-2B employers to keep H-2B wages lower than the average OES wage rate that would otherwise be required under the 2013 Interim Final Rule.

Specifically, in the nine months between July 1, 2013, and March 31, 2014, employers increased their submissions of private wage surveys for H-2B prevailing wage determinations by 3,182 percent, as compared with the 12 months leading up to the federal court decision that invalidated the previous H-2B wage rule. In 21.1 percent of those determinations, the certified H-2B wage was lower than the Level 1 H-2B prevailing wage, which is the 17th percentile wage by occupation and local area (according to U.S. Department of Labor OES wage survey data), and 94.4 percent of the determinations were for a wage that was lower than the Level 2 (34th percentile) wage (whereas the Level 3 wage is generally considered the local average wage, or roughly the 50th percentile wage).

Introduction/background

The Essential Worker Immigration Coalition (EWIC)—a lobbying group representing the interests of employers—claims that it is “concerned with the shortage of both semiskilled and unskilled (“essential worker”) labor” and thus “supports policies that facilitate the employment of essential workers by U.S. companies that are unable to find American workers” (EWIC 2015). Representatives of other influential corporate lobbying groups, including the U.S. Chamber of Commerce and ImmigrationWorks USA, have made similar claims.

The main policy solution these groups advocate is the deregulation and expansion of the H-2B temporary foreign worker program (also often referred to as a “guestworker” program), which permits U.S. employers to temporarily hire workers from abroad with H-2B nonimmigrant visas for lower- and semi-skilled occupations that are non-agricultural and seasonal in nature.

In 2010, the U.S. Chamber and ImmigrationWorks USA published a joint report touting the economic benefits of the H-2B program. In the foreword to the report they claim that “many American businesses could not function without the H-2B program” (U.S. Chamber and ImmigrationWorks USA 2010). Yet the report cites a survey showing that H-2B employers believe “the annual cap of 66,000 H-2B visas is too low to meet business needs” and “the program is so complicated and difficult to apply for that it discourages many small businesses from using it” (U.S. Chamber and ImmigrationWorks USA 2010, 3). These employer groups have lobbied Congress to expand the H-2B program and have lobbied and litigated to prevent federal agencies from implementing additional regulations that would raise the minimum wages they must pay and provide more rights and protections to U.S. and foreign H-2B workers (Luban 2015). The employer groups have largely succeeded in these efforts, having blocked almost all of the Obama administration’s proposed H-2B reforms from being implemented from 2011 to April 2015.1

A second policy response advocated by the U.S. Chamber, ImmigrationWorks USA, and EWIC is the creation of a new and much larger temporary foreign worker program to fill year-round jobs in lower- and semi-skilled non-agricultural occupations. Unlike in the H-2B program, temporary foreign workers employed in the United States under the EWIC plan would be allowed to work year round instead of only seasonally, as well as to switch to another job with a different employer under an employer registration protocol. Guestworkers in this program also would eventually be allowed to apply for permanent residence if they met certain requirements. Under the EWIC plan, the new temporary foreign worker program would have an annual numerical limit that starts “at 400,000 a year to keep up with demand” (EWIC 2007).

On more than one occasion, a program resembling the EWIC proposal has been seriously considered by Congress. In May 2005, a bill authored by Senators Edward Kennedy and John McCain that would have comprehensively reformed the U.S. immigration system, included provisions that would have created a new temporary foreign worker program with an initial annual limit of 400,000 that could be adjusted upward after the first year in response to employer demand.2 More recently, in the context of spring 2013 Senate negotiations to craft comprehensive immigration reform legislation, representatives of the U.S. Chamber and the AFL-CIO (the nation’s largest trade union federation) agreed to a legislative proposal to create a new, year-round temporary foreign worker program for non-agricultural lower-skilled occupations (Parker and Greenhouse 2013).3 The agreed-upon framework became part of the Border Security, Economic Opportunity, and Immigration Modernization Act (also known by its bill number, S. 744), which passed the Senate on June 27, 2013, by a vote of 68 to 32 (Silverleib 2013). If S. 744 had become law, a new temporary foreign worker program known as the “W-1” visa program would have eventually permitted employers to hire a maximum of 200,000 additional guestworkers per year.

While Congress continues to debate whether to expand existing temporary foreign worker programs and whether to create a much larger new program for employers who claim there are labor shortages in lower- and semi-skilled occupations, there is no credible evidence that such labor shortages exist. In fact, little has been written regarding the current state of the most common H-2B occupations, in terms of employment levels, wages, and unemployment rates. Other than employer anecdotes, no credible data or labor market metrics have been presented by non-employer-affiliated groups or organizations—let alone by disinterested academics—proving the existence of labor shortages that could justify a large expansion of non-agricultural lower- and semi-skilled temporary foreign worker programs. This report collects and assesses the available evidence on employment, wages, and unemployment rates in the top 15 certified H-2B occupations in fiscal 2014, for the 2004–2014 period. This report does not, however, attempt to conduct a detailed national, regional, or local labor shortage analysis or make a determination about the existence of shortages in particular H-2B occupations.

This report reviews the average wages certified nationwide in the top 15 H-2B occupations in fiscal years 2012, 2013, and 2014, and compares them with the average wages nationwide for each occupation and with the average wage in each state where an H-2B job was certified. The report also explains the different prevailing wage rules in place during the different fiscal years, and explores how they may have affected the results. Specifically, the report calculates the difference between the H-2B wage certified by the U.S. Department of Labor and the average national or state wage, and this gap in most cases represents how much employers can save on their wage bill, on average, by hiring an H-2B worker instead of a U.S. worker earning the local average wage for the occupation. The impact of employer-submitted private wage surveys (i.e., wage surveys that were not conducted by the U.S. Department of Labor) on determining H-2B prevailing wages is also assessed.

Data and methodology

Data presented on H-2B occupations are from the labor certification data sets provided annually and made publicly available by the Office of Foreign Labor Certification (OFLC), U.S. Department of Labor (DOL) (OFLC 2015). Filing for a labor certification is the first step in the process that an employer must complete if he or she wishes to hire an H-2B worker. Basically, the employer identifies a job he or she wants to fill with an H-2B visa holder and specifies the wage that the future H-2B visa beneficiary will earn if hired; the specified wage is also the wage the employer must promise to pay in advertisements targeting U.S. workers before the employer may hire an H-2B worker from abroad. The DOL reviews the request to make sure that the wage matches the correct wage rate for the job in its own database, unless a wage is set for the job through a collective bargaining agreement (in that case, the wage specified in the agreement is the H-2B wage), or is otherwise justified by a private survey submitted by the employer and approved by DOL.4

In the available H-2B data, requests for labor certification from employers are either labeled as certified, partially certified (meaning some fraction of the total number of the workers requested is certified, while the rest are denied), or denied (entirely) by DOL. Some records in the data may also be labeled as “certification expired” or “partial certification expired” (however not every year of DOL data on H-2B includes records labeled as certification expired or partial certification expired). H-2B labor requests that are certified may then be sent by employers to the United States Citizenship and Immigration Services (USCIS), which is part of the Department of Homeland Security (DHS), for review. USCIS may approve or deny the H-2B petition submitted by the employer. H-2B petitions that are approved by USCIS are then forwarded to the U.S. Department of State (DOS), which determines whether or not to issue an H-2B visa to an individual foreign worker after a consular interview, unless the H-2B worker is already in the United States in another immigration status (in that case, USCIS can adjust the worker’s status to H-2B).

These requests populate the DOL database, which is able to provide data on the pledged wages attached to each request for labor certification—the hourly wage rates contained in the requests that were certified but not yet approved by USCIS are referred to in this report as the certified H-2B wage. It is important to note that the number of certified H-2B labor requests is not the same as the final number of H-2B visas issued in a fiscal year (which are tallied by DOS) (Bureau of Consular Affairs 2015) or the number of H-2B “admissions,” i.e., persons crossing through a port of entry into the United States with a valid H-2B visa (which are tallied by DHS) (Monger 2013).

Because some of the requests for labor that DOL certifies do not result in an H-2B visa for a worker, the number of labor certifications is always higher than the number of petitions approved by USCIS and the eventual number of visas issued by DOS. For example, in fiscal 2013, DOL certified 82,307 requests for H-2B workers, of those USCIS approved 79,219 workers, and DOS issued 57,600 H-2B visas. (U.S. law also specifies that the number of H-2B visas issued in a fiscal year may not exceed 66,0005—this is commonly referred to as the annual H-2B “cap”—however certain exemptions6 may allow the total number to be higher than 66,000 in a given fiscal year.) This is a limitation inherent in H-2B data because DOS does not publish any data on the employers who received H-2B visas, the occupations of the H-2B workers, or the wages employers have promised to pay the H-2B workers issued visas by DOS. H-2B data from approved USCIS H-2B petitions that are disaggregated by Standard Occupational Classification code (SOC) and listing wages promised to be paid to H-2B workers would not be as reliable as DOS visa data, but would be preferable to DOL labor certification data because they would reveal a narrower subset of employers likely to eventually get an H-2B visa. Unfortunately, USCIS does not publish those data either. USCIS only publishes the total number of H-2B petitions it has approved, along with other limited information found in a report to Congress that it is required by law to publish every year (USCIS 2015). The total number of H-2B petitions USCIS approved for fiscal years 2009 to 2013 was also published in a March 2015 report by the United States Government Accountability Office (GAO 2015, 21).

The same GAO report criticized USCIS’s data collection, noting that while DOL uses modern SOC codes for classifying H-2B occupations, USCIS then converts those 840 occupations used by DOL to a different classification system that has a much smaller and broader subset of 83 detailed occupations (GAO 2015, 22). Data on wages to be paid by H-2B employers are collected by USCIS on the Form I-129 petition, but because they are not stored in electronic format (the data are stored in paper files), they are inaccessible to the government’s own auditors, or to the public, even with the use of a Freedom of Information Act request. As a result, DOL labor certification data are the most accurate and reliable public source of information for examining the occupations, work locations, and wages of H-2B workers.

The H-2B data used in this report exclude all H-2B labor certification records labeled as “denied,” “certification expired,” and “partial certification expired,” because denied and expired records were not likely to become approved USCIS H-2B petitions or visas issued by the State Department. As a result, there may be a discrepancy between the total number of H-2B certifications listed in this report for fiscal 2012, 2013, and 2014 and the total number of certifications found in the raw data sets published by DOL online, as well as on DOL’s annual “Selected Statistics” fact sheets,7 which report all certification records, including those listed as expired. This may also result in lower totals by occupation in this report than are reported by DOL in the Selected Statistics fact sheets.

The data on the wages of all workers in an occupation for the entire United States or by state comes from the Bureau of Labor Statistics’ Occupational Employment Statistics (OES) survey data (BLS 2015). These data are appropriate for comparing with certified average H-2B wages because most certified H-2B wages are based on OES wage data reported in the Online Wage Library at the Foreign Labor Certification Data Center, and these OES wage data correspond to the SOC codes for the jobs. The wages reported in the Online Wage Library are based on OES survey data for nearly every occupation (by SOC code) and region in the United States, therefore comparing certified H-2B wages with OES wages is comparing apples to apples in the vast majority of cases.8

When calculating occupational unemployment rates, the SOC codes do not correspond perfectly with the occupational codes used in the Current Population Survey (CPS) monthly household survey microdata, published jointly by the U.S. Census and the Bureau of Labor Statistics. CPS microdata are needed to calculate the unemployment rates for the top H-2B occupations. However, the H-2B occupations with the 15 largest numbers of certifications in fiscal 2014 match up reasonably well with the same or similar occupations found in the CPS data using the government’s crosswalks between the occupations, even though the occupational titles may differ slightly.

Top 15 H-2B occupations in FY2014

Table 1 shows the 15 occupations with the largest number of approved labor certifications for H-2B workers in fiscal 2014, for the entire United States. The top 15 occupations account for 67,978 labor certifications out of 83,843 total certifications, 81 percent of all labor certifications in fiscal 2014. The largest occupation by far is Landscaping and Groundskeeping Workers, with 34,159 workers certified, making up 40.7 percent of all H-2B labor certifications in fiscal 2014. The second largest occupation is Forest and Conservation Workers, with 6,753 certifications (8 percent of the total). Together, the top two H-2B occupations accounted for nearly half of all H-2B certifications (48.8 percent). The third and fourth largest H-2B occupations were Maids and Housekeeping Cleaners and Amusement and Recreation Attendants, each of which accounted for approximately 6 percent of all H-2B certifications. The fifth-largest H-2B occupation, Meat, Poultry, and Fish Cutters and Trimmers, made up 3.5 percent of all H-2B certifications and the sixth-largest, Construction Laborers, accounted for 2.9 percent. The rest of the top 15 H-2B occupations in fiscal 2014 are listed in Table 1.

Table 1

Lower-skill and low-wage occupations in the United States

According to the Organization for Economic Cooperation and Development (OECD), 25 percent of all jobs in the United States are “low-wage” jobs (defined as earning less than two-thirds of the national median hourly wage); the largest share of low-wage jobs among all OECD countries (Schmitt 2012, 1). (The United Kingdom has the next highest share at 21 percent, and Belgium has the lowest, with 4 percent.)

In 2013, 27.5 percent of all workers in the United States were earning “poverty-level” wages; this share has increased by 2.4 percentage points since 2000 (EPI 2013).9 From 2000 to 2012, the bottom 60 percent of wage earners in the United States on average experienced negative or zero real wage growth (Shierholz and Mishel 2013), and from 2000 to 2014, the median income for non-elderly households fell 12.3 percent (Mishel and Davis 2015). Even the most educated workers have not seen their wages rise since 2007; in fact, workers at all education levels have experienced stagnant or declining wages since 2007 (Gould 2015). Despite these dire statistics, low-wage occupations have been growing faster than middle-income or high-paying occupations since the end of the financial crisis of 2007–2009 (the “Great Recession”). According to the National Employment Law Project, low-wage occupations “constituted 21 percent of recession losses, but 58 percent of recovery growth” (NELP 2012). Many of the occupations that experienced high employment growth in 2014 (from BLS 2015a, Table 1.4), as well as the occupations projected to be the “fastest growing” between 2014 and 2024 (from BLS 2015b, Table 1.3) are in low-wage occupations that require little or no postsecondary education. While low-wage-earning workers in lower-skilled occupations in the United States are a significant share of the workforce and are in many cases employed in rapidly growing occupations, most have seen their wages stagnate or decline for many years.

Education/skills of workers in the top 15 H-2B occupations

Many of the top 15 H-2B occupations require minimal or no education and training. The Occupational Information Network (O*NET), which is developed and maintained by U.S. Department of Labor’s Employment and Training Administration (ETA), provides detailed background information on each of these occupations.10 O*NET assigns each occupation a “Job Zone” number which corresponds to “a group of occupations that are similar in: how much education people need to do the work; how much related experience people need to do the work; and how much on-the-job training people need to do the work.”11 The O*NET website has a brief descriptor of each Job Zone:

• Job Zone 1 – occupations that need little or no preparation
• Job Zone 2 – occupations that need some preparation
• Job Zone 3 – occupations that need medium preparation
• Job Zone 4 – occupations that need considerable preparation
• Job Zone 5 – occupations that need extensive preparation12

Of the 15 top H-2B occupations, nine are in O*NET Job Zone 1 (60 percent), four are in Zone 2 (26.6 percent), one is in Zone 3, and one is in Zone 4 (Table 2). According to O*NET, Job Zone 1 occupations “may require a high school diploma or GED certificate” and have employees who need “little or no previous work-related skill, knowledge, or experience” and “anywhere from a few days to a few months of training.” Job Zone 2 occupations “usually require a high school diploma” and have employees who usually need “[s]ome previous work-related skill, knowledge, or experience” and “anywhere from a few months to one year of working with experienced employees.” The O*NET survey data on the education levels of all workers who are employed in the United States in the top 15 H-2B occupations show that the vast majority either possess a high school diploma or its equivalent (but no additional education), or have less than a high school diploma. Therefore, because 13 of the top 15 H-2B occupations require little education and training, it can safely be said that they are lower-skilled occupations. The two possible exceptions are: Coaches and Scouts, in which over a quarter (28 percent) of workers reported possessing an associate’s degree, nearly half (47 percent) possess a bachelor’s degree, and 15 percent possess a master’s degree. The other is Nonfarm Animal Caretakers, in which 18 percent of workers reported possessing a bachelor’s degree. However, a worker may have a bachelor’s or master’s degree even though the job the worker performs requires little or no education. For example, many jobs in these two occupations, such as grooming horses at a stable (Nonfarm Animal Caretakers) or coaching a community college softball team (Coaches and Scouts), are unlikely to require a bachelor’s or master’s degree in any field.

Table 2

Employment change in the top 15 H-2B occupations, 2004–2014

Table 3 shows the U.S. employment levels in 2004 to 2014 for all occupations and for the top 15 H-2B occupations (by Standard Occupational Classification) in fiscal 2014. Table 4 shows the change in employment from 2004 to 2014 for all occupations in the United States, and for all workers in the United States who were employed in the occupations making up the top 15 H-2B occupations in FY 2014.

Table 3

Table 4

From 2004 to 2014, employment in all occupations grew 5.5 percent, averaging 0.5 percent per year. Over the same period, the top 15 H-2B occupations had widely varying rates of employment growth. Six experienced employment declines; seven experienced growth that was positive and above the 5.5 percent growth rate for all occupations; and two experienced growth that was lower than the percentage change for all occupations.

Amusement and Recreation Attendants, Coaches and Scouts, Waiters and Waitresses, Nonfarm Animal Caretakers, and Cooks, Restaurant, all grew by more than 10 percent from 2004 to 2014. Three occupations experienced employment growth far exceeding the overall growth of 5.5 percent for all occupations from 2004 to 2014: Nonfarm Animal Caretakers (99.5 percent), Coaches and Scouts (72.3 percent), and Cooks, Restaurant (44.3 percent). The three occupations with employment declines from 2004 to 2014 were Fishers and Related Fishing Workers (-57.4 percent); Forest and Conservation Workers (-24.8 percent), and Packers and Packagers, Hand (-20.5 percent). The top H-2B occupation in FY 2014, Landscaping and Groundskeeping Workers, grew by only 1.0 percent from 2004 to 2014 (an average of 0.1 percent per year).

Average hourly wages in top 15 H-2B occupations, 2004-2014

Table 5 shows the change in hourly wages for all workers in the United States and workers in the top 15 H-2B occupations from 2004 to 2014, adjusted to 2014 dollars. As the table shows, there was no significant wage growth for workers; wages were stagnant (growing less than 1 percent annually) or declined for workers in all of the top 15 H-2B occupations between 2004 and 2014.

Table 5

For all workers in the United States in all occupations, wages rose by just $0.40 in real terms (adjusted to 2014 dollars), just 1.8 percent over the decade. For workers in 10 of the top 15 H-2B occupations, wages declined, between $0.13 and $0.93 in 2014 dollars. The five occupations that saw slight hourly wage increases were Forest and Conservation Workers (by $0.04, or 0.3 percent), Maids and Housekeeping Cleaners ($0.02, 0.2 percent), Waiters and Waitresses ($0.80, 8.3 percent), Fishers and Related Fishing Workers ($0.82, 4.7 percent), and Dining Room and Cafeteria Attendants and Bartender Helpers, ($0.54, 5.8 percent). While workers in these occupations experienced real wage growth between 2004 and 2014, it was insignificant; wages in each of the five occupations grew by much less than 1 percent per year. Nationwide, workers in the other 10 top H-2B occupations were actually worse off in 2014 than they were 10 years earlier.

Unemployment rates in top 15 H-2B occupations, 2004–2014

Unemployment rates in H-2B occupations are calculated from Current Population Survey basic monthly microdata, which are jointly maintained by the U.S. Census and the Bureau of Labor Statistics. These data are not classified by SOC code, but instead use Census codes. However, the H-2B occupations with the 15 largest numbers of certifications in fiscal 2014 match up reasonably well with the same or similar occupations found in the CPS data using the government’s crosswalks between the occupations, even though the occupational titles may differ slightly in some cases.

Figure A shows the average unemployment rates in each of the occupations listed, for 2004–2005, 2006–­2007, and 2013–­2014. Two years of data were pooled together to increase sample sizes. The first two periods listed were chosen because they exclude the years of the recession that began in 2008 and its aftermath, and 2013–2014 was chosen because those years represented the most recent data available.

Figure A

Figure A shows that the unemployment rate rose in all but one of the top 15 H-2B occupations between 2004 and 2014. The average unemployment rate of Helpers-Production Workers was 10.8 percent during 2004–2005 and ended up at 10.3 percent during 2013–2014. Although the unemployment rate in this occupation has declined since 2004, it has declined by only one-half of a percent, and remained very high in 2014.

The average annual unemployment rate for all workers in the United States in 2014 was 6.2 percent. During 2013–2014, none of the 15 H-2B occupations was at or below the overall U.S. unemployment rate for 2014. Three occupations—Nonfarm Animal Caretakers; Athletes, Coaches, Umpires, and Related Workers; and Butchers and Other Meat, Poultry, and Fish Processing Workers, had an unemployment rate that was roughly about one percentage point higher than the national unemployment rate, while the other 12 occupations had much higher unemployment rates. Nine of the occupations had unemployment rates 10 percent or higher in 2013–2014, with the highest being Construction Laborers at 14.7 percent. Grounds Maintenance Workers, which corresponds to Landscaping and Groundskeeping Workers (SOC code 37-3011), the top H-2B occupation by far, had an average unemployment rate of 12.7 percent during 2013–2014, more than double the national unemployment rate.

Are there labor shortages at the national level among the top 15 H-2B occupations?

The claim by the Essential Worker Immigration Coalition that it is “concerned with the shortage of both semi-skilled and unskilled (‘essential worker’) labor” begs the question of whether there is any evidence that semiskilled and unskilled labor are in short supply. And since EWIC and other employer groups look to the H-2B visa as the remedy for labor shortages, it is reasonable to think about this in the context of the top H-2B occupations. However, it should be noted at the outset that determining whether a labor shortage exists in a particular occupation can be a difficult and inexact science. Numerous books and articles have been written on how to determine the existence of a shortage.13 While the definitions vary, the governments of various developed countries maintain and regularly update shortage occupation lists based on quantitative and qualitative labor market data and analysis.14 They rely on these analyses when crafting legislation and policy related to the workforce, the labor market, and immigration.

How to define and determine a labor shortage

A detailed explanation of how to conduct a labor shortage analysis is beyond the scope of this report. As mentioned, a substantial body of literature already exists on the subject, and in fact, this report will not conduct a detailed shortage analysis or make a determination about the existence of shortages in H-2B occupations. Exact methodologies for determining labor shortages may vary, and a discussion about their relative efficacy is also beyond the scope of this report. There are, however, basic definitions that many labor market economists can agree on. A good example is the simple definition offered by economists Philip Martin from the University of California, Davis, and Martin Ruhs from Oxford University, who have summarized the three essential elements required to establish a shortage (Martin and Ruhs 2011). Martin and Ruhs assert that industries and occupations reporting labor shortages should have (1) rising real wages relative to other occupations, (2) faster-than-average employment growth, and (3) relatively low and declining unemployment rates.15 The preceding sections of this report provide the available evidence relating to these three factors.

(1) Are real wages in top 15 H-2B occupations rising relative to other occupations?

Real wages are rising in only three of the top 15 H-2B occupations, and this rise is not significant. According to OES data, wages across all occupations stagnated in the United States between 2004 and 2014, rising only $0.40 in real terms (2014 dollars), 1.8 percent for all occupations. And as Table 5 shows, the story was mostly similar or worse in the top 15 H-2B occupations, where wages declined in real terms in 10 of the top 15 occupations. While wages increased in real terms in five of the top 15 occupations, the increases were insignificant: less than $1 in all cases, and less than a nickel in two of the occupations. In three occupations (Waiters and Waitresses, Fishers and Related Fishing Workers, and Dining Room and Cafeteria Attendants and Bartender Helpers) wages did rise faster than they did for all occupations, but rose at far less than even 1 percent per year.

(2) Is there faster-than-average employment growth in top 15 H-2B occupations?

There is faster-than-average employment growth in less than half of the occupations. As Tables 3 and 4 show, from 2004 to 2014, seven of the top 15 occupations experienced employment growth that exceeded the 5.5 percent increase for all occupations, six experienced declines; and two experienced growth lower than the rate for all occupations. Three occupations experienced employment growth far exceeding the overall growth of 5.5 percent for all occupations from 2004 to 2014: Nonfarm Animal Caretakers (99.5 percent), Coaches and Scouts (72.3 percent), and Cooks, Restaurant (44.3 percent); however, wages in these three fastest-growing occupations declined over the same 10-year period.

(3) Do the top H-2B occupations have relatively low and declining unemployment rates?

Rather than declining, unemployment rates increased in all but one of the top 15 H-2B occupations between 2004–2005 and 2013–2014, and all averaged very high unemployment rates in 2013–2014. The unemployment rate declined in one occupation, Helpers-Production Workers, which had an average unemployment rate of 10.8 percent during 2004–2005 and 10.3 percent during 2013–2014. Although the unemployment rate in this occupation declined by half a percentage point between 2004–2005 and 2013–2014, the unemployment rate remained very high in 2013–2014. In 11 of the top 15 H-2B occupations, the unemployment rate dropped from 2004–2005 to 2006–2007, but then rose significantly between 2006–2007 and 2013–2014.

Such high unemployment rates suggest a loose labor market—an oversupply of workers rather than an undersupply—in the top 15 H-2B occupations. The unemployment rates presented in this report may underestimate how many workers cannot find work in the occupation because the official rates do not count workers who are no longer actively seeking employment in the occupation (either because they found a job in another occupation or because they gave up looking for work in the occupation).

Summing up the results of the three tests

While seven of the top 15 occupations experienced employment growth that exceeded the 5.5 percent increase for all occupations, the fact that wages were stagnant or declined, combined with persistently high unemployment rates, makes it highly unlikely that labor shortages exist at the national level in any of the top H-2B occupations. This does not mean that no labor shortages exist anywhere in the United States in these occupations—it is entirely possible and even likely that shortages exist in some states or localities—but the high national unemployment rates in H-2B occupations suggest that even the employers experiencing a local labor shortage might find available U.S. workers if they recruited outside their city, region, or state, and if they offered more attractive wages and benefits (including transportation and housing).

H-2B wages compared with the wages of all workers in top 15 H-2B occupations and the significance of prevailing wage regulations

Two additional issues deserve to be explored in depth. Are the wage rates that employers are required to pay H-2B workers high enough to attract U.S. workers and comply with the Immigration and Nationality Act’s statutory requirement that an H-2B worker not be hired unless “unemployed persons capable of performing such service or labor cannot be found in this country”?16 And are the prevailing wage regulations promulgated by the U.S. Departments of Labor and Homeland Security adequate to prevent downward pressure on the wages of U.S. workers who are employed in the top H-2B occupations?

This section reviews the average wages certified nationwide in the top 15 H-2B occupations in fiscal years 2012, 2013, and 2014, and compares them with the average wages nationwide for each occupation, as well with the average wage in each state where an H-2B job was certified. The difference between the H-2B certified wage and the average national or state wage in most cases represents how much employers can save on their wage bill, on average, by hiring an H-2B worker instead of a U.S. worker.

Since 2010, the rules governing the legally defined “prevailing wage” that employers are required to pay H-2B workers (corresponding to occupation and local area) have been modified several times by the departments of Labor and Homeland Security, and have been the subject of litigation and congressional appropriations riders. As a result, different sets of prevailing wage regulations were in place during fiscal years 2012, 2013, and 2014. This section explains the different prevailing wage rules in the different fiscal years, and explores how they may have affected the results. Finally, the impact of employer-submitted private wage surveys (i.e., wage surveys that were not conducted by the U.S. Department of Labor) on determining H-2B prevailing wages is also assessed.

The 2008 H-2B prevailing wage rule

In 2008, the U.S. Department of Labor (DOL) issued a regulation establishing the methodology for determining the prevailing wage that employers would be required to pay their H-2B employees. DHS and DOL describe the 2008 wage methodology in the preamble to H-2B regulations jointly promulgated in 2015 (DHS and DOL 2015, 24148):

The 2008 rule provided that the prevailing wage would be the collective bargaining agreement (CBA) wage rate if the job opportunity was covered by an agreement negotiated at arms’ length between a union and the employer; the Occupational Employment Statistics (OES) wage rate if there was no CBA; a survey if an employer elected to provide an acceptable survey; or a wage rate under the Davis-Bacon Act (DBA), 40 U.S.C. 276a et seq., or the McNamara-O’Hara Service Contract Act (SCA), 41 U.S.C. 351 et seq., if one was available for the occupation in the area of intended employment. See 20 CFR 655.10 (2009). In the absence of the CBA wage, the employer could elect to use the applicable SCA or the DBA wage in lieu of the OES wage. See 20 CFR 655.10(b) (2009). The 2008 rule and the agency guidance implementing it required that when prevailing wage determinations were based on the OES wage survey, which is compiled by the Bureau of Labor Statistics (BLS), the wage had to be structured to contain four tiers to reflect skill and experience.⁶

Footnote 6: The 2008 rule required that when the prevailing wage was based on the OES, it should reflect skill levels. The agency’s implementing guidance required that the prevailing wage contain four wage tiers based on skill level. As a result, we refer throughout this rule to the 2008 rule’s requirement of four wage tiers.

Because the OES survey captures no information about actual skills or responsibilities of the workers whose wages are being reported, the four-tiered wage structure, adapted from the statutorily required four tiers applicable to the H-1B visa program under section 212(p)(4) of the INA, 8 U.S.C. 1182(p), was derived by mathematical formula as follows to reflect “entry level,” “qualified,” “experienced,” and “fully competent” workers: Level 1 is the mean of the lowest-paid 1/3, or approximately the 17th percentile; Level 2 is approximately the 34th percentile; Level 3 is approximately the 50th percentile; and Level 4 is the mean of the highest-paid 2/3, or approximately the 67th percentile.

In addition, wage methodology guidance published by the Employment and Training Administration (ETA) at DOL in 2009 provided “policy and procedural guidance” for private wage surveys submitted by employers for H-2B prevailing wage determinations (ETA 2009).

On January 19, 2011, the Obama administration issued a final rule to modify the 2008 H-2B wage methodology and 2009 wage guidance that would have set the H-2B prevailing wage “as the highest of the OES arithmetic mean wage for each occupational category in the area of intended employment; the applicable SCA/DBA wage rate; or the CBA wage” and “eliminated the use of employer-provided surveys as alternative wage sources, except in limited circumstances.” (DHS and DOL 2015, 24148) (The limited circumstances included in which the H-2B job was not represented in the DBA, SCA, or OES, and the surveys submitted had to meet the methodological standards in the 2008 rule.) However, due to legal challenges in federal court, the administration postponed implementation of the 2011 final wage rule (DOL 2011a), and was also prevented by Congress from using any funds to implement, administer, or enforce the January 2011 wage rule as a result of enacted appropriations legislation (DOL 2011b).17 As a result, the four-tiered 2008 wage methodology remained in effect, and the use of of private wage surveys under the terms of the 2008 wage rule and 2009 wage guidance continued to be permitted throughout all of fiscal 2012.

Fiscal 2012 H-2B wage data

Table 6 shows the top 15 H-2B occupations in fiscal 2012, and the nationwide average hourly wage for certified H-2B workers in each of the occupations. The 2012 OES average hourly wage for all workers in the occupation nationwide is listed next to the H-2B wage. In tables 6 through 11, the final column shows the difference between the average hourly certified H-2B wage and the average hourly OES wage (nationwide or by state); this is what employers save, on average, by hiring an H-2B worker instead of a worker who is paid the average (national or state) wage for the occupation. A negative value in the employer hourly wage savings column represents an H-2B job that was, on average, certified at a higher wage rate than the corresponding OES national or state average hourly wage.

Table 6

Table 6 shows that in each of the top 15 H-2B occupations in fiscal 2012, the average hourly wage certified nationwide for H-2B workers was lower than the OES average hourly wage for all workers in the occupation. The biggest wage savings for employers was found in the First-Line Supervisors/Managers of Housekeeping and Janitorial Workers occupation; employers could save nearly $9 per hour on average by hiring an H-2B worker instead of a worker earning the national average for the occupation. In the top two occupations of Landscaping and Groundskeeping Workers and Forest and Conservation Workers, the average hourly savings were over $3. If for example, an employer hired an H-2B landscaper to work for 40 hours per week for nine months (approximately 36 weeks) at $3 per hour less than the local average wage, the employer would save $4,320.

These findings are consistent with what DHS and DOL described in 2013. Employers were allowed to pay the Level 1, 17th percentile and Level 2, 34th percentile wage—both of which are below the local average wage for the job the H-2B worker would perform—and were in fact taking advantage of this wage rule in order to pay their H-2B workers wage rates that were well below average:

According to the distribution of the 59,694 H-2B prevailing wage determinations the Department of Labor issued based on the Occupational Employment Statistics (OES) wage survey in FY 2011 and 2012,¹⁶ 72.3 percent of H-2B prevailing wage determinations based on the OES were at Level I. The percentages of H-2B prevailing wage determinations based on the OES at Levels II, III, and IV were 14.4, 5.9, and 7.4, respectively. In over 90 percent of those cases, the H-2B prevailing wage was determined at the wage rate lower than the mean of the OES wage rates for the same occupation. [Emphasis added]

Footnote 16: In FY 2011 and 2012, a total of 72,037 prevailing wage determinations were issued by the Department of Labor’s National Prevailing Wage Center (NPWC) for employers seeking wage rates for H-2B workers. Of the 72,037, 59,694 determinations (82.9%) were based on the OES and 12,343 determinations were based on a collective bargaining agreement (CBA), the Davis-Bacon Act (DBA), or the Service Contract Act (SCA) prevailing wage, or employer-submitted wage surveys. (DHS and DOL 2013, 24057)

The results are similar when the average wages in each state and for each occupation for which data are available are compared with the average certified H-2B wage for the corresponding state and occupation. Table 7 (which can be found at the end of this report) shows that in the vast majority of cases, H-2B workers on average were certified to be paid lower wages than the state average. In 27 instances (representing a total of 675 workers), the average certified H-2B wage was higher than the state OES average wage for the occupation.

The 2013 H-2B prevailing wage rule, private wage surveys, and fiscal 2013 H-2B wage data

For approximately the first half of fiscal 2013, the 2008 H-2B wage methodology regulation and 2009 wage guidance remained in effect. But on April 24, 2013, the DHS and DOL issued a joint interim final rule (IFR) that was effective on the day it was published (DHS and DOL 2013), which eliminated the 2008 four-tiered wage methodology, and modified the regulation to require that:

If the job opportunity is not covered by a CBA, the prevailing wage for labor certification purposes shall be the arithmetic mean, except as provided in paragraph (b)(4) of this section, of the wages of workers similarly employed in the area of intended employment. The wage component of the BLS Occupational Employment Statistics Survey (OES) shall be used to determine the arithmetic mean, unless the employer provides a survey acceptable to OFLC under paragraph (f) of this section. (DHS and DOL 2013, 24061)

Although the April 24, 2013, IFR required employers to pay the “arithmetic mean,” meaning the average hourly wage (which is in most cases identical to the Level 3 wage), and no longer permitted employers to pay their H-2B workers the 17th (Level 1) or 34th (Level 2) percentile wages, the 2013 IFR continued to permit the use of private wage surveys submitted by employers to set prevailing wage levels under the terms of the 2008 wage rule and 2009 wage guidance; something that the final 2011 wage methodology regulation was much more restrictive about permitting (but which never became effective). In a 2014 case, the United States Court of Appeals for the Third Circuit considered the legality and continued use of private wage surveys, and noted that “DOL allowed this unlimited use of private surveys despite its 2011 findings that such surveys are unreliable and should only be used in extraordinary circumstances.”18 The wage methodology that employers were required to use from April 24, 2013, through the rest of the fiscal year (ending on September 30, 2013) was the wage methodology promulgated in the 2013 IFR (the arithmetic mean by occupation and local area) along with DOL-accepted private wage surveys under the terms of the 2008 wage rule and 2009 wage guidance.

Table 8 shows the top 15 H-2B occupations in fiscal 2013, and the nationwide average hourly wage for certified H-2B workers in each of the occupations. As with Table 6, the final column shows the difference between the average hourly certified H-2B wage and the OES average hourly wage for all workers in the occupation; in other words what employers save, on average, by hiring an H-2B worker who is paid the certified wage instead of a U.S. worker who is paid the average wage for the occupation.

Table 8

Table 8 shows that in 14 of the top 15 H-2B occupations in fiscal 2013, the average hourly wage certified for H-2B workers was lower than the OES nationwide average hourly wage for the occupation. The biggest savings was in the Construction Laborers occupation; employers could save $6.39 per hour on average by hiring an H-2B worker instead of a worker earning the national average for the occupation. In the top two occupations, the average hourly wage savings were again over $3. The only occupation where on average, the certified hourly H-2B wage was higher than the national OES average hourly wage was Cooks, Restaurant, where the average certified H-2B wage was $0.31 an hour higher.

The results are similar when comparing the average wages in each state and for each occupation for which data are available with the average certified H-2B wage for the corresponding state and occupation in fiscal 2013. Table 9 (which can be found at the end of this report) shows that in the vast majority of cases, H-2B workers on average were certified to be paid lower wages than the state average. In 34 instances representing a total of 2,403 workers, the average certified H-2B wage was higher than the state OES average wage for the occupation. While 2,403 H-2B certifications is still a very small share of all the labor certifications in the top 15 for fiscal 2013—accounting for only 3.5 percent of certifications in the top 15—it represents a larger share than in fiscal 2012, when only 1.2 percent of H-2B certifications in the top 15 were in occupations in a state where the average certified H-2B wage was higher than the state OES average wage for the occupation.

The 2013 H-2B prevailing wage rule and fiscal 2014 H-2B wage data

On October 1, 2013, at the beginning of fiscal 2014, the prevailing wage rule laid out in the 2013 DHS/DOL Interim Final Rule (IFR) had been effective for just over five months. Therefore, during the entirety of fiscal 2014, employers were required to follow the wage methodology in the April 24, 2013, IFR, along with DOL-accepted private wage surveys under the terms of the 2008 wage rule and 2009 wage guidance. This allows us to compare a full year of H-2B wage data under the 2013 IFR wage rule and private wage surveys with the OES average wages for the top 15 H-2B occupations.

The most obvious shift in the fiscal 2014 data displayed in Table 10 is that one-third of the top 15 H-2B occupations were on average, certified at an hourly wage that was higher than the national OES average hourly wage for the occupation, compared with only one occupation in fiscal 2013 and zero occupations in fiscal 2012. It is possible that the 2013 IFR requiring that employers pay the local OES average wage (unless a collective bargaining agreement existed for the job or if an alternative wage survey was accepted by DOL) may have raised average certified H-2B wages enough for this to occur. However, if employers were in fact paying the local average wage to their H-2B workers after implementation of the 2013 IFR, one could reasonably expect that the significant wage savings employers get by hiring an H-2B worker instead of a U.S. worker earning the local average wage would mostly disappear.

Table 10

But hourly wages for H-2B Landscaping and Groundskeeping workers—the largest H-2B occupation, accounting for over a third of all H-2B jobs certified in fiscal 2014—were on average certified at a much lower hourly wage than the national OES average hourly wage for the occupation: $2.59 less. That means employers still saved significantly on their wage bills by hiring H-2B landscapers instead of local workers earning the local average wage. Employers hiring H-2B workers for other jobs in the top six, such as in seafood processing (as part of the Meat, Poultry, and Fish Cutters and Trimmers occupation), construction (Construction Laborers), and working for traveling fairs and carnivals (Amusement and Recreation Attendants), also saw significant wage savings despite the 2013 IFR requiring that H-2B workers be paid the local average wage.

Again, because the 2013 IFR required that the local average wage be paid, it would have been reasonable to expect that average certified H-2B wages would have risen higher than the wages shown in Table 10. However, two factors related to the H-2B prevailing wage regulations are likely to have reduced the average wage levels for which H-2B jobs were certified at in fiscal 2014. One is the requirement that employers pay their H-2B workers the collectively bargained–for wage, if one exists, even if it is lower than the OES average wage. As the New York Times reported in September 2015, a federal investigation is ongoing into whether an employer-created-and-controlled union was purporting to represent workers while bargaining with employers to keep wages low for H-2B workers for traveling fairs and carnivals (Meier 2015).19

However, likely the main reason that wages certified for H-2B workers in fiscal 2013 and 2014 did not increase enough to achieve parity with the state and national average OES wages for all workers is that employers were allowed to continue submitting private wage surveys to determine H-2B prevailing wages under the terms of the 2008 wage rule and 2009 wage guidance. The CATA v. Perez decision in the Third Circuit noted in no uncertain terms that H-2B employers responded to the higher prevailing wage requirements in the 2013 IFR by substantially increasing the number of private wage surveys they submitted to DOL in order to keep certified H-2B wages low:

Congress has charged DOL with the duty to ensure that it grants certifications only if they do not adversely affect wages and working conditions of United States workers, and it is the burden of DOL to be mindful of and honor that charge. However, employers increasingly have been submitting private surveys authorized by Section 655.10(f) in order to obtain a wage rate that is lower than the OES wage rate indicates would be appropriate—the wage rate DOL itself has determined is necessary to avoid an adverse effect on foreign and domestic employee’s wages. The 2009 Wage Guidance therefore establishes criteria contrary to both the letter and spirit of 5 U.S.C. § 706(2)(A) and (C), and DOL’s use of it in the consideration of labor certification applications is unlawful.20

A media report from Bloomberg BNA (Francis 2014) on the CATA v. Perez decision highlighted the increase in the use of private wage surveys in response to the 2013 IFR:

In the 12 months leading up to the March 2013 CATA decision striking down the 2008 H-2B wage rule, employers seeking labor certification for H-2B visas submitted a total of 49 applications using private surveys to determine the prevailing wage, the court said. By contrast, employers submitted 1,559 applications using private surveys in the nine months between July 1, 2013, and March 31, 2014—a 3,182 percent increase.

According to the court, 21.1 percent of those prevailing wage determinations certified wages less than the average wage for the lowest skill level on the OES survey, and 94.4 percent of the determinations included wages lower than the OES’s Level II.

As the data revealed in the CATA v. Perez decision and reported by Bloomberg show, a significant number of employers began to request that DOL approve their submitted private wage surveys—by an increase of 3,182 percent soon after promulgation of the 2013 IFR—and in 21.1 percent of those determinations, the certified wage was lower than even the Level 1, 17th percentile wage for the position (by occupation and local area), and 94.4 percent of the determinations were for a wage that was lower than the Level 2, 34th percentile wage.

The aforementioned wage differential between the average OES wage and the average certified H-2B wage for Landscaping and Groundskeeping workers in fiscal 2014 is almost entirely explained by the new and increased use of private wage surveys by landscaping employers after promulgation of the 2013 IFR. Landscaping employers:

did not submit any employer wage surveys in the year prior to April 2013 despite being the industry employing the most H-2B employees. In the nine-month period from July 2013 to March 2014, 1,240 prevailing wage determinations for landscape workers (SOC Code 37-3011) were based on employer surveys, accounting for 42.7% of all the prevailing wage determinations made for that occupation during that period. DOL approved 97.7% of those surveys at wage rates below the OES Skill Level II wage rate.21

This is clear evidence that the shift to the use of private wage surveys was a systematic response by H-2B employers to keep H-2B wages lower than the local average OES wage rate that would otherwise be required under the 2013 IFR.

The results by state and occupation for fiscal 2014 are similar to the two previous years’ differences between the average certified H-2B wage and the average OES wage for the occupation in the state. Table 11 (which can be found at the end of this report) shows that in the vast majority of cases, again H-2B workers on average were certified to be paid lower wages than the state average. However, in fiscal 2014, the share of top 15 H-2B workers whose certified H-2B wage was higher than the state OES average wage for the occupation was greater than in fiscal 2012 or fiscal 2013. In fiscal 2014 there were 86 instances in which the H-2B average hourly wage was certified at a higher average hourly wage than the state OES wage for a particular occupation in a state. These instances represented 6,145 workers out of a total of 67,978 H-2B labor certifications in the top 15 occupations, or 9 percent.

The 2015 H-2B prevailing wage rule

On April 29, 2015, the Department of Homeland Security (DHS) and the Department of Labor (DOL) jointly issued another wage rule to establish the prevailing wage methodology for the H-2B program (DHS and DOL 2015). The April 29, 2015, H-2B Final Wage Rule became effective on the day it was published and superseded the April 24, 2013, Interim Final Rule (IFR). The April 29, 2015, rule is similar to the 2013 IFR in its requirement that employers pay their H-2B workers the OES average wage unless the job is covered by a collective bargaining agreement. It differs by prohibiting employers from choosing the Davis Bacon or Service Contract Act wage rates as a source for the prevailing wage unless the work performed by the H-2B worker is covered by a government contract. The April 29, 2015, wage rule also puts additional restrictions on the use of employer-provided wage surveys; wage data sources other than the OES will be considered for the purpose of establishing an H-2B prevailing wage only if the survey falls into one of the following categories:

(i) The survey was independently conducted and issued by a state, including any state agency, state college, or state university;

(ii) The survey is submitted for a geographic area where the OES does not collect data, or in a geographic area where the OES provides an arithmetic mean only at a national level for workers employed in the SOC;

(iii)(A) The job opportunity is not included within an occupational classification of the SOC system; or

(B) The job opportunity is within an occupational classification of the SOC system designated as an “all other” classification.

If the survey falls into one of these categories, then additional methodological requirements for the survey are listed in the remaining subsections of 20 C.F.R. 655.10(f).

Because the 2015 wage rule was published relatively recently, it will be difficult to assess the impact of the rule in the same comparative manner used in this report until future years of H-2B data are published. The 2015 wage rule has some obvious flaws, however, which could lead to results similar to past years when H-2B wages were mostly certified at below-average wages. The 2015 rule continues to allow paying H-2B workers an hourly wage rate that is lower than the local OES average if a CBA applies, and the 2015 rule still permits the use of non-OES wage surveys. While 20 C.F.R. 655.10(f) restricts which non-OES surveys may be used to establish an H-2B prevailing wage, still permitted are surveys “conducted and issued by a state, including any state agency, state college, or state university.” Employers and employer groups might respond by requesting that state agencies and/or universities conduct new wage surveys in certain regions and occupations, and may even fund such surveys—and therefore perhaps exert undue influence on the results—since nothing in the H-2B regulations prohibits requesting that a wage survey be conducted by a public agency or a university and then privately funding it.

Amendments to H-2B rules in the Consolidated Appropriations Act of 2016

In addition, the 2015 wage rule was amended in Congress through a legislative rider included in the Consolidated Appropriations Act of 2016 (an omnibus bill to fund the government during fiscal 2016), which included a number of changes to the H-2B program (Lipinski 2015a; Siskind 2015; H-2B Workforce Coalition 2015; Mikulski 2015), and was signed into law in December 2015. The substance of the amendment that affects the H-2B wage rule is found in Section 112, and requires that:

The determination of prevailing wage for the purposes of the H-2B program shall be the greater of—(1) the actual wage level paid by the employer to other employees with similar experience and qualifications for such position in the same location; or (2) the prevailing wage level for the occupational classification of the position in the geographic area in which the H-2B nonimmigrant will be employed, based on the best information available at the time of filing the petition. In the determination of prevailing wage for the purposes of the H-2B program, the Secretary [of Labor] shall accept private wage surveys even in instances where Occupational Employment Statistics survey data are available unless the Secretary determines that the methodology and data in the provided survey are not statistically supported.22 [Emphasis added]

As a result, employers will likely be permitted to use private wage surveys in a much broader range of circumstances, and this may result in H-2B workers being paid wages that are below the OES local average wage for their jobs. As of the time of publishing this report, the DOL has only given a preliminary indication of how it will interpret, implement, and enforce some of the December 2015 amendments to the H-2B program in the Consolidated Appropriations Act of 2016 (DOL 2016). DOL has not, however, explained in detail how it will implement the provisions relating to private wage surveys. DOL has also not yet indicated whether it will publish new regulations to implement these changes, either as an interim final rule or as a regulation that is subject to notice and comment procedures under the Administrative Procedure Act. H-2B wage levels in fiscal 2016 will depend much on DOL’s interpretations and actions, and any possible litigation that results.

One of the other notable changes to the H-2B program in the Consolidated Appropriations Act of 2016 is found in Section 565, commonly referred to as the “returning worker exemption,” which exempts foreign workers who participated in the H-2B program in fiscal 2013, 2014, or 2015 from being counted under the program’s annual numerical limitation of 66,000. This could lead to a large increase in the number of H-2B workers in the United States; as a result of the returning worker exemption, the size of the H-2B program could as much as quadruple. However, previous years in which the returning worker exemption was the law of the land suggest the number of H-2B workers is more likely to double or triple, but ultimately will depend on employer demand. In fiscal 2007 for example, the last year the returning worker exemption was in place, nearly 130,000 H-2B visas were issued (Bruno 2015).

In addition, DOL has been prohibited in fiscal 2016 from using appropriated funds to enforce H-2B regulations that require “employers of H-2B workers to provide at least the same wages and other working conditions as they provide to H-2B workers to certain U.S. workers performing substantially the same work identified in the labor certification or performed by the H-2B workers,” or to enforce the rule requiring employers “to offer workers full-time employment for a total number of work hours equal to at least three-fourths of the workdays of each 12-week period (or 6-week period if the employment covered by the job order is less than 120 days)” (DOL 2016). While DOL cannot enforce these rules during fiscal 2016, the substantive rules remain in place even in fiscal 2016. The Consolidated Appropriations Act of 2016 also prevents DOL from using funds to audit H-2B applications or to conduct assisted or supervised recruitment (i.e., where DOL helps employers search for willing and available U.S. workers).

Most of the December 2015 amendments to the H-2B program, including the amendment to the 2015 wage rule, will remain in place for all of fiscal 2016. If no further amendments are made through standalone legislation or appropriations legislation to fund the U.S. government in fiscal 2017, the original 2015 wage rule (as promulgated in April 2015) would become effective again at the beginning of fiscal 2017 and the restrictions on using appropriated funds to enforce the rule would expire.

Conclusion

The evidence presented here—flat wages and persistent high unemployment rates in the top 15 H-2B occupations for the past decade—sheds doubt on claims that there are labor shortages in the top 15 H-2B occupations. Members of Congress and the federal agencies tasked with administering the H-2B temporary foreign worker program should take these data points into consideration when employer groups and other corporate representatives urge them to modify the H-2B wage methodology or rules relating to the recruitment of U.S. workers. One sensible policy response could be to reform the H-2B program so that the availability of work visas is tied to occupations and regions that are experiencing proven and documented labor shortages; but that would also require that an entity of the U.S. government be tasked with assessing and declaring labor shortages, and perhaps publishing and continually updating shortage occupation lists, according to a selected methodology.

The increase in the share of H-2B wages that were on average certified at a wage higher than the state or national OES average wage from fiscal 2012 to fiscal 2014 suggests that the prevailing wage rules put in place in the middle of fiscal 2013 (by the 2013 DHS/DOL Interim Final Rule) may have put some limited upward pressure on H-2B wages. To recap, the 2013 IFR requires employers to pay the local average wage unless a collective bargaining agreement is applicable, or if the U.S. Department of Labor approves the use of a wage survey that it did not conduct. The intention of the 2013 Interim Final Rule was indeed to prohibit employers from continuing to pay their H-2B workers wages that were much lower than the average being paid to local U.S. workers in the same occupations. However, the use of private wage surveys by employers to determine H-2B prevailing wages in the 2013 IFR—which continued to use the 2008 wage rule and 2009 wage guidance on private wage surveys—has prevented certified H-2B wages from closing the gap with the average wages paid to other similarly situated workers in the United States.

The 2015 H-2B Final Wage Rule continues to permit wage surveys that were not conducted by the U.S. Department of Labor in some circumstances; and whether employers will continue to be allowed to underpay their H-2B employees vis-à-vis U.S. wage standards will depend on how the 2015 Final Wage Rule is implemented by the U.S. Department of Labor, if and when it becomes effective. In fiscal 2016, it is all but certain that the likely increased use of private wage surveys to set H-2B wages as a result of the December 2015 appropriations riders will lower the wages paid to H-2B worker to levels far below the local averages paid to similarly situated U.S. workers.

In addition to the December 2015 appropriations riders which became law, multiple legislative proposals were introduced in Congress in late 2015 (Lipinski 2015b) that would reform the H-2B program. Specifically the proposals would permanently reinstitute the use of private wage surveys and the four prevailing wage skill levels for determining H-2B wages that were introduced in the 2008 wage rule and 2009 wage guidance. If such legislative proposals or additional appropriations riders were enacted, they would likely ensure that the temporary foreign workers employed through the H-2B program will continue to be underpaid for the foreseeable future, which would continue to put downward pressure on the wages of similarly situated U.S. workers employed in the top H-2B occupations.

Acknowledgements

The author is grateful to Arthur Read, David Griffith, Ross Eisenbrey, and Meredith Stewart for the insightful comments and observations they provided during the drafting of this document. The author is also grateful for the valuable research assistance provided by Will Kimball and Tanyell Cooke. However, the author is solely responsible for any errors or omissions.

Endnotes

1. The 2015 wage rule (along with numerous other H-2B reforms) was published by DHS and DOL in April 2015 and was able to come into effect (was not postponed by litigation or appropriations legislation). However, in December 2015, Congress made changes to these rules through riders to omnibus appropriations legislation, as discussed at the end of this paper.

2. See Title III—Essential Worker Visa Program, Secure America and Orderly Immigration Act, S.1033, 109th Cong. (2005–2006).

3. For a detailed listing of the elements of the W-1 visa program in S. 744, see Costa (2013).

4. The employer must also demonstrate certain steps have been met; for example that the job was advertised to U.S. workers at the same wage in the labor certification. For more background about the labor certification process, see DOL’s website on Foreign Labor Certification, “H-2B Certification for Temporary Non-Agricultural Work.”

5. INA § 214(g)(1)(B); 8 U.S.C. § 1184(g)(1)(B)

6. H-2B workers who have their H-2B visas extended for longer than the original validity period will not be counted again against the H-2B annual cap. Also, fish roe processors, fish roe technicians, and supervisors of fish roe processing are exempt from the annual cap (see Pub. L. No. 108-287, § 14006, 118 Stat. 951, 1014 (2004)), and from November 28, 2009, until December 31, 2019, workers performing temporary labor or services in the Commonwealth of the Northern Mariana Islands (CNMI) or Guam are also exempt from the annual cap (see 48 U.S.C. § 1806(a)(2) as amended by sec. 10 of Pub. L. 113-235; 48 U.S.C. § 1806(b)).

7. See e.g., OFLC (2015b).

8. The one notable exception where OES data may be lacking is in the second-largest H-2B occupation, Forest and Conservation Workers, (SOC code 45-4011). In recent New Jersey District Court litigation in Comité de Apoyo a Los Trabajadores Agricolas (CATA) v. Perez case (1:15-cv-04014-RBK-JS Doc. 21-3, filed July 15, 2015, and see also Docs. 1-2 and 21-2), the plaintiffs argue that the Bureau of Labor Statistics did not sample critical industries employing H-2B workers in SOC code 45‐4011, and therefore, the OES wages reported do not calculate a valid average wage rate for workers in 11 states in which prevailing wage determinations are made for workers employed in SOC 45-4011. Specifically, plaintiffs note that excluded from the OES survey “are almost all of the industries that hire 96% of H-2B forestry workers, including: North American Industry Classification System (NAICS) 1131 (Timber Tract Operations); NAICS 1132 (Forest Nurseries and Gather of Forest Products); NAIS 114 (Fishing, Hunting and Trapping); and NAICS 1154 (Support activities for Forestry). The exclusion of the industries that employ most forestry workers calls into question the validity of the OES forestry wage even in those states where one is reported.” Plaintiffs argue that Service Contract Act (SCA) wage rates are more appropriate for setting H-2B wage rates in forestry jobs because “The SCA survey…does not exclude the industries that hire H-2B forestry workers and, as a result, reports wages specific to H-2B forestry jobs in virtually all areas.” (CATA) v. Perez (Doc. 21-3, at 34).

9. The poverty-level wage in the figure cited is calculated using an estimate of the four-person weighted average poverty threshold in 2011 of $23,010 (based on the 2010 threshold updated for inflation). This is divided by 2,080 hours to obtain a poverty-level wage of $11.06 in 2011. The poverty-level wage is roughly equal to two-thirds of the median hourly wage. This figure is deflated by CPI-U-RS (Consumer Price Index Research Series Using Current Methods) to obtain the poverty-level wage levels for other years. The threshold is available at the U.S. Census Bureau website.

10. According to its website, O*NET “is the nation’s primary source of occupational information…containing information on hundreds of standardized and occupation-specific descriptors. The database, which is available to the public at no cost, is continually updated by surveying a broad range of workers from each occupation.” See Occupational Information Network (O*NET) website, “About O*NET.”

11. Occupational Information Network (O*NET) website, “O*NET OnLine Help, Job Zones,”

12. Id.

13. See e.g., Barnow, Trutko, and Piatak (2013) and Downs (2009)

14. See e.g., the Australian Government’s Skilled Occupations List, and the UK government’s Tier 2 Shortage Occupation List.

15. It must be noted that the labor market data metrics used for determining whether a labor shortage exists can also be usefully supplemented with additional evidence. For example, when assessing whether a labor shortage exists, the United Kingdom’s Migration Advisory Committee (MAC) also interviews employers, workers, labor unions, and other stakeholders to get a more complete picture than can be gathered purely from national-level labor market statistics. (The MAC refers to these surveys as “bottom-up” evidence.) These qualitative data can be useful and round out the analysis when labor market data do not provide a clear-enough picture on their own (Martin and Ruhs 2011). Nevertheless, if unemployment rates in an occupation are exceptionally high and wages do not rise for a prolonged period of time, those two factors are strong evidence that a labor shortage does not exist.

16. Immigration and Nationality Act, § 101(a)(15)(H)(ii)(b) [8 U.S.C. §1101(a)(15)(H)(ii)(b)].

17. Multiple appropriations bills that were enacted continued to prevent DOL from using funds to enforce the 2011 wage rule; see Consolidated and Further Continuing Appropriations Act, 2012, Pub. L. 112-55, 125 Stat. 552, Div. B, Title V § 546 (2011); Consolidated Appropriations Act, 2012, Pub. L. 112-74, 125 Stat. 786, Div. F, Title I § 110 (2011); Continuing Appropriations Resolution, 2013, H.J. Res. 117, 112th Cong., 126 Stat. 1313 (2012); Consolidated and Further Continuing Appropriations Act, 2013, Pub. L. 113-6, 127 Stat 198, Div. F, Title 5 (2013).

18. Comité de Apoyo a Los Trabajadores Agricolas (CATA) v. Perez, No. 14-3557, Doc. No. 003111811504 (3d Cir. Dec. 5, 2015), at 15.

19. For further discussion, see also Costa (2015).

20. Comité de Apoyo a Los Trabajadores Agricolas (CATA) v. Perez, No. 14-3557, Doc. No. 003111811504 (3d Cir. Dec. 5, 2015), at 36–37.

21. Comité de Apoyo a Los Trabajadores Agricolas (CATA) v. Perez, No. 14-3557, Doc. No. 003111811504 (3d Cir. Dec. 5, 2015), at 26–27 (footnotes omitted).

22. Consolidated Appropriations Act of 2016, H.R. 2029, 114th Cong., § 112 (2015). Pub. L. No. 114-113.

Data sources

The tables and figure in this report draw from the following data sources

Bureau of Labor Statistics (BLS) Occupational Employment Statistics (OES). 2015. May 2014 National Occupational Employment and Wage Estimates.

Current Population Survey basic monthly microdata. Various years. Survey conducted by the Bureau of the Census for the Bureau of Labor Statistics [machine-readable microdata file]. Washington, D.C.: U.S. Census Bureau.

Office of Foreign Labor Certification (OFLC) (U.S. Department of Labor). 2015a. OFLC Performance Data.

O*NET OnLine. www.onetonline.org (accessed on July 8, 2015).

References

Barnow, Burt, John Trutko, and Jaclyn Schede Piatak. 2013. Occupation Labor Shortages: Concepts, Causes, Consequences, and Cures. W.E. Upjohn Institute for Employment Research.

Bruno, Andorra. 2015. The H-2B Visa and the Statutory Cap: In Brief. Congressional Research Service. December 11.

Bureau of Consular Affairs (U.S. Department of State). 2015. Nonimmigrant Visa Statistics.

Bureau of Labor Statistics (BLS) (U.S. Department of Labor). 2015a. Table 1.4: Occupations with the Most Job Growth, 2014 and projected 2024.

Bureau of Labor Statistics (BLS) (U.S. Department of Labor). 2015b. Table 1.3: Fastest Growing Occupations, 2014 and Projected 2024.

Bureau of Labor Statistics (BLS) Occupational Employment Statistics (OES). 2015. May 2014 National Occupational Employment and Wage Estimates.

Costa, Daniel. 2013. Future Flows and Worker Rights in S. 744: A Guide to How the Senate Immigration Bill Would Modify Current Law. Economic Policy Institute. November 12.

Costa, Daniel. 2015. H-2B Wage Rule Loophole Lets Employers Exploit Migrant Workers. Economic Policy Institute. September 10.

Department of Homeland Security (DHS) and Department of Labor (DOL). 2015. Wage Methodology for the Temporary Non-Agricultural Employment H–2B Program, Docket No. ETA–2013–0003], RIN 1205–AB69. 80 Fed. Reg. 24146. April 29.

Department of Homeland Security (DHS) and Department of Labor (DOL). 2013. Wage Methodology for the Temporary Non-Agricultural Employment H–2B Program, Part 2, RIN 1205–AB69, 78 Fed. Reg. 24047. April 24.

Department of Labor (DOL). 2011a. Wage Methodology for the Temporary Non-Agricultural Employment H–2B Program; Postponement of Effective Date, RIN 1205–AB61, 76 Fed. Reg. 59896. September 28.

Department of Labor (DOL). 2011b. Wage Methodology for the Temporary Non-Agricultural Employment H–2B Program; Delay of Effective Date, RIN 1205–AB61, 76 Fed. Reg. 73508. November 29.

Department of Labor (DOL) (Employment and Training Administration, Office of Foreign Labor Certification). 2016. Emergency Guidance: Implementation of 2016 DOL Appropriations Act. January 5.

Downs, Anna. 2009. Identifying Shortage Occupations in the UK. Economic and Labour Market Review 3(5): 23–29.

Economic Policy Institute (EPI). 2013. “Share of Workers Earning Poverty-level Wages, by Gender, 1973–2013.” The State of Working America.

Employment and Training Administration (ETA) (U.S. Department of Labor. 2009. Prevailing Wage Determination Policy Guidance, Nonagricultural Immigration Programs. Revised November 2009.

Essential Worker Immigration Coalition (EWIC). 2007. “Cut in Temporary Worker Program Weakens Immigration Reform.” May 24.

Essential Worker Immigration Coalition (EWIC). 2015. Website home page accessed December, 2015.

Francis, Laura. 2014. “DOL’s Use of Private Wage Surveys In H-2B Program Struck Down by 3rd Cir.” Bloomberg BNA. December 8.

Gould, Elise. 2015. Even the Most Educated Workers Have Declining Wages. Economic Policy Institute. February 20.

H-2B Workforce Coalition. 2015. “H-2B Workforce Coalition Applauds H-2B Provisions in End-of-Year Congressional Spending Legislation.” Website accessed December 16, 2015.

Lipinski, Jed. 2015a. “Foreign Guest Workers to Triple under New Federal Budget Bill.” The Times-Picayune. December 21.

Lipinski, Jed. 2015b. “Foreign Guest Worker Visas Targeted in Cassidy, Boustany Bills.” The Times-Picayune. November 11.

Luban, Rachel. 2015. “DOL Freezes Guestworker Program After Court Ruling.” In These Times. March 17.

Martin, Philip, and Martin Ruhs. 2011. Labor Shortages and U.S. Immigration Reform: Promises and Perils of an Independent Commission. International Migration Review, Volume 45, Issue 1 (174–187), Spring.

Meier, Barry. 2015. “Union Accused of Betraying Migrant Carnival Workers.” New York Times. September 1.

Mikulski, Sen. Barbara. 2015. “Mikulski Fights to Support Maryland’s Seafood Industry and Eastern Shore Jobs in FY16 Spending Bill.” Email from Matt Jorgenson on Behalf of Sen. Barbara Mikulski, Office of Senator Barbara Mikulski. On file with the author. December 16.

Mishel, Lawrence, and Alyssa Davis. 2015. Income Stagnation in 2014 Shows the Economy Is Not Working for Most Families. Economic Policy Institute. September 16.

Monger, Randall. 2013. Nonimmigrant Admissions to the United States: 2012, Annual Flow Report. Office of Immigration Statistics, Department of Homeland Security. August.

National Employment Law Project (NELP). 2012. The Low-Wage Recovery and Growing Inequality. August.

Parker, Ashley, and Steven Greenhouse. 2013. “Labor and Business Reach Deal on Immigration Issue.” New York Times. March 30.

Office of Foreign Labor Certification (OFLC) (U.S. Department of Labor). 2015a. OFLC Performance Data.

Office of Foreign Labor Certification (OFLC) (U.S. Department of Labor). 2015b. H-2B Temporary Non-Agricultural Labor Certification Program – Selected Statistics, FY 2014.

Schmitt, John. 2012. Low-Wage Lessons. Center for Economic and Policy Research. January.

Shierholz, Heidi, and Lawrence Mishel. 2013. A Decade of Flat Wages: The Key Barrier to Shared Prosperity and a Rising Middle Class. Economic Policy Institute. August 21.

Silverleib, Alan. 2013. “Senate Passes Sweeping Immigration Bill.” CNN. June 28.

Siskind, Greg. 2015. “Siskind Summary: Important Immigration Provisions Included in Omnibus Appropriations Bill.” Greg Siskind on Immigration Law and Policy (blog). December 22.

United States Citizenship and Immigration Services (USCIS), U.S. Department of Homeland Security. 2015. H-2B Nonagricultural Temporary Worker Visa and Status, Fiscal Year 2014 Semiannual Report to Congress, Part 2: October 1, 2013 – September 30, 2014. February 11.

United States Government Accountability Office (GAO). 2015. H-2A and H-2B Visa Programs: Increased Protections Needed for Foreign Workers. March.

U.S. Chamber of Commerce (U.S. Chamber) and ImmigrationWorks USA. 2010. The Economic Impact of H-2B Workers. October 28.

Tables 7, 9, and 11: State-by-state wage comparisons for H-2B occupations

Table 7

Table 9

Table 11

The latest attempt to assess the economic impact of RTW laws is a study from Deskins, Bowen, and Christiadi (2015) associated with the West Virginia University School of Business (the “WVU study” henceforth). In this latest study, the authors claim to identify the causal effect of RTW laws on employment growth rates by examining a panel of state-level data on employment from 1990 to 2013. In doing so, they claim that RTW laws lead to faster employment growth. While they would appear on first glance to have an impressive dataset to tackle an ambitious question, the WVU study is fraught with several problems, outlined here and described in more detail below:

1. The WVU study does not have sufficient variation in RTW status within states during the study period to support the strong causal claims it makes about RTW laws. Convincing analysis of employment trends (either levels or growth rates) should rely on identification from states that switch from RTW to non-RTW or from non-RTW to RTW in the panel data. Between 1990 and 2010, there is only one state that meaningfully changed its RTW status, which does not provide enough variation to make a determination on employment growth rates.
2. The WVU study authors mistakenly claim that Texas became RTW in 1993 and Utah became RTW in 1995. While there is no evidence why they claim that Utah became RTW in 1995 rather than 1955, in 1993 Texas did pass legislation that modified the RTW law that had been on the books since 1947. Correctly reclassifying those states as RTW during the entire study period greatly reduces both the magnitude and statistical significance of the posited relationship between employment growth and RTW status.
3. The WVU study fails to include state fixed effects in its analysis. State fixed effects are the industry standard in conducting analysis of this type because they account for characteristics that are particular to a given state and not controlled for in other variables in the model. Employing fixed effects, Jepsen et al. (2014) and Eren and Ozbeklik (2016) find no such causal relationship running from RTW status to improved economic outcomes across states.

On their own, those first three arguments are enough to allow serious researchers and conscientious policymakers to disregard the WVU study results. When the appropriate adjustments are made to its model specification, the relationship between RTW and employment growth disappears.

Even outside these fatal considerations, there remain other problems with the WVU analysis:

4. The WVU regression analysis fails to acknowledge the fact that state data are highly correlated from one year to the next, so even in its multivariate regression model, it is likely that the growth rate (conditional on the controls) is correlated over time, as is the policy “treatment” variable (RTW status). If one erroneously treats state-year observations as fully independent of one another, then one will very likely underestimate standard errors and may overstate the statistical significance of any regression results.
5. The WVU study appears to use some incorrectly measured data and improperly employs others. For instance, the authors suggest their variable of interest is total employment, but they appear to have limited their data to private-sector employment. However, when we pull state-level employment data from the same source they employ, we find different employment levels than they report. This calls into question their proper collection of reliable data on other variables.

There is insufficient variation in RTW status to assess growth rate differences

The WVU study authors claim to be able to identify the causal effect of RTW laws on employment growth rates by examining a panel of state-level data on employment from 1990 to 2013. The first thing to note about this claim is how different it is from most other assessments of the effect of RTW legislation. Most previous assessments have posited a relationship between levels of employment (or wages) and RTW laws, not growth rates.

For example, the most detailed studies on the correlation of RTW laws and wage levels across U.S. states are Gould and Shierholz (2011) and Gould and Kimball (2015), who find that wage levels in RTW states are roughly 3 percent lower than in other states even after controlling for a comprehensive range of wage determinants besides RTW status. As we will highlight below, the nonstandard use of growth rates instead of levels in the WVU study becomes particularly problematic upon realizing that all of the positive effects of RTW status in the WVU dataset come from states that adopted RTW laws decades before the dataset begins.

Such effects seem hard to credit. States either are or are not RTW, and they can only choose to adopt RTW status once. Whatever economic mechanism links RTW status with employment should really be a one-time, discrete shock. If RTW status, for example, leads to lower wage levels because the hamstringing of union power reduces workers’ bargaining power, then perhaps the lower wage level should lead some employers to migrate to RTW states. But this employment migration to lower-wage states should in turn lead only to a one-time, level shift in employment (as opposed to a change in the rate of growth). Granted, as states move to the new wage or employment levels resulting from RTW status, there could be a temporary change in relative growth rates relative to non-RTW states. But, as we point out below in more detail, the timespan over which the WVU study implicitly posits that these adjustments would have to take place is just enormously, even implausibly, long.

Finally, besides the issues raised by using growth rates rather than levels as the dependent variable, a number of statistical issues remain. In theory, the long panel dataset assembled by the WVU authors (48 states over 20 years) could provide enough observations to enable economic relationships to be well-estimated. However, their data set does not actually solve the most vexing problem in trying to assess the effect of RTW status on employment growth: the relatively small number of independent observations on states’ pre- and post-RTW performance.

Contrary to study claims, Texas and Utah did not switch RTW status during the study period

In their sample, the WVU study authors posit that Utah, Oklahoma, and Texas switched from non-RTW to RTW status between 1990 and 2010. It should be noted that Texas merely “modified” its already existing RTW law in 1993 (Collins 2014). The original legislation was enacted in 1947; only minor changes were made in 1993, which did not expand RTW during the period in question.1 Thus, it’s inappropriate to label Texas as a “switcher” state in the study period because it became RTW in 1947. In fact, the 1993 act was titled “non-substantive changes to the Texas Labor Code.” And then there is the question of Utah. It is unclear why the WVU study lists Utah’s RTW legislation as occurring in 1995, when the National Right to Work Legal Defense Foundation, among other sources, clearly lists the enactment of Utah’s RTW law as 1955 (NRTW).

Before moving on to the multivariate regressions, the WVU study illustrates employment growth in switching states, including those outside the scope of its regression analysis. Bartik (2016) calls into question the WVU study on the grounds that 10 states that switched to RTW saw no improvement in employment growth in subsequent years. In an effort to better understand the WVU authors’ data and results, we replicate the figures in question for the three states that they claimed to have switched in their period of analysis, using the Current Employment Statistics and Federal Reserve Economic Data.

As shown in Figure A, Texas had only trivially faster employment growth (well under half a percentage point more rapid growth) in the 20 years after the authors posited that RTW status was passed. Figure B shows that Utah saw average growth of 1.4 percentage points lower following the supposed RTW switch. The actual remaining switcher, Oklahoma (Figure C), saw employment growth in the 10 years after RTW status was passed that was nearly 1.8 percentage points slower than what prevailed in the 10 years before. We acknowledge that there are a number of reasons why employment growth would be lower in the latter years in Oklahoma. For instance, it includes the years of the Great Recession and this is why multivariate regression analysis is important. However, this naive look at the employment changes in the one state that actually switched its RTW status calls into question the WVU study results.

Figure A

Figure B

Figure C

What this means is that the WVU study’s results are driven by a number of states that adopted RTW decades before the dataset even begins. For example, the vast majority of RTW states (18 of 25) adopted RTW status before 1960 (Collins 2014). The idea that adopting right-to-work status in the 1940s or 1950s strongly influenced employment growth in the 1990s and 2000s does not seem particularly persuasive given the huge range of other potential influences on employment growth across states in the second half of the 20th century.

Therefore, the WVU study claim that right-to-work status generates faster employment growth is totally based on the employment performance of states that have had right-to-work status for many decades. Yet, the adoption of right-to-work legislation would be expected, at best, to have a one-time effect occurring over a few years, not a decades-long impact.

The need for state fixed effects

The authors regress employment growth in a given state in one year against RTW status three years before. While this nonstandard regression framework essentially means that only a state’s RTW status before 2011 will be assessed, their regression results show a positive and statistically significant relationship between employment growth in a state/year observation and RTW status three years previous. While this relationship holds even when a number of controls are included, the authors are missing key state explanatory variables, notably state fixed effects.

Previous studies have demonstrated that RTW states have a number of characteristics besides their RTW status that are similar, making the potential problem of omitted variables driving the results particularly worrisome. In relatively long panel data a very common sensitivity test to probe the durability of regression results is the inclusion of state fixed effects—indicators that control for characteristics that are particular to a given state and not controlled for specifically in the remaining control variables.

A recent study at the University of Kentucky (Jepsen et al. 2014) examined economic development trends across states and found no impact of right-to-work status. The model used in this study used fixed effects “to capture unobserved differences in states that may affect economic growth. These effects include things such as a state’s climate, access to overseas markets or its citizens’ work ethic.”

Furthermore, Timothy Bartik (2016), a leading economic development expert at Upjohn Institute, recently reviewed the WVU study and concluded that it “does not provide any convincing evidence that a state that adopts RTW laws will as a result experience faster job growth.” Bartik reached this conclusion in part because the WVU study failed to include state fixed effects. Bartik also notes that “the most rigorous recent study that looks at Right-to-Work,” which focuses on Oklahoma, found no significant effect on employment growth (Eren and Ozbeklik 2016).

The WVU study authors argued against the inclusion of state fixed effects in their regression analysis on the grounds that because only three (arguably only one) states “switched” from non-RTW to RTW status between 1990 and 2010, that statistical identification in the fixed-effect regressions is based on only this very small number of “switcher” states. But this is the entire point of why their panel dataset is the wrong tool to use to assess the causal impact of RTW status on employment growth—despite the large number of state/year observations, it just doesn’t have enough pre- and post-RTW observations within states to make strong inferences.

To assess the effect of the correct classification of Texas and Utah and the inclusion of state-level fixed effects, we begin by replicating Table 3 of the WVU study to the best of our ability. We were able to match its results in Models 1 and 2, but did not have the additional variable to replicate Model 3 in its entirety. Our regression results are found in Table 1.2

The first and fourth columns replicate the WVU study results. When state fixed effects are included in the WVU study’s baseline regression as shown in the second column, the relationship between employment growth and (three-years-previous) RTW status gets smaller and becomes statistically insignificant.

This failure to find a statistically significant relationship between RTW and employment growth when a standard control such as state fixed effects is included is the most salient thing to know about the WVU study. State fixed effects are a completely standard robustness check for panel data and the WVU results fail this check, meaning one should be quite skeptical about their final results.

Table 1

In the third column, we go one step further and classify Texas and Utah correctly as RTW states during the entire period. The coefficient gets even smaller. The fifth column of Table 1 displays Model 2 with state fixed effects. Here, the magnitude of the result increases and remains statistically significant. However, once we correctly reclassify Texas and Utah, the economic and statistical significance disappears.

Uncorrected autocorrelation of errors overstates statistical significance

The WVU regression analysis fails to acknowledge the fact that state data are highly correlated from one year to the next so even in its multivariate regression model, it is likely that the growth rate (conditional on the controls) and the policy variable (RTW) is correlated over time. For example, if employment growth for one state in one year is above its conditional mean, then it’s likely that it will be above its mean in the next period. This suggests that there is serial correlation in the error terms. Traditionally when this is corrected, this increases standard errors. If standard errors go up, then confidence intervals get wider and statistical significance goes down.

The WVU study makes no mention of adjusting its standard errors for this serial correlation. Correcting standard errors in the presence of serial correlation is standard practice and likely would have increased standard errors. If one erroneously treats state-year observations as fully independent one another, then one will very likely underestimate standard errors and may overstate the statistical significance of any regression results.

We used a simple (and almost surely insufficient) correction to their Model 2 specification, adding in both state fixed effects and robust standard errors (accounting for the serial correlation).3 Those results are shown in the last column of Table 1. Even employing the model that wrongfully includes Texas and Utah as switchers, the WVU study results are nullified.

Other issues with the WVU data and methodology

The WVU study appears to use some incorrectly measured data and improperly employs others. For instance, the authors suggest their measured key variable of interest is total employment. But, the published Current Employment Statistics data on state-level employment has differed by 15–30 percent from their data.4 When we investigated further, it appears that they limited their data to private-sector employment. When we tried to match their employment levels using the Current Employment Statistics, we still found significant differences. We found their state employment levels vary anywhere from 0 percent to 7.8 percent from what is found in the CES. Given the lack of clarity and discrepancy in measurement, it calls into question the validity of other variables in their model.

Other studies that have attempted to identify correlations between RTW status and economic variables have tended to look at cross-sectional variation—sometimes in growth rates over relatively long (5- or 10-year) periods, rather than annually. We experimented with regressing state-level employment growth over the 1990–2000 and 2000–2007 periods on a number of controls. We also included some control variables that other studies have indicated may be correlated with employment growth across states, including the number of sunny days per year, the relative median rent in a state, and the starting level of employment in a state. In these regressions (available upon request) the RTW variable often became statistically insignificant depending on the precise constellation of other controls included.

Conclusion

In the United States, 25 of the 50 states had adopted RTW policies by 2015. This could conceivably give one enough data to assess the causal impact of RTW policies by looking at economic variables before and after RTW adoption and using controls such as state fixed effects. However, 18 of the RTW states had adopted these laws before 1960. This means that state-level data on economic outcomes (employment or wages, for example) would need to be found that extended quite far back in history. Until such data is assembled, it will be hard indeed to make strong claims about the causal impact of RTW status. But the latest WVU study does not acknowledge these difficulties and presents claims about the impact of RTW status that are just not supported by the data.

In short, while a naive assessment of the WVU dataset seems to indicate a large number of observations (48 states by 25 years), the small number of RTW “switchers” in the data (namely one) make it impossible to reliably identify the causal impact of RTW on employment growth. Our reasonable set of robustness test yields no relationship between RTW status and employment growth.

Endnotes

1. EPI analysis of Texas Labor Code, Title 3, Chapter 101, revisions effective Sept. 1, 1993. Also, see: http://www.lrl.state.tx.us/legis/billsearch/billdetails.cfm?billtypedetail=HB&billnumberdetail=752&legSession=73-0

2. Replication possible with dataset supplied by WVU study authors upon request, December 2015

3. The very limited extent of their panel data (specifically, the small number of states “switching” RTW status) does not allow the type of rigorous correction for the problem of serial correlation suggested in Bertand et al. (2004). In fact, the WVU study’s panel dataset is even too small to allow for the test Bertrand et al. (2004) submit as appropriate for a panel dataset with a “small” number of switchers.

4.  Email correspondence with WVU study authors, December 17, 2015.

Appendix: Measuring employment in states

—This report was amended January 19 to include the text of this appendix.

The WVU study’s most important findings are those regarding the impact of right-to-work status on state employment growth. However, we tried but could not replicate the study authors’ data with what is available from the Bureau of Labor Statistics website.

The WVU study’s authors describe their work by saying “we examine overall employment” (page 11). Their data appendix says they measure “Employment Growth” by “Year t to t+3 growth rate in total employment, by state.” They list their data source as “Current Employment Statistics Survey, US Bureau of Labor Statistics.”

We interpret “total employment” to be total nonfarm employment since that is the most aggregate category available from BLS. Appendix Table 1 provides a comparison of the WVU study’s data for 1990 and 2013, the beginning and end of the period of their study, with the data we developed from BLS for those years, both private sector and nonfarm (which should match their “total employment”). For the U.S. as a whole, total nonfarm employment in 1990 was 108.3 million, nearly 20 million more than the WVU study’s employment total of 89.8 million. This is why we also benchmarked their data to private-sector employment, which, for 1990, turns out to be close, just over 200,000 off. So, it appears that the WVU study may simply have mislabeled their results and relied on private-sector employment rather than total employment. However, the data comparison in the latest year, 2013, is not as close; there is a 1.5 million difference between the WVU study’s employment number (111.7 million) and our retrieval of BLS data on private-sector employment (113.2 million). BLS data on state employment comes rounded to the nearest hundred; however, the WVU study employment data was provided without rounding. We tried to replicate its data by averaging three years, which could produce unrounded values, but this didn’t explain the discrepancy.

What matters for this study is whether employment growth (not levels) is measured correctly. Simple tabulations of the growth of employment from 1990 to 2013 in each state using the WVU study’s data and the BLS private-sector employment data show that there are several large discrepancies. For instance, our downloaded BLS private-sector employment data show that West Virginia employment grew by 21.6 percent while the WVU study’s data suggest a 17.1 percent growth. There are also sizeable discrepancies of at least four percentage points in Arizona, California, and Idaho. The average discrepancy is only 1.1 percent but this reflects some states where employment growth is overstated by the WVU study and some states where employment growth is understated.

We conclude that at minimum their data are mislabeled as “total employment” when they should be labeled “private employment.”

Appendix Table 1

Appendix Table 2

References

Bartik, Timothy. 2016. “What do we know about right-to-work laws and state prosperity? Not much, because of limited variation over time in which states are RTW states.” Investing in Kids (blog). January 5.

Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan. 2004. “How Much Should We Trust Difference-in-Difference Estimators?” Quarterly Journal of Economics, 119(1), pp. 249–275

Bureau of Labor Statistics (U.S. Department of Labor) Current Employment Statistics Program. Various years. Regional and State Employment and Unemployment [database].

Collins, Benjamin. 2014. Right to Work Laws: Legislative Background and Empirical Research. Congressional Research Service report.

Deskins, John, Eric Bowen, and Christiadi. 2015. The Economic Impact of Right to Work Policy in West Virginia. Bureau of Business and Economic Research report.

Eren, Ozkan, and Serkan Ozbeklik. 2016. “What Do Right-to-Work Laws Do? A Case Study Analysis Using Synthetic Control Method.” Journal of Policy Analysis and Management, vol. 35, 173–194.

Federal Reserve Bank of St. Louis. Federal Reserve Economic Data (FRED) [database].

Gould, Elise, and Heidi Shierholz. 2011. The Compensation Penalty of “Right-to-Work” Laws. Economic Policy Institute, Briefing Paper No. 299.

Gould, Elise, and Will Kimball. 2015. “Right-to-Work” States Still Have Lower Wages. Economic Policy Institute, Briefing Paper No. 395.

Jepsen, Christopher; Sanford, Kenneth; and Troske, Kenneth R. 2008. Economic Growth in Kentucky: Why Does Kentucky Lag Behind the Rest of the South?. Center for Business and Economic Research, University of Kentucky.

National Right to Work Legal Defense Foundation, Inc. (NRTW) “Right to Work States: Utah.”

Introduction and key findings

Since the beginning of the 20th century, U.S. states have used minimum-wage laws to help ensure that regular employment provides the means to a decent quality of life. For decades following its enactment in 1938, the federal minimum wage provided this protection for workers across the United States. Yet in recent decades, the buying power of the federal minimum wage has eroded because policymakers have failed to raise it such that its value keeps pace with rising prices. As a result, a parent who works full time and is paid the federal minimum wage does not currently earn enough through work to be above the poverty line (Cooper 2015a). Faced with this inadequate federal standard, lawmakers in many states have adopted higher minimum wages to better reflect state-specific costs of living, and to help alleviate the wage inequality that has grown considerably over the past generation, in part as a result of eroding minimum wages (Mishel 2014).

In July 2015, an executive-appointed New York wage board created to make recommendations related to fast food workers’ pay recommended that the minimum wage for these workers be gradually raised to $15 per hour by 2018 in New York City and by 2021 throughout the rest of the state. In September, the acting state labor commissioner signed an order adopting the wage board’s recommendations for fast food workers. Shortly thereafter, the state’s governor, Andrew Cuomo, called for a legislative increase in the state minimum wage that would raise the wages of all workers in the state to at least $15 per hour. As a first step toward this goal, Gov. Cuomo announced in November his intent to raise the minimum pay of state employees to $15 an hour on a schedule similar to that of the fast food wage order.

This report analyzes the likely effects of a statewide increase in the minimum wage—in terms of the workers who would be affected and the resulting change in their pay—if the increase were implemented along a schedule similar to that of the fast food wage order. In particular, the report analyzes the affected workforce by age, gender, race and ethnicity, education levels, work hours, family status, household composition, and family income.

Key findings include:

• Raising the minimum wage to $15 per hour by 2018 in New York City and by mid-2021 throughout the rest of New York state would directly or indirectly lift wages for 3.2 million workers­­—about 37 percent of all workers in the state.
  • In New York City, the increase would raise the wages of 1.4 million workers, approximately 35 percent of the city’s workforce.
  • Outside of New York City, the increase would lift pay for 1.7 million workers, roughly 38 percent of wage earners elsewhere in the state.
• Over the phase-in period of the increases, affected workers would receive $15.3 billion in additional wages. Once the increase to $15 is reached, the average affected worker would earn roughly $4,800 more in annual pay than she does today (assuming no change in the number of work hours).
• The workers who would benefit from the higher minimum wage do not fit the stereotype of low-wage workers being teenagers from affluent families working part time.
  • A mere 5.2 percent of affected workers are teenagers. Nearly 95 percent are 20 years old or older, and more than three-quarters are 25 or older.
  • The majority of affected workers (52.7 percent) are women.
  • Statewide, roughly half of affected workers are persons of color. However, within New York City, more than three-quarters of affected workers are persons of color, and statewide, workers of color would benefit disproportionately from the increase. More than half of all Hispanic or Latino workers in the state would receive a raise, as would 40.5 percent of all black or African American workers.
  • Of workers who would receive a raise, two-thirds work full time, more than half (52.3 percent) have some college experience, and nearly a third (33.0 percent) have children.
  • Low-income households would benefit disproportionately from the increase. More than a third (37.1 percent) of affected workers come from families either in poverty or “near poverty,” defined as having income less than 200 percent of the poverty line. Over three-fourths of workers in or near poverty would get a raise.
  • The workers who would benefit earn, on average, half of their family’s total income. More than a quarter (27.0 percent) of affected workers are the sole providers of their family’s income.
• Three major industries have more than 400,000 affected workers each: retail trade, restaurants, and the group of human service workers and care providers in the home-based and residential care, social assistance, and child care sectors.

Background

On September 10, 2015, Gov. Andrew Cuomo proposed increasing the New York state minimum wage in several stages to $15 an hour, which would make New York the first state with a statewide $15 minimum wage. This followed a decision by a Cuomo-appointed wage board to recommend a $15 minimum wage for employees of fast food chains.

The wage board’s action was informed by testimony at hearings around the state in which scores of workers, advocates, and experts described the inadequacy of pay in the fast food industry and the damaging effects such low pay has on workers, families, and their communities. Workers described having to work multiple jobs to make ends meet, sacrificing time with family, and struggling to afford rent, food, health care, and other necessities (Karlin 2015). Low pay among fast food workers also forces many of these workers to rely on public assistance programs to supplement their inadequate earnings, with 60 percent of workers in the industry using some form of means-tested government assistance (Allegretto et al. 2013).

The Economic Policy Institute’s Family Budget Calculator (Gould, Cooke, and Kimball 2015) shows that in the least expensive area of New York—the Buffalo/Niagara Falls metro area—a single, childless adult working full time, year round, requires an hourly wage of at least $13.48 to achieve a modest but adequate standard of living. Using inflation projections from the New York State Budget Office, by 2021 workers in this region, and in all other areas of the state, will require an hourly wage of $15 or more to be able to afford a modest yet adequate living standard.1 In New York’s metropolitan areas, a single adult will need an hourly wage of $15.72 or more by 2021 to meet her basic needs. Parents who are raising children will need an hourly wage much greater than $15 in 2021, even if both parents are working to support one child.2 The significant gap between current minimum-wage levels and the wages required to achieve a modest yet adequate living standard reflects the extent to which lawmakers have let the minimum wage erode.

Under legislation enacted in 2013, New York’s minimum wage was raised above the current $7.25 federal minimum wage and reached $9.00 on December 31, 2015. At $9.00 an hour, the state minimum wage is still 20.3 percent below its inflation-adjusted value in 1970, when its value (in 2014 dollars) was $11.29 per hour (using national measures of inflation).3 However, adjusting for changes in nationwide prices (inflation) does not reflect the higher cost of living in New York compared with the national average.

To match the 1970 value of the minimum wage accounting both for national price changes and the higher cost of living in New York would require a New York minimum wage of $14.27 in 2016.4 Adjusting that level for projected inflation in New York, as forecast by the State Division of the Budget, would bring the state minimum wage to $15.01 per hour in 2018, effectively the same minimum-wage level and implementation year set by the wage board for fast food workers in New York City.

Another critical benchmark for considering the appropriateness of a $15 minimum wage, and the state economy’s capacity to support it, is the growth in average worker productivity since 1970. Over the past 40 years, average labor productivity—the value of goods and services produced from each hour of work—has grown steadily across the United States, yet real (inflation-adjusted) hourly pay for the vast majority has barely budged (Bivens et al. 2014). This is due, in part, to the falling real value of state and federal minimum wages (Mishel 2014). In New York, had the 1970 state minimum wage of $11.29 in 2014 dollars been raised at the same growth rate as average U.S. labor productivity, it would be $21.40 today. This is not to say that the state minimum wage should be this high, but it does indicate that had lawmakers taken a different path over the past four decades, it could have been this high given growth in workers’ average output per hour.

Critics of raising the minimum wage sometimes argue that measuring changes in the minimum wage against changes in average labor productivity is inappropriate because productivity among low-wage workers has not grown as much as average labor productivity. This criticism misses the point. Rising average labor productivity provides the means for improving living standards for all workers. To observe that improvements in productivity would allow for a minimum wage of over $21 today is to say that had the benefits of productivity improvements been shared more equally, rather than being concentrated among the highest-paid earners and wealthiest households, the lowest-paid workers in our economy could be making over $21 an hour.5 It is a result of policy decisions that they are paid far less.

Finally, in assessing the suitability of a proposed minimum wage, analysts will also often compare the proposed minimum to the median hourly wage of full-time workers, a ratio known as the Kaitz index. (See Cooper, Schmitt, and Mishel 2015 for more information on the Kaitz index and historical Kaitz values of the U.S. minimum wage.) Based upon the American Community Survey (ACS) data used in this study, a minimum wage of $15 in 2021 would equal 58 percent of the projected median wage of full-time workers in New York state, under the conservative assumption that wages at the median grow no faster than projected inflation.6 This would be slightly above the high point of the federal minimum wage in 1968, when it equaled roughly 55 percent of the national median wage of full-time workers (although this difference may not be meaningful, since these estimates are produced using different data sets). Moreover, individual states have had minimum-to-median wage ratios as high as 67 percent, using the earliest available state-level data (Zipperer and Evans 2014). Additionally, a half dozen developed countries have national minimum wages equal to at least 57 percent of their median wage (Cooper 2015b).

While Gov. Cuomo has not yet released a specific timetable for the proposed phased-in statewide minimum-wage increase to $15, the schedule will likely resemble the fast food minimum-wage increase shown in the left side of Table 1. This report analyzes how raising the state minimum wage—applicable to all industries—to $15 over roughly the same period as the fast food wage order would affect New York workers and their pay. The right side of Table 1 shows the schedule of minimum-wage increases modeled in this report. Note that in the years after the New York City minimum wage has reached $15 per hour, this analysis assumes that the minimum wage would be indexed to inflation such that the state labor department would automatically make adjustments each year to preserve the inflation-adjusted value of the wage floor. This type of automatic indexing is the best way to ensure that the real value of a minimum-wage income does not erode over time. It removes the need for policymakers to repeatedly legislate further increases, and it allows businesses to plan for a small and predictable increase in the wage floor each year. Currently, 15 states and the District of Columbia use automatic inflation indexing.7

Table 1

Demographic characteristics of affected workers

Raising New York’s minimum wage in stages to $15 by mid-2021 would lift pay directly or indirectly for nearly 3.2 million New York workers, comprising 36.6 percent of workers employed in wage and salaried positions throughout the state.8

Figure A shows the number of workers who would receive a raise as the minimum wage is gradually increased, combining the affected workers in both New York City and elsewhere in the state. In the first step, simulated to occur on April 1, 2016, the minimum is raised from $9.00 to $10.50 in New York City and to $9.75 in the rest of the state, raising pay for 2.06 million workers in the state. This includes 1.6 million workers who would directly benefit—meaning their current pay rate as of January 1, 2016, would be between $9.00 and $9.74 outside of New York City and between $9.00 and $10.49 in New York City. An additional 460,000 workers would indirectly benefit, meaning they would likely receive a raise through spillover or “ripple” effects because their current pay is just above the new minimum wage levels reached on April 1. Raising the minimum wage typically results in wage increases for workers further up the wage ladder because employers want to maintain some progression in their internal pay scales (Wicks-Lim 2006).

Figure A

With each successive increase, the cumulative number of workers who would benefit grows. By the second step (simulated to occur on December 31, 2016), when the minimum wage reaches $10.75 outside of New York City and $12.00 in New York City, 1.9 million workers would directly receive a raise, and 511,000 would indirectly receive a raise. Thus, the total number of affected workers would rise to 2.4 million in the second step, to 2.7 million in the third step a year later, and so on in the subsequent steps, reaching just under 3.2 million in the final step on July 1, 2021.

Of the 3.2 million New York workers who would benefit from the minimum-wage increase, 1.4 million work in New York City (45.6 percent of the total affected population), and 1.7 million work in the downstate suburbs and upstate (54.4 percent of affected workers throughout the state). Among all New York City workers, 34.8 percent would benefit from the increase. Among workers elsewhere in the state, 38.3 percent would receive a raise. See Appendix Table A1 for details on the number of workers directly and indirectly affected at each step of the minimum-wage phase-in for both city and non-city workers.

Age

The low-wage workers likely to benefit from increasing the minimum wage are often stereotyped as teenagers earning discretionary spending money. Although this would not justify paying them wages significantly lower than those paid to their counterparts a generation ago, this stereotype is false. In fact, as the tendency for young adults to attend college has grown over the years, teenagers account for a relatively small portion of the New York workforce and represent only 5.2 percent of those who would be affected by increasing the minimum wage to $15. Nearly 95 percent of affected workers are at least 20 years old.

Figure B indicates that over three-fifths of those affected are between the ages of 25 and 54. Among affected workers, far more are age 40 or older (40.9 percent) than are under age 25 (23.8 percent).

Figure B

Gender

Women account for 52.7 percent of workers who would be affected by the minimum-wage increase, slightly higher than their share of the New York workforce. As shown in Figure C, 39.0 percent of New York’s working women would get a pay increase if the state minimum wage were raised to $15 by mid-2021, as would about one-third (34.3 percent) of the state’s working men. About 37 percent of working mothers and 24.3 percent of working fathers would benefit. The affected rates are even higher for single parents: Over 45 percent of single working mothers would get a raise from a $15 state minimum wage, as would 39.3 percent of single working fathers.

Figure C

Race/ethnicity

When the state is considered as a whole, about half (49.1 percent) of those who would benefit from a minimum-wage increase to $15 are white, non-Hispanic workers, as shown in Figure D. Latino workers of any race make up the next largest share, at just under a quarter (24.6 percent) of the total affected population. Black or African American workers are 15.2 percent of the total, Asians comprise 8.9 percent, and workers of other races or ethnicities make up the remaining 2.2 percent.

Figure D

Despite being a smaller share of the total affected population statewide, workers of color would benefit from a minimum-wage increase at significantly higher rates than would white, non-Hispanic workers. The bar chart in Figure D shows the share of each race or ethnic group that would receive a raise if the minimum wage were increased to $15 by mid-2021. As the figure shows, 40.5 percent of all black or African American workers would receive higher pay, as would more than half of all Latino workers. More than a third (37.5 percent) of Asian workers would receive a raise—while 30.9 percent of white, non-Hispanic workers would receive higher pay.

Because the racial and ethnic composition of New York City is different than that of the rest of the state, there are meaningful differences in the composition of the workers likely to be affected by the proposed minimum-wage increase to $15. Among the workforce in New York City, Hispanic workers comprise the largest share of the affected population, at 36.7 percent of workers likely to get a raise. White, non-Hispanic workers make up just less than one quarter (24.0 percent) of the affected workforce. Black or African American workers account for 21.5 percent, and Asians are 15.3 percent of those affected. Demographic breakdowns for the affected populations in the state as a whole, in New York City, and outside New York City can be found in Appendix Tables A2, A3, and A4, respectively.

Education

Many of the workers who would benefit from increasing the New York minimum wage to $15 by 2021 have more education than is commonly acknowledged. As shown in Figure E, more than half of affected workers have at least some college experience, with nearly 10 percent having an associate degree and almost 20 percent holding a bachelor’s degree or higher.

Figure E

The bar graph in Figure E shows the share of workers at each education level who would receive a raise from increasing the state’s minimum wage to $15 by 2021. Not surprisingly, workers with lower levels of education are far more likely to be affected. Two-thirds (67.0 percent) of workers with less than a high school education would receive a pay increase, as would half (50.8 percent) of all workers with only a high school diploma. Just over one-third (35.4 percent) of those with an associate degree stand to benefit from the higher minimum wage, as do about one in six New York workers with a bachelor’s degree or higher.

Hours of work

Many New York workers who would benefit from a state minimum-wage increase to $15 also work longer hours than is commonly acknowledged; they are not predominantly working part time or in after-school jobs. As shown in the pie chart in Figure F, two-thirds of affected workers work full time, defined as working at least 35 hours per week. Another 24.7 percent work between 20 and 34 hours per week, and only 8.4 percent work less than 20 hours per week.

Figure F

It is important to note that many individuals who work less than full time are not opting for fewer hours by choice. Some are limited by a lack of available work, while circumstances prevent others from seeking full-time employment, such as the need to care for a family member, or a lack of adequate work supports that might facilitate a full-time schedule (e.g., access to child care, paid leave, or a flexible work schedule). For these workers, an increase in their hourly rate of pay is arguably even more important, as it could provide resources that could enable them to work more hours.

The bar chart in Figure F shows that 30.5 percent of full-time workers in the state are likely to benefit from raising the minimum wage to $15, compared with nearly two-thirds of those working 20­–34 hours per week and more than half of those who work less than 20 hours per week.

Household and family income

The great majority of New York workers who would benefit from increasing the minimum wage come from families of modest means. As shown in Figure G, about 43 percent of affected workers have total household incomes of less than $50,000, and nearly 63 percent have household incomes of less than $75,000. While these levels of household income may seem high relative to a minimum-wage income, overall household incomes of New York workers, including those who commute from other states, tend to be higher than elsewhere in the country. Less than a quarter (24.3 percent) of all New York workers (including those who commute) have total household incomes below $50,000, and only 42 percent have household incomes less than $75,000. This means that workers from the least-well-off households would disproportionately benefit from the wage hike.

Figure G

Often, minimum-wage opponents say that the minimum wage is “poorly targeted” because some of the workers who would benefit from a minimum-wage hike come from middle-income families. Considering how much of the net job change since the onset of the Great Recession has occurred in low-wage sectors, it is perhaps not surprising that many middle-income families have a low-wage earner.9 Yet the fact that the minimum wage provides protection to workers at all levels of family income is not a failing of the law; rather, it reflects the minimum wage’s role as a labor standard. The minimum wage prevents exploitation of workers, regardless of their socioeconomic background. No worker, regardless of his family income level, should have to work for unacceptably low wages. Moreover, the fact that both low- and middle-income families alike stand to benefit from an increase in the state’s minimum wage underscores that the failure to adequately raise the minimum wage over the past 45 years has hurt both low- and middle-income families.

Nevertheless, when looking at family income of the affected workforce (listed in Appendix Table A2), it becomes clear that many of the workers who would benefit from a higher state minimum wage are desperately in need of higher incomes.10 Just less than a quarter of affected workers are either in poverty or have family incomes within 150 percent of the poverty line. In fact, 37.1 percent of affected workers are either in poverty or what experts often describe as “near poverty,” having total family income less than 200 percent of the poverty line. Roughly 18 percent of all New York workers have family incomes that fall into this “near poverty” range, indicating again that the workers with the greatest family need would disproportionately benefit from a higher state minimum wage.

Figure H shows the share of workers grouped by ratio of their family income to the poverty line who would benefit from the higher minimum wage. Three quarters of workers in poverty would get a raise, as would 78.2 percent of workers with family incomes between 101 percent and 200 percent of the poverty line. In contrast, just over 20 percent of workers with family incomes at or above 300 percent of the poverty line would be affected by the policy change.

Figure H

Family status and children

Many of the New York workers who would benefit from increasing the state minimum wage to $15 are supporting families and children. As shown in the pie chart in Figure I, more than one-third (35.8 percent) of the affected workers are married, and just less than one-third (33.0 percent) of affected workers have children.

Figure I

The bar chart portion of Figure I shows the share of each group of workers—by their family status—who would benefit from increasing the state minimum wage to $15 by 2021. More than a quarter of all working married parents in New York would get a pay raise. For single parents facing the challenge of raising one or more children on their own and having to juggle the demands of child care and work, low pay only compounds their difficulties. In New York, over 340,000 working single parents—roughly 44 percent of all working single parents in the state—would get a raise if the state minimum wage were increased to $15.

Statewide, 1.3 million children have a working parent who would benefit from a minimum-wage increase to $15 by 2021—equaling 34.3 percent of all children in the state.

Researchers also suspect that the increased share of jobs paying low wages in recent years has contributed to a rising share of young people delaying marriage, postponing having children, and continuing to live with their parents longer into adult life.11 Thus, it is not surprising that many unmarried or childless young adults in their 20s and 30s would benefit from an increase in the minimum wage.

The importance of affected workers’ pay to their total family incomes

Low-wage workers are sometimes characterized as “secondary earners,” suggesting that their work earnings are discretionary or inconsequential to their family’s financial health. The data show that this is not at all the case; the workers who would benefit from increasing the minimum wage to $15 by 2018 in New York City or by 2021 in the rest of the state are often the primary breadwinners for their families. On average, workers who would benefit from increasing the minimum wage to $15 earn 50.3 percent of their family’s total income. Among workers age 25–39—who constitute 35.3 percent of all affected workers—their earnings account for nearly 58 percent of their family’s total income. More than 855,000, or 27.0 percent, of all affected workers are the sole providers of income for their families.

Raising the state minimum wage to $15 would provide a significant boost to the incomes of affected workers. Once the $15 minimum is fully phased-in, affected workers would receive, on average, $4,800 more in annual income (assuming no change in work hours).

10 economic sectors account for nearly three-fourths of affected workers

Table 2 shows the 10 industries in New York that account for the largest numbers of workers who would be affected by increasing the state minimum wage to $15 by 2021. Combined, these 10 sectors account for just over 2.3 million, or 73.1 percent, of the 3.2 million affected workers.

Table 2

Retail trade and restaurants have long been recognized as the largest employers of low-wage workers likely to be affected by raising the minimum wage. The New York Fast Food Wage Board has already acted to institute a wage order that is expected to raise pay for an estimated 132,000 workers employed by large chain restaurants. The wage order would cover roughly one-third of all restaurant workers projected to be affected by a phased-in $15 minimum wage. (Note that these workers are included in the figures presented here; see endnote 8 for further detail.) Retail has, by far, the greatest number of workers who would be affected by the proposed minimum-wage increase, with over 555,000 workers likely to get a raise, or 57.7 percent of all retail trade workers. Within retail, just over two-thirds of all grocery store workers (not listed in the table) would be affected by raising the minimum wage to $15.

A large group of workers providing health and human services would also be affected by a statewide increase in the minimum wage. Among these workers are home healthcare workers in the ambulatory care sector, nursing home workers and providers of services to the developmentally disabled in the residential care sector, and child care and other human service providers in the social assistance sector. Most of these low-wage health and human service workers are employed by nonprofits under contract to state and local governments or are funded under a Medicaid-supported program. In total, there are approximately 420,000 New York workers in these three health and human service sectors who would benefit from raising the minimum wage to $15, representing nearly half (48.4 percent) of all workers in these sectors. Among child care workers, who are included in social assistance, over 60 percent would be affected, and if home healthcare workers were broken out of the broader ambulatory sector, their share likely would be over 90 percent since they are among the lowest-paid workers in the state.12

About 20 percent of local government employees would be affected by the higher minimum wage, although this is one of the largest employment sectors in the state. In total, roughly 172,000 local government workers, including those working for school districts, would be affected by the proposed increase.

Other sectors among the 10 largest employers of affected workers include educational services; finance, insurance, and real estate; construction; and transportation and warehousing.

Tables showing the breakdown of affected workers by industry in New York as a whole, in New York City, and in the rest of the state can be found in Appendix Tables A5, A6, and A7, respectively.

Conclusion

Since its inception in the Great Depression, a strong minimum wage has been recognized as a key labor market institution that, if effectively maintained, can provide the foundation for equitable and adequate pay for American workers. However, the failure to regularly and adequately raise the minimum wage over the past five decades is one of several policy failures that have denied a generation of American workers more significant improvement in their quality of life. In fact, the erosion of the minimum wage has left low-wage workers today earning significantly less (in inflation-adjusted terms) than their counterparts 50 years ago.

In the absence of changes at the federal level, state policymakers are taking action into their own hands by updating state labor standards to reflect today’s economy and the needs of workers and their families. Raising the New York minimum wage in several steps to $15 would restore its value to a level that ensures full-time work is a means to escape poverty—and would provide more than a third of New York’s workers with a long-overdue improvement in their standard of living.

About the author

David Cooper is an economic analyst with the Economic Policy Institute. He conducts national and state-level research on a variety of issues, including the minimum wage, employment and unemployment, poverty, and wage and income trends. He also provides support to the Economic Analysis and Research Network (EARN) on data-related inquiries and quantitative analyses. David has been interviewed and cited by numerous local and national media for his research on the minimum wage, poverty, and U.S. economic trends. He holds a Master of Public Policy degree from Georgetown University.

Appendix A: Additional tables and figures

Appendix Table A1

Appendix Table A2

Appendix Table A3

Appendix Table A4

Appendix Table A5

Appendix Table A6

Appendix Table A7

Appendix B: Technical documentation and methodology

EPI’s minimum-wage simulation model relies on 2014 data from the American Community Survey (ACS) published by the U.S. Census Bureau, and harmonized by Ruggles et al. (2015) at the Minnesota Population Center. The ACS is the largest annual survey conducted by the Census Bureau, interviewing more than 2.3 million households throughout the United States. The survey measures a host of demographic, social, and economic characteristics of the U.S. population, and was created to replace the long-form version of the decennial census.

The large size of the ACS sample makes it ideal for studying state and sub-state areas. In addition, because the ACS is the only major public statistical survey containing information on individuals’ place of work—as opposed to simply recording respondents’ place of residence—it is ideal for assessing changes in labor policy. However, the ACS only records workers’ annual income and wages. Therefore, to study the effects of changes in hourly wages, we must first impute hourly wage information for all workers in the sample using their reported annual wage income, usual hours of work per week, and weeks worked in the previous year. One further complication is that the ACS asks respondents to record the number of weeks they worked in the previous 12 months by selecting one of several intervals (e.g., 1–13 weeks, 14–26 weeks, 50–52 weeks, etc.). Thus, in order to impute an hourly wage, we must first construct a discrete value for respondents’ weeks worked in the previous year. To do this, we use the Annual Social and Economic Supplement to the Current Population Survey (CPS-ASEC), which does include information on workers’ discrete weeks worked in the previous calendar year. Using linear regression, we estimate how various demographic, work, and economic variables vary with respondents’ weeks worked per year, utilizing separate regression models for each interval. We then apply the regression coefficients from those models to the ACS data to predict workers’ discrete weeks worked in the previous year.

We estimate respondents’ hourly wage by dividing their annual wage income by their usual hours of work per week, multiplied by their predicted weeks worked in the previous year. This imputation process is not ideal, as imputing hourly wages in this way compounds measurement error in the three variables used to derive hourly wages. Indeed, imputed hourly wages fall below the statutory minimum wage for roughly 10 percent of the ACS worker sample. However, the ACS is the only publicly available data set with the adequate sample size and place of work information needed to conduct this type of analysis.

We restrict the ACS sample to individuals age 16 and older, who are currently employed and for whom imputed wage values are greater than $0.93 or less than $185. In all calculations of workers affected by increases in the minimum wage, we exclude all observations with imputed hourly wages less than 50 percent of the statutory minimum wage prior to the simulated increase.

Prior to the simulation in this analysis, we adjust wage values to reflect the increase in New York’s state minimum wage to $9.00, which occurred on December 31, 2015. Observations with imputed hourly wages below $9.00 are increased in proportion to the change in the minimum wage. For example, if an observation has an imputed hourly wage of $8.80 in 2014, when the state minimum wage was $8.75, their hourly wage is adjusted to equal ($8.80/$8.75) x $9.00 = $9.05. In addition, we assume natural nominal wage growth prior to the first increase equal to the rate of nominal average wage growth for workers in the bottom 20 percent of the hourly wage distribution for New York, prorated for the number of months between the midpoint of the sample period and the month prior to the first simulated increase in the state minimum wage.

We also assume population growth between the data period and the proposed first increase. ACS person weights are adjusted by the projected annual New York state population growth rate from 2015 to 2020, as estimated by the Cornell Population Center, of 0.154 percent (Cornell University 2011). This annual growth rate is prorated by the number of months that occur between the midpoint of the data and the month that the first proposed minimum-wage increase would occur.

Having made these adjustments, “directly affected” workers are identified as those workers with hourly wages between 50 percent of the statutory minimum wage prior to the proposed increase, and the proposed minimum wage applicable in the worker’s jurisdiction of work (i.e., within New York City or elsewhere in the state). We identify “indirectly affected” workers as those workers whose wages are greater than or equal to the proposed new minimum wage, but less than 115 percent of the dollar value of the proposed increase—hereafter referred to as the “indirectly affected cutoff.” This cutoff point is chosen to reflect the findings of Dube, Giuliano, and Leonard (2015), which observed minimum-wage spillover or “ripple” effects for workers earning 15 percent above newly implemented minimum wages. For example, for an increase from $9.00 to $9.75, directly affected workers have a wage between $4.50 and $9.74. Indirectly affected workers would be those workers with wages between $9.75, inclusive, and $11.21, exclusive. The indirectly affected cutoff in this case would be $11.21.

After each step, if an individual is predicted to be either directly or indirectly affected, her wage is adjusted to reflect her implied raise. For directly affected workers with hourly wages of at least 90 percent of the statutory minimum wage prior to the simulated increase, their raise is equal to the greater of: 1) the difference between the new minimum wage and their existing wage; or 2) one-quarter of the difference between their existing wage and the indirectly affected cutoff. We allow for these two possibilities because we believe it is likely that workers close to, yet still below, the new minimum wage prior to the increase will receive a larger raise than simply an increase to the new minimum. For example, if a worker was earning $9.70 per hour prior to an increase in the minimum wage from $9.00 to $9.75, it stands to reason that her employer will give her more than a $0.05 raise, particularly if lower-paid colleagues are receiving raises as large as $0.75. For workers earning between 50 percent and 89 percent of the minimum wage prior to the increase, they are given a raise proportional to the increase in the minimum. For example, a worker earning 75 percent of the minimum wage prior to the increase would receive a raise that brought her to 75 percent of the new minimum wage.

For all indirectly affected workers, their raise is modeled as one-fourth of the difference between their existing wage and the indirectly affected cutoff. For example, an indirectly affected worker previously earning $10.00 in the above scenario would receive a raise of 0.25 x ($11.21-$10.00), or $0.30.

Having counted these directly and indirectly affected workers, the program iterates to the next proposed increase. Again, weights are adjusted to reflect the predicted population growth between the first and second increments in the proposed minimum-wage increase. Wage values are again adjusted to reflect natural nominal wage growth; however, all workers who received a raise as a result of the higher minimum wage are given only 50 percent of the assumed natural nominal wage growth applied to all other observations. The same method for identifying directly and indirectly affected workers is applied, and the counts are recorded. The model iterates in this fashion for all remaining steps.

Two additional controls are applied throughout the simulation. First, workers in occupations that customarily receive tips as the majority of their earnings are coded with hourly wage values equal to the applicable subminimum wage for tipped workers, or “tipped minimum wage.” We make this adjustment so that reported changes in wage values reflect changes in the wages required to be paid by employers or tipped staff. See Allegretto and Cooper (2014) for more information on tipped minimum wages, and the occupations that are considered tipped occupations. In New York, the tipped minimum wage is equal to 83 percent of the full minimum wage; thus in all steps, workers in tipped occupations are set to 83 percent of the applicable minimum wage in each step. Second, because state legislatures have no jurisdiction over federal employees, federal government workers are excluded from the directly and indirectly affected groups throughout this analysis. However, federal government employees are included in counts of the total workforce.

Finally, no controls are made in this analysis for workers already scheduled to receive a raise as a result of the fast food wage order or the governor’s executive order to raise wages for state employees. The fast food wage order applies to “chain” fast food restaurants, defined as those with at least 30 locations nationwide. The New York Fast Food Wage board estimated that 62 percent of fast food workers in the state were in chain locations; however, the data used in this study do not allow for us to identify chain versus non-chain employees. With roughly 213,000 fast food workers in the state and an estimated 62 percent in “chain” restaurants, this means that we may be overestimating the population likely to be affected by a statewide minimum-wage increase, independent of the fast food wage order, by roughly 132,000 workers. Additionally, we estimate that roughly 67,000 state employees would likely receive a raise under a statewide minimum-wage increase, independent of the governor’s executive order. Consequently, the results may overstate the total affected workforce resulting from a statewide minimum-wage increase by as much as 199,000 workers, or roughly 6.3 percent of the total.

Endnotes

1. Author’s calculations using New York State Budget Office Projections for the state CPI, 2014–2021, and data from the Economic Policy Institute’s Family Budget Calculator (Gould, Cooke, and Kimball 2015).

2. See Cooper (2015c) or National Employment Law Project (forthcoming 2016).

3. Author’s calculations using the CPI-U-RS

4. Author’s calculations using the CPI-U-RS, Regional Price Parity (RPP) data from Bureau of Economic Analysis (2015), and New York State Budget Office projections for the state CPI 2014–2016. Such an adjustment would not be appropriate if New York prices were equally high, relative to the national average, in 1970. However, data on housing costs from the U.S. Census Bureau indicate that that is very likely not the case. Housing costs are the largest driver of regional price differences. Data from the U.S. Census Bureau show that from 1970 to 2010, median gross rents rose by 78 percent in constant dollars nationally (author’s calculations using data from U.S. Census Bureau 2014 and American FactFinder). However, in New York state, median gross rents rose by 89 percent in constant dollars over the same period, indicating that overall prices in New York have almost certainly grown more rapidly than the national average.

5. One indication of the trend where the average worker has not been benefitting from the growth in the state’s economy is that, from 2001 to 2013, business profits per worker in New York state increased by 61 percent, while labor compensation per worker rose by only 34 percent, about the same as the increase in consumer prices over that period (Fiscal Policy Institute 2015).

6. Author’s analysis of American Community Survey microdata, 2014.

7. For information on all state and local minimum-wage levels, see http://www.epi.org/minimum-wage-tracker/.

8. This figure includes workers in fast food, although most will receive wage increases as a result of the Fast Food Wage Board order. The wage order applies to “chain” fast food restaurants, defined as those with at least 30 locations nationwide. The wage board estimated that 62 percent of fast food workers in the state were in chain locations; however, the data used in this study do not allow for us to identify chain versus non-chain employees. Consequently, we do not control for those workers in fast food who will already receive wage increases. With roughly 213,000 fast food workers in the state and an estimated 62 percent in “chain” restaurants, this means we may be overestimating the population likely to be affected by a statewide minimum-wage increase, independent of the fast food wage order, by roughly 132,000 workers (4 percent of the total affected population). We also do not control for the governor’s executive order raising pay for state employees, as the exact timeframe for these increases has not been specified. An estimated 67,000 state employees would be affected in the proposed increase in this study.

9. Fiscal Policy Institute (2014), 102.

10. Family income measures a smaller unit than household income, as households can contain multiple coresident families.

11. See Dietz (2014) and Jacobsen and Mather (2011). According to the latter, the share of young adults 25–34 who are married fell from 55 percent in 2000 to 46 percent in 2011, and the number living at home rose from 4.7 million in 2007 to 5.9 million in 2011.

12. According to BLS Occupational Employment Statistics, the 90th percentile hourly wage in 2014 was $12.57 among the 81,000 home health care workers within New York’s ambulatory care industry.

References

Allegretto, Sylvia, and David Cooper. 2014. Twenty-Three Years and Still Waiting for Change. Economic Policy Institute Briefing Paper #379.

Allegretto, Sylvia, Marc Doussard, Dave Graham-Squire, Ken Jacobs, Dan Thompson, and Jeremy Thompson. 2013. Fast Food, Poverty Wages: The Public Cost of Low-Wage Jobs in the Fast Food Industry. U.C. Berkeley Center for Labor Research and Education.

American Community Survey. Various years. U.S. Census Bureau data set.

Bureau of Economic Analysis (U.S. Department of Commerce). 2015. Real Personal Income for States and Metropolitan Areas. Regional Price Parities [data tables].

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 Briefing Paper No. 378.

Bureau of Labor Statistics (U.S. Department of Labor) Occupational Employment Statistics program. Various years. Current Occupational Employment and Wages [economic news release].

Cooper, David. 2015a. Raising the Federal Minimum Wage to $12 by 2020 Would Lift Wages for 35 Million American Workers. Economic Policy Institute Briefing Paper #405.

Cooper, David. 2015b. “A $12 Minimum Wage Would Bring the United States in Line with International Peers.” Working Economics (Economic Policy Institute blog), May 6.

Cooper, David. 2015c. “Testimony before the New York State Department of Labor Fast Food Wage Board.” Testimony delivered June 22 in Albany, N.Y.

Cooper, David, John Schmitt, and Lawrence Mishel. 2015. We Can Afford a $12.00 Federal Minimum Wage in 2020. Economic Policy Institute Briefing Paper #398.

Cornell University. 2011. New York State Projection Data by County. Program on Applied Demographics.

Current Population Survey Annual Social and Economic Supplement microdata. Various years. Survey conducted by the Bureau of the Census for the Bureau of Labor Statistics [machine-readable microdata file]. Washington, D.C.: U.S. Census Bureau.

Dietz, Robert. 2014. “The Great Delay.” U.S. News and World Report, December 17.

Dube, Arindrajit, Laura Giuliano, and Jonathan Leonard. 2015. Fairness and Frictions: The Impact of Unequal Raises on Quit Behavior. IZA Discussion Paper No 9149.

Fiscal Policy Institute. 2014. The New York State Economic and Fiscal Outlook 2014-2015.

Fiscal Policy Institute. 2015. Business Profits in New York State Have Grown Much Faster Than Wages Since 2001; Minimum Wage Hike Is a Good Corrective. Fiscal Policy Institute Data Brief.

Gould, Elise, Tanyell Cooke, and Will Kimball. 2015. “Family Budget Calculator.” Economic Policy Institute.

Jacobsen, Linda A., and Mark Mather. 2011. “A Post-Recession Update on U.S. Social and Economic Trends.” Population Reference Bureau.

Karlin, Rick. 2015. “New York Fast Food Wage Board Hears Testimony About Potential Mandate of Higher Minimum Wage.” Albany Times Union, June 22.

Mishel, Lawrence. 2014. “The Tight Link Between the Minimum Wage and Wage Inequality.” Working Economics (Economic Policy Institute blog), January 27.

National Employment Law Project. 2016 (forthcoming). “How Much Do New York’s Workers Need? At Least $15 Per Hour—Both Upstate and Down.”

New York Department of Labor. 2015. “Order of Acting Commissioner of Labor Mario J. Musolino on the Report and Recommendations of the 2015 Fast Food Wage Board,” September 10.

Ruggles, Steven, Katie Genadek, Ronald Goeken, Josiah Grover, and Matthew Sobek. 2015. Integrated Public Use Microdata Series: Version 6.0 [Machine-readable database]. Minneapolis: University of Minnesota.

U.S. Census Bureau, Housing and Household Economic Statistics Division. 2011. Historical Census of Housing Tables.

U.S. Census Bureau. Various years. American Community Survey Table B25064. Generated using American FactFinder on December 20, 2015.

Wicks-Lim, Jeannette. 2006. Mandated Wage Floors and the Wage Structure: New Estimates of the Ripple Effects of Minimum Wage Laws. Political Economy Research Institute at the University of Massachusetts Amherst, Working Paper Number 116.

Zipperer, Ben, and David Evans. 2014. Where Does Your State’s Minimum Wage Rank Against the Median Wage? Washington Center for Equitable Growth.

These disparate estimates can be mostly reconciled by noting that the FU/U estimate is of anyone “engaged in supplemental, temporary, project- or contract-based work, within the past 12 months”3 and even includes people who “freelance” but do not have any 1099 income—stretching this group beyond recognition. In contrast, the BLS estimate reflects those whose primary job is or primary income comes from self-employment. Even the BLS estimate stretches the concept of self-employment because it includes people who themselves are employers: only 11.3 million self-employed persons, representing 7.7 percent of total employment, work for themselves and have no paid employees.4

As with most measurement issues, it is important to be clear what is being asked and what one hopes to measure and then examine whether the measurement is accurate. The major problem with the FU/U studies (a 2014 edition was commissioned by Freelancers Union and Elance-oDesk, which was renamed Upwork by 20155) is that the claims of a fundamental shift in “work” as more people make their living freelancing are based on data about how many people earn any income from freelancing even if they clearly rely primarily on other work to support themselves.

In fact, the FU/U survey results actually contradict their claim that freelancing as a way to make a living is rapidly expanding. For instance, although not in any of the press releases, the surveys show 1.8 million fewer independent contractors in 2015 than in 2014, an 8.5 percent drop. (Of the five types of freelancers identified by FU, “independent contractors,” defined as someone without an employer who does freelance, temporary, or supplemental work on a project-to-project basis, comes closest to the BLS concept of people obtaining their primary income from self-employment.) Plus, using the FU/U broadest definition of freelancers (all five categories) the surveys show that the share of “freelancers” in the economy was essentially the same in 2015 (34.2 percent)6 as in 2014 (34.0 percent).

Therefore, despite the hype around the FU/U freelancer estimates it is reasonable to conclude from the FU/U data themselves that freelancing is not exploding as a way that people earn their living in the U.S. today and that the share of the workforce freelancing for supplementary income is not even growing. As the analysis in the paper will show, an alternative headline for the FU/U study could easily be, “Only 12.3 percent of workers primarily rely on freelancing to support themselves,” rather than emphasizing that 34 percent of workers do some freelancing.

What’s the question the surveys answer?

To gauge whether 53–54 million is a good estimate of freelancers one must first determine what question the estimate is answering. Based on what the two studies emphasize in their opening pages, the purpose seems to be to assess how many people earn their living through freelancing, meaning how many people have freelancing as their main way of supporting themselves, their equivalent of a “job.” For instance, the 2014 study opens with the following headline and subhead:

“The way we work is changing. Gone are the days of the traditional 9-to-5. We’re entering a new era of work — project-based, independent, exciting, potentially risky, and rich with opportunities.”

The opening paragraph starts, “For the past two decades, Americans have felt a shift to a new kind of work. The traditional “full-time job with benefits” was becoming less and less common, replaced by a new gig-to-gig, project-to-project worklife. But while many could feel this shift, it was difficult to quantify.”

This description of the study’s findings and their implications focuses on freelancing as “the work” that is replacing full-time employment. Given that full-time jobs are the primary way people obtain income to support themselves it would seem the study is attempting to quantify whether more people are using freelancing as a way to support themselves.

The 2015 study maintains the same theme in its opening headline and subhead: “We have entered a new era of work in this country. Freelancing is becoming a more prevalent, viable option for workers— a trend that spans across borders, industries and occupations.”

If one presumes that work is the primary activity undertaken to generate one’s income then this opening implies that freelancing has become a more prevalent option for people to support themselves (earn a living) in the “new era.”

The two studies clearly position themselves as quantifying a shift in how people support themselves—from relying on regular, full-time jobs to freelancing. But the studies’ methodology for counting the number of freelancers does not match the stated purpose because it does not assess the number of people who primarily support themselves through freelancing. Rather, the estimated 53 million plus freelancers are anyone “engaged in supplemental, temporary, project- or contract-based work, within the past 12 months.” That is, anyone with any income generated by freelancing, whether it is their main source of income or not, is counted as a freelancer. Thus, someone with a full-time job earning a little supplemental income as a freelancer—a “moonlighter”—is counted as one of the 53 million plus freelancers. In fact, the reports so thoroughly scrounge survey participants for “freelancers” that even people who do not receive any 1099 income are counted. It may be interesting to know how many people obtain some income from freelancing—a market for Upwork or other firms— but such an estimate does not support claims about how work (how people earn their living) is changing in America.

How many people rely on freelancing as their main income source?

The FU/U surveys’ breakdowns of freelancers allow us to assess the number of people freelancing as their main source of income. The five FU categories of freelancing are presented in the text box and their numbers and shares are presented in Table 1. There are two panels in Table 1. In Panel A the FU/U survey findings are converted to numbers of people using the total labor force age 16 and older, as the FU/U studies do.7 Panel B uses total employment of those 18 and older. The panel B estimates based on employment are more appropriate. Each analysis starts from the findings regarding the share of each sample who are freelancers (e.g., the 34 percent estimate for the 2014 survey). The FU/U analysis multiplies this share by the number of people in the labor force to obtain 53 million in 2014 and 53.7 million in 2015. However, the labor force includes those who are employed as well as those who are unemployed. The convention for identifying the different types of work is to examine the number and shares of total employment, as panel B does. The data presented in panel B also adjust total employment to reflect workers age 18 or older to match the FU/U survey design (16- to 17-year-olds make up 1.1 percent of BLS estimates of total employment).8 Benchmarking against employment rather than the labor force reduces the estimate of freelancers in 2014 by 7.2 percent, from 53 million to 49.2 million.

Table 1

Definitions of freelancer subcategories

It would have been useful if the FU/U surveys identified the share of each person’s total income from freelancing or 1099 work so that the share of the sample primarily relying on freelance income to get by could be straightforwardly determined. One wonders why the surveys did not do this given the interest in the answer. Instead, to determine how many people primarily earn their living through freelancing necessarily requires judging which of the five subgroups have that characteristic.

“Independent contractors” seem to be the only freelancers primarily supporting themselves through self-employment, because they do not have any employer (though in the FU/U surveys they do not necessarily receive 1099 income). The other four categories do not capture people who primarily support themselves through self-employment. “Freelance business owners” have paid employees so they do not fit. “Moonlighters” actually have just one employer but simply do some freelancing for supplementary income (they do not necessarily receive 1099 income). “Temporary workers” have just one “temporary status” job, but do some freelancing for supplementary income (they do not necessarily receive 1099 income); according to their survey responses they do not do any work outside of their primary position (the temp job) to earn additional money. The “diversified” group is actually two segments, though there is no information about the size of each segment. Members of both segments have more than one employer. Those in the first segment receive 1099 income. Those in the second segment do not receive any 1099 income but are categorized as freelancers because each considers him or herself “an independent professional” and has done some freelance work. It is highly unlikely that this second segment primarily relies on self-employment income. As to the first segment, without knowing whether their 1099 income exceeds that obtained from their employers we can’t know whether self-employment provides the primary income source. On the whole, the diversified group does not fit the description of those primarily supporting themselves through self-employment.

The employment trends for 2014 to 2015 for each subgroup are presented in Panel B. Independent contractors, the only group that clearly primarily supports itself through self-employment, account for about two-fifths of all freelancers and fell from 19.6 million to 18.1 million between 2014 and 2015, a 7.7 percent drop. As the bottom panel shows, these independent contractors constituted 13.5 percent of all employment in 2014 but fell to just 12.3 percent in 2015. The Freelancers Union/Upwork studies’ headlines about self-employment becoming an increasingly important option for many workers do not match these data trends, nor do the data support the notion that we will soon become a nation of freelancers.

Reconciling BLS and FU/U estimates

Still open is the question of whether the FU/U survey results for independent contractors tell us something that the BLS does not. At first blush the estimates might seem close. The FU/U survey tells us that in 2014 13.5 percent of total employment (19.6 million of the 144,791,000 employed persons age 18 or older) was in self-employment as a primary source of income, 3.4 percentage points higher than the BLS estimate of 10.1 percent. The gap is even larger once one recognizes that nearly a fourth of the BLS estimate of the self-employed is made up of employers with their own paid employees. In contrast, FU/U independent contractors do not have any employees. So, the better comparison is to the 11.3 million BLS self-employed with no paid employees, who constituted 7.7 percent of total employment in 2014.9 That indicates a more considerable 8.3 million gap between the FU/U survey and BLS, or 5.8 percent of total employment in 2014.

What can explain the gap? One technical issue is that the FU/U survey asks people whether they freelanced over the last 12 months whereas the BLS estimates the number of people self-employed each month—a cross-section estimate. In 2013, for instance, the number of people who worked at some point over the year (156,987,000)10 was roughly 9 percent greater than the average employment (143,929,000) in all 12 months of 2013,11 which suggests that the FU/U estimate would be expected to be about 9 percent higher. Correcting for that would explain just 1.6 million of the 8.3 million gap.12 The only other clear difference is that the FU/U survey counts people as independent contractors even if they have not received any 1099 income. This may pump up the FU/U estimate relative to BLS.

There is much discussion among analysts that BLS may not capture all the ways that people bring in income—that many more people earn some self-employment income than what BLS captures. This may be so, but is irrelevant for discussions about the “future of work” and possible shifts in how people will support themselves. After all, there is no apparent reason to believe that BLS does not adequately capture the number of people who earn self-employment income as their primary source of income, and that is the metric that should drive any analysis regarding the “future of work.”

Conclusion

Based on current trends there is no reason to believe that in the near or intermediate future a large and growing share of people will obtain their main source of income from freelancing or doing gig work. The Freelancers Union/Upwork studies provide no evidence to the contrary since their claims are based on a count of freelancers that included anyone who does any freelancing, even when it is not the main source of income, and many who do not even receive 1099 income.

That does not mean that there are not important issues regarding the economic security of freelancers. Certainly any repeal of the Affordable Care Act would limit freelancer access to quality and affordable health care. There is a growing interest in providing portable benefits for freelancers and other types of workers. There is also the desire to help freelancers get paid for their work and paid in a timely fashion, an issue raised by a New York City Council bill that the Freelancers Union backs.13 Freelancers definitely need an advocacy group such as the Freelancers Union.

But any assessment that self-employment has been stable does not speak to whether employee misclassification as independent contractors has risen.14 Blue-collar construction workers on major federal public housing projects who are inappropriately and illegally treated as independent contractors—as recently documented in the award-winning McClatchy series15­­—probably don’t respond to BLS surveys by answering that they are self-employed. Nor should an analysis of self-employment minimize the growth in overall nonstandard work and the fissuring of the workplace,16 including the spread of subcontracting to shield major firms from liability for shady employment practices and from the demands of low-wage workers. There needs to be attention to all of these issues and to the larger question of how to generate robust wage growth. Hype about freelancing and gig work limits our focus and perhaps distracts from these broader issues.

About the author

Lawrence Mishel, a nationally recognized economist, has been president of the Economic Policy Institute since 2002. Prior to that he was EPI’s first research director (starting in 1987) and later became vice president. He is the co-author of all 12 editions of The State of Working America. He holds a Ph.D. in economics from the University of Wisconsin at Madison, and his articles have appeared in a variety of academic and non-academic journals. His areas of research are labor economics, wage and income distribution, industrial relations, productivity growth, and the economics of education.

Endnotes

1.  Freelancing in America: A National survey of the new workforce, Edelman Berland (commissioned by Freelancers Union and Elance-oDesk), 2014; Freelancing in America: 2015, Edelman Berland (study commissioned by the Freelancers Union and Upwork), 2015. The data from the 2015 online survey of 7,107 U.S. adults were collected July 30 to August 14, 2015.

2. Author’s computations of BLS data on self-employed, both incorporated and unincorporated, and total employment.

3. “Freelancing in America: 2015: Results Deck,” Edelman Berland (study commissioned by the Freelancers Union and Upwork), October 1, 2015.

4. Lawrence Mishel, “Claims of an exploding gig economy are contradicted by the lack of growth of self-employed workforce,” Economic Policy Institute Economic Snapshot, October 28, 2015

5.  Freelancing in America: A National survey of the new workforce, Edelman Berland

6. Author’s analysis of FU/Upwork survey results, in Table 1 of this report.

7. The FU/U survey sample includes respondents age 18 and older but the survey benchmark is the labor force, which captures those age 16 and older.

8. Table 1 uses the FU/U estimates of the number of people in each category as presented in each year’s study, which sum to the total, to compute shares of the total. Panel B uses the shares estimated in Panel A and the total employed (18+) to compute the number of people in each category.

9. Lawrence Mishel, “Claims of an exploding gig economy are contradicted by the lack of growth of self-employed workforce.”

10. Bureau of Labor Statistics, “Work experience of the population—2013,” news release, December 16, 2014.

11. Bureau of Labor Statistics, “Household data annual averages: 1. Employment status of the civilian noninstitutional population, 1944 to date” (online data table).

12. The latest data are for 2013. The number of workers with any work experience in 2013 was 156,987,000 whereas total employment, on average across the 12 months, was 8.3 percent less, 143,929,000. If the 19.6 million independent contractors estimate for 2014 was overstated by this then it would be 8.3 percent, or 1.6 million, less.

13. Chester Soria, “New York City Council bill would fine, jail employers who stiff freelancers,” Metro US, December 7, 2015.

14. Francoise Carre, Independent contractor misclassification, Economic Policy Institute Briefing Paper #403, 2015.

15. The Sidney Hillman Foundation, “McClatchy & ProPublica win October Sidney for exposing multi-billion dollar tax fraud by worker misclassification,” October, 2014.

16. In his book The Fissured Workplace, published in February 2014, David Weil explains how the strategy of large corporations to split off functions that were once managed internally has led to stagnating wages and benefits for workers.

• Chinese imports entering through Wal-Mart in 2013 likely totaled at least $49.1 billion and the combined effect of imports from and exports to China conducted through Wal-Mart likely accounted for 15.3 percent of the growth of the total U.S. goods trade deficit with China between 2001 and 2013.
• The Wal-Mart-based trade deficit with China alone eliminated or displaced over 400,000 U.S. jobs between 2001 and 2013.
• The manufacturing sector and its workers have been hardest hit by the growth of Wal-Mart’s imports. Wal-Mart’s increased trade deficit with China between 2001 and 2013 eliminated 314,500 manufacturing jobs, 75.7 percent of the jobs lost from Wal-Mart’s trade deficit. These job losses are particularly destructive because jobs in the manufacturing sector pay higher wages and provide better benefits than most other industries, especially for workers with less than a college education.
• Wal-Mart has announced plans to create opportunities for American manufacturing by “investing in American jobs.” To date, very few actual U.S. jobs have been created by this program, and since 2001, the growing Wal-Mart trade deficit with China has displaced more than 100 U.S. jobs for every actual or promised job created through this program.

China has achieved its rapidly growing trade surpluses by manipulating its currency: it invests hundreds of billions of dollars per year in U.S. Treasury bills, other government securities, and private foreign assets to bid up the value of the dollar and other currencies and thereby lower the cost of its exports to the United States and other countries. China has also repressed the labor rights of its workers and suppressed their wages, making its products artificially cheap and further subsidizing its exports. Wal-Mart has aided China’s abuse of labor rights and its violations of internationally recognized norms of fair trade by providing a vast and ever-expanding conduit for the distribution of artificially cheap and subsidized Chinese exports to the United States.

China trade and U.S. job loss

Exports support jobs in the United States, and imports displace them. Thus, the net effect of trade flows on employment must be based on an analysis of the trade balance. This Briefing Paper calculates the employment effects of growing goods trade deficits by using an input-output model that estimates the direct and indirect labor requirements of producing output in a given domestic industry. The model includes 195 U.S. industries, 77 of which are in the manufacturing sector.1

The model estimates the labor that would be required to produce a given volume of exports, and the labor that is displaced when a given volume of imports is substituted for domestic output.2 The job losses presented here represent an estimate of what total employment levels would have been in the absence of growing trade deficits.3

U.S. exports to China in 2001 supported 161,400 jobs, but U.S. imports displaced production that would have supported 1,127,700 jobs, as shown in the bottom half of Table 1. Therefore, the $84.1 billion goods trade deficit in 2001 displaced nearly 1 million jobs in that year. Net job displacement rose to 4,123,400 in 2013. Growth in trade deficits with China has reduced demand for goods produced in every region of the United States and has led to job displacement in all 50 states and the District of Columbia. The overall China trade and job loss estimates in this report are based on the findings reported in Kimball and Scott (2014).

Table 1

Wal-Mart’s role

Given its enormous size and the fact that it sells manufactured goods, which have been the primary Chinese export to the U.S. in recent years, it is natural to try to estimate the role of Wal-Mart as a conduit for Chinese trade. We find that a conservative estimate is that Wal-Mart accounted for approximately 11.2 percent of total U.S. goods imports from China between 2001 and 2013. This estimate is based on published reports on Wal-Mart trade with China between 2001 and 2004, including Wal-Mart’s own estimates of its imports from China, on more recent published data on ocean trade (by company), and on the relationship between total Wal-Mart sales in the United States and personal consumption expenditures on goods from the GDP accounts (BEA 2015).

Wal-Mart provided its own estimate for the value of imports from China in its fiscal year ending January 31, 2004 (Wal-Mart 2007). Most of these goods were imported in 2003, and the Wal-Mart share of total imports from China in that year was 11.9 percent. Bianco and Zellner (2003) and Bianco (2006) have also attempted to construct estimates of Wal-Mart’s imports from China and have reported imports that yield shares that are similar to Wal-Mart’s own estimates, with the lowest share reported as 11.2 percent in 2004. Since 2007, evidence strongly suggests that this share has not shrunk (and may have risen). For example, The Journal of Commerce produces annual reports of total U.S. imports and exports of goods via ocean container transport.4 While this is a partial and incomplete accounting, it does show that Wal-Mart was the top U.S. importer of ocean container freight in every year between 2001 and 2013, and its share of top 100 imports remained stable in a range from 12.1 percent to 14.8 percent of total imports of the top 100 importers.5

Limited data on total imports by company are also available from shipments data collected by the U.S. Customs and Border Protection agency.6 Data on Wal-Mart imports are available for only two comparable months in the study period: November 2007 and 2012. The available information reports total imports in both kilograms and container equivalents (twenty-foot equivalent units or TEUs). The Wal-Mart share of total imports from China increased in both kilograms and TEUs in this period (Panjiva.com 2015). In short, the 2003 share of imports accounted for by Wal-Mart as estimated by the company itself (11.2 percent) has likely only grown since then. However, for this report we make the conservative assumption that it has remained stable.

But a stable share of Wal-Mart imports implies rapid growth in volumes. U.S. goods imports from China increased $336.1 billion between 2001 and 2013, as shown in the top half of Table 1, an increase of 329 percent. If Wal-Mart’s share of U.S. imports from China remained stable in this period at 11.2 percent, this implies that its imports increased from $11.4 billion in 2001 to $49.1 billion in 2013, an increase of $37.6 billion. As it is a retailer and not a manufacturer, Wal-Mart likely exports only a negligible amount to China. Our best estimate is that Wal-Mart accounts (at most) for roughly 1.0 percent of total U.S. exports to China.7 This in turn implies that Wal-Mart was responsible for a $36.7 billion increase in the U.S. trade deficit with China between 2001 and 2013.

The Wal-Mart trade deficit displaced 125,800 jobs in 2001 and 541,300 jobs in 2013.8 Thus, Wal-Mart was responsible for displacing at least an additional 415,400 U.S. jobs between 2001 and 2013, as shown in the bottom half of Table 1 and in Figure A. While Wal-Mart was responsible for 11.2 percent of U.S. imports in this period, it was responsible for 13.2 percent of the U.S. job losses due to growing trade deficits with China (Table 1). Since Wal-Mart’s exports to China were negligible, the rapid growth of its imports had a proportionately bigger impact on the U.S. trade deficit and job losses than overall U.S. trade flows with China (since the rest of U.S. trade with China does include significant U.S. exports to that country). On average, each of the 4,835 stores Wal-Mart operated in the United States in fiscal 2014 (Wal-Mart Stores Inc. 2014) was responsible for the loss of about 86 U.S. jobs due to the growth of Wal-Mart’s trade deficit with China between 2001 and 2013.

Figure A

These job loss estimates are conservative because goods sold at Wal-Mart are primarily durable and nondurable consumer goods, such as furniture, apparel and textiles, toys, and sporting goods. These are particularly labor-intensive manufacturing industries and support more jobs per $1 billion of imports than more capital-intensive goods such as machine tools, motor vehicles and parts, and aircraft and parts imported by other U.S. firms.

Job losses in manufacturing account for 75.7 percent of total jobs displaced due to the growing U.S. trade deficit with China in this period (Kimball and Scott 2014, Table 3). Jobs in the manufacturing sector pay higher wages and provide better benefits than most other industries, especially for workers with less than a college education. Manufacturing also employs a greater share of such workers than other sectors (Scott 2013).

The job displacement estimates in this study are conservative. They include only the jobs directly or indirectly displaced by trade, and exclude jobs in domestic wholesale and retail trade or advertising; they also exclude re-spending employment.9 They also do not account for the fact that during the Great Recession of 2007–2009, and continuing through 2013, jobs displaced by China trade reduced wages and spending, which led to further job losses.

Further, the labor-market effects of the U.S. trade deficit with China are not limited to job loss and displacement and the associated direct wage losses. Competition with low-wage workers from less-developed countries such as China has driven down wages for workers in U.S. manufacturing and reduced the wages and bargaining power of similar, non-college-educated workers throughout the economy, as previous EPI research has shown (Bivens 2013). The affected population includes essentially all workers with less than a four-year college degree—such workers make up roughly 70 percent of the workforce, or about 100 million workers (U.S. Census Bureau 2015).

The workers affected by this job displacement include millions whose jobs were not lost but whose wages were held down because of increased labor market competition with the job losers. As earlier EPI research has shown, trade with China between 2001 and 2011 displaced 2.7 million workers, who suffered a direct loss of $37.0 billion in reduced wages alone when re-employed in non-traded industries in 2011 (Scott 2013). In addition, the nation’s 100 million non-college educated workers suffered a total loss of roughly $180 billion due to increased trade with low-wage countries. These indirect wage losses were nearly five times greater than the direct losses suffered by workers displaced by China trade, and the pool of affected workers was nearly 40 times larger (100 million non-college-educated workers versus 2.7 million displaced workers).10

Wal-Mart’s U.S. manufacturing promises

In 2013 Wal-Mart announced a plan to purchase “$250 billion in products that support the creation of American jobs” by 2023 by increasing purchases of U.S. manufactured goods (Loeb 2013, Wal-Mart 2015a).11 To date, very few actual U.S. manufacturing jobs have been created as a result of this commitment. Wal-Mart remains, by far, the top importer of ocean shipping containers in the United States with total imports of more than 775,000 container-equivalents (TEUs) in 2014, exceeding total imports by Target, the number two importer, by more than 250,000 TEUs (48.7 percent, more than total Target imports) (Journal of Commerce 2015). In addition, about two-thirds of what Wal-Mart calls American-made goods are actually groceries, which support few U.S. manufacturing jobs (Alliance for American Manufacturing 2015).

In 2015, Wal-Mart’s publicly available list of manufacturing jobs that have been or will be created in the United States includes fewer than 4,100 specific U.S. manufacturing jobs, and many of those are promised jobs that firms “will create” up to 10 years in the future (Wal-Mart 2015c). Since 2001, Wal-Mart’s growing trade deficit with China has displaced more than 100 U.S. jobs for every job that Wal-Mart has created in the United States through its “Invest in American Jobs” program.” Meanwhile, the U.S. goods trade deficit with China increased by $23.9 billion (7.5 percent) in 2014 (Scott 2015). Continuing growth in that trade deficit and in Wal-Mart imports will likely displace many times more manufacturing jobs than Wal-Mart creates in the United States over the next decade.

Conclusion

The growing goods trade deficit with China displaced 3.2 million U.S. jobs in the United States between 2001 and 2013, and it has been a prime contributor to the crisis in manufacturing employment over the past 15 years. Due to its own growing trade deficit with China, Wal-Mart alone was responsible for the loss of more than 400,000 U.S. jobs, 13.2 percent of total U.S. jobs lost in this period. The current unbalanced U.S.-China trade relationship is bad for both countries, and Wal-Mart has played a major role in creating that imbalance. The United States is piling up foreign debt, losing export capacity, and facing a more fragile macroeconomic environment.

Meanwhile, China has become dependent on the U.S. consumer market for employment generation, has suppressed the purchasing power of its own middle class with a weak currency, and, most importantly, has purchased trillions of dollars of hard-currency reserves in low-yielding, government securities and other financial assets, instead of investing these funds in public goods that could benefit Chinese consumers and workers. In order to artificially and illegally hold down the value of its currency, and thereby lower the cost of its exports to the United States and other countries, China has purchased nearly $5 trillion in U.S. Treasury bills and other government securities and private assets (IMF 2015, SWFI 2015) since it entered the WTO in 2001. It has also repressed the labor rights and wages of its workers, making its exports artificially cheap, further subsidizing its exports. Wal-Mart has aided China’s abuse of labor rights and its violations of internationally recognized norms of fair trade behavior by providing a vast and growing conduit for the distribution of artificially cheap and subsidized Chinese exports to the United States.

The U.S. relationship with China needs fundamental change: addressing the exchange rate policies and labor standards issues in the Chinese economy should be important national priorities. Wal-Mart’s huge reliance on Chinese imports illustrates that many powerful economic actors in the United States benefit from China’s unfair trading system. Wal-Mart’s gain, however, is not the country’s gain, as Wal-Mart’s imports have contributed to the ever-growing trade deficit that imperils future economic growth.

—The author thanks Josh Bivens and Ross Eisenbrey for comments; Elizabeth Glass for research assistance; and Molly McGrath, Kevin Rudiger, and Aditya Pande for data analysis.

About the author

Robert E. Scott is director of trade and manufacturing policy research at the Economic Policy Institute. He joined EPI as an international economist in 1996. Before that, he was an assistant professor with the College of Business and Management of the University of Maryland at College Park. His areas of research include international economics and trade agreements and their impacts on working people in the United States and other countries, the economic impacts of foreign investment, and the macroeconomic effects of trade and capital flows. He has a Ph.D. in economics from the University of California-Berkeley.

Endnotes

1. See Kimball and Scott (2014, 6 and “Appendix: Methodology,” 25–27) for further details.

2. This report distinguishes exports produced domestically and re-exports—which are goods produced in other countries, imported into the United States, and then re-exported to other countries, in this case to China. Re-exports do not support domestic employment because they are not produced domestically and they are excluded from the model used here. See Table 1 for information about the levels of U.S. re-exports to China in this period.

3. This model assumes that everything else is held constant; the trade and job loss estimates shown here are based on counterfactual simulations.

4. The complete list of Journal of Commerce citations for 2004–2015, covering calendar year trade between 2003 and 2014, is available on request.

5. Wal-Mart (2007) reports that it “estimates about $18 billion worth of products were purchased from China [in the fiscal year ending 2004] … about $9 billion imported from direct sources and about $9 billion from indirect.” These data are for Wal-Mart’s fiscal year ending on January 31, 2004, and were 11.9 percent of U.S. consumption imports from China in 2003, when most of those goods were imported. The following estimates all assume that Chinese imports are for Wal-Mart fiscal years (FY), and are compared with total U.S. imports in the preceding calendar years. Bianco and Zellner (2003) report that Wal-Mart imports from China totaled $12 billion (11.8 percent of U.S.-China imports) in FY 2002. Bianco (2006) reports that Wal-Mart imports from China were $22 billion in FY2005 (11.2 percent of China imports). Bianco’s estimates for FY 2004 replicate the estimate provided by Wal-Mart (2007) for its FY2004 imports from China. Based on these estimates, Table 1 assumes, conservatively, that Wal-Mart maintained a stable 11.2 percent share of U.S. goods imports from China between 2001 and 2013.

Between 2003 and 2013, overall Wal-Mart net sales in the United States rose from $208.8 billion to $336.6 billion (Wal-Mart Stores Inc. various years), rising from 7.7 percent of total U.S. personal consumption expenditures on goods in 2003 to 8.8 percent in 2013 (BEA 2015). Thus, Wal-Mart was a major and growing channel for the distribution of both domestic and imported goods in the United States in this period. Wal-Mart was also the single largest U.S. importer of goods imported from all countries via ocean container freight in 2014 (Journal of Commerce 2015), and was responsible for 12.1 percent of the total containers imported by the top 100 companies in that year. These data suggest that Wal-Mart’s share of total China imports likely increased between 2003 and 2013. Thus, the estimate of jobs displaced by Wal-Mart’s China trade in Table 1 likely represent a lower-bound estimate of actual jobs displaced.

6. Under U.S. rules, companies are allowed to petition Customs and Border Protection (CBP) to avoid disclosure of company names on bills of lading that accompany each shipment. Periodically, gaps appear in these disclosure petitions, making importing companies known for short periods of time. Comparable Wal-Mart data are available only for November 2007 and 2012 from this database.

7. This calculation is based on the ratio of total Wal-Mart international sales per square foot times an estimate of total Wal-Mart square footage in China, in various Wal-Mart fiscal years (Wal-Mart Stores Inc. 2002, 2006, 2014). Wal-Mart reports state that “over 95 percent of the merchandise in our stores in China is sourced locally” (Wal-Mart 2015b). Export estimates in this paper assume that sales per store in China were equal to the average per square foot for all Wal-Mart international stores times estimated total Wal-Mart square footage in China, and that all Wal-Mart imports into China came from the United States (the average Wal-Mart store in China was 2.3 to 2.8 times larger than the average Wal-Mart international store, based on data reported by Wal-Mart Stores Inc. (various years)). This is clearly an upper bound on total Wal-Mart exports to China because it assumes that all Wal-Mart imports into China originated in the United States, which is highly unlikely.

Wal-Mart had 6,107 international stores at the end of FY2014 and total international sales of $136.5 billion in 2014, or about $22.3 million per store. Wal-Mart had 405 stores in China, with estimated total sales of $25.6 billion in FY2014, and total imports of $1.0 billion (reported as U.S. exports in 2013 in Table 1). Assuming that all these imports were shipped from the United States, Wal-Mart was responsible for 0.9 percent of total U.S. exports to China in 2013.

8. These estimates assume that jobs supported and displaced by Wal-Mart’s China trade were directly proportional to total jobs supported and displaced by total U.S. exports to and imports from China in 2001 and 2013, as estimated by Kimball and Scott (2014).

9. Direct jobs displaced refer to jobs displaced within a given industry, such as motor vehicles and parts. Indirect jobs displaced are those displaced in industries that supply inputs to that industry, such as primary metal (e.g., steel), plastics and rubber products (e.g., tires and hoses), transportation, and information. Re-spending employment results from the spending of wages by employed workers. It is one form of a macroeconomic multiplier.

10. Author’s calculations from the estimated $1,800 wages lost by a median-wage non-college educated worker per year (Bivens 2013) times the 68.1 percent of the workforce made up of workers with less than a four-year college degree (U.S. Census Bureau 2015) times total number of U.S. workers employed (on average) in 2014 from the Bureau of Labor Statistics (BLS 2015) (yielding roughly 100 million non-college educated workers).

11. The initial Wal-Mart commitment was to purchase $50 billion in “U.S. products,” a figure that was subsequently increased to $250 billion (Loeb 2013, Walmart 2015a).

References

Alliance for American Manufacturing. 2015. “ANOTHER Walmart Made in America Infographic Needed Some Work, So We Fixed It.” Manufacture This (blog of the Alliance for American Manufacturing). July 7.

Bianco, Anthony. 2006. The Bully of Bentonville: How the High Cost of Wal-Mart’s Everyday Low Prices Is Hurting America. New York: Currency/Doubleday.

Bianco, Anthony, and Wendy Zellner. 2003. “Is Wal-Mart Too Powerful?” Business Week. October 3.

Bivens, Josh. 2013. Using Standard Models to Benchmark the Costs of Globalization for American Workers Without a College Degree. Economic Policy Institute, Briefing Paper #354.

Bureau of Economic Analysis (BEA). 2015. “Table 1.1.5 Gross Domestic Product.” (Accessed October 27).

Bureau of Labor Statistics (BLS). 2014. “Employment, Hours, and Earnings from the Current Employment Statistics Survey (National): Manufacturing Employment.” [Excel file].

Bureau of Labor Statistics (BLS). 2015. “Labor Force Statistics from the Current Population Statistics: Employment Level.” [Excel file].

Bureau of Labor Statistics, Employment Projections program (BLS–EP). 2014a. “Special Purpose Files—Employment Requirements Matrix; Chain-Weighted (2005 dollars) Real Domestic Employment Requirements Table for 2001” [Excel sheet, converted to Stata data file]. http://www.bls.gov/emp/ep_data_emp_requirements.htm

Bureau of Labor Statistics, Employment Projections program (BLS–EP). 2014b. “Special Purpose Files—Industry Output and Employment – Data for Researchers, Industry Output.” [CSV File, converted to Excel sheet and Stata data file].  http://www.bls.gov/emp/ep_data_industry_out_and_emp.htm

International Monetary Fund (IMF). 2015. International Financial Statistics. [CD-Rom, August 2015], Washington, D.C.: International Monetary Fund.

Journal of Commerce. 2015. “JOC Top 100 Importers in 2014: U.S. Foreign Trade Via Ocean Container Transport.” May 28.

Kimball, William, and Robert E. Scott. 2014. China Trade, Outsourcing and Jobs: Growing U.S. Trade Deficit with China Cost 3.2 Million Jobs between 2001 and 2013, with Job Losses in Every State. Briefing Paper #385. Washington, D.C.: Economic Policy Institute.

Loeb, Walter. 2013. “How Walmart Plans to Bring Manufacturing Back to the United States.” Forbes. November 12.

Panjiva.com. 2015. United States Trade Data (subscription data service). (Excel sheets accessed October 23).

Scott, Robert E. 2007. The Wal-Mart effect: Its Chinese imports have Displaced Nearly 200,000 U.S. Jobs. Issue Brief #235. Economic Policy Institute.

Scott, Robert E. 2013. Trading Away the Manufacturing Advantage: China Trade Drives Down U.S. Wages and Benefits and Eliminated Good Jobs for U.S. Workers. Briefing Paper #367. Economic Policy Institute.

Scott, Robert E. 2015. “Increased U.S. Trade Deficit in 2014 Warns Against Signing Trade Deal without Currency Manipulation Protection.” Economic Indicator: Trade and Globalization. Economic Policy Institute, February 5.

Sovereign Wealth Fund Institute (SWFI). 2015. “Sovereign Wealth Fund Rankings: Largest Sovereign Wealth Funds by Assets under Management.” Accessed October 27.

U.S. Census Bureau. 2013. “American Community Survey: Special Tabulation Over 45 industries, Covering 435 Congressional Districts and the District of Columbia (113th Congress Census Boundaries), Plus State and US Totals Based on ACS 2011 1-year file” [spreadsheets received March 6].

U.S. Census Bureau. 2015. “Current Population Survey, Historical Time Series, Table A-2: Percent of People 25 Years and Over Who Have Completed High School or College, by Race, Hispanic Origin and Sex: Selected Years 1940 to 2011” [Excel file].

U.S. International Trade Commission (USITC). 2014. “USITC Interactive Tariff and Trade DataWeb” [Excel files].

Wal-Mart Stores Inc. 2002. “Walmart 2002 Annual Report.”

Wal-Mart Stores Inc. 2006. “Walmart 2006 Annual Report.”

Wal-Mart Stores Inc. 2007. “Wal-mart Facts: Outsourcing.” [Html page downloaded April 3—available on request].

Wal-Mart Stores Inc. 2014. “Walmart 2014 Annual Report.”

Wal-Mart Stores Inc. 2015a. “Opportunity: US Manufacturing.”

Wal-Mart Stores Inc. 2015b. “Walmart China Factsheet.”

Wal-Mart Stores Inc. 2015c. “Walmart U.S. Manufacturing Announcements.”

Executive summary

In the past three decades, the Supreme Court has engineered a massive shift in the civil justice system that is having dire consequences for consumers and employees. The Court has enabled large corporations to force customers and employees into arbitration to adjudicate practically all types of alleged violations of countless state and federal laws designed to protect citizens against consumer fraud, unsafe products, employment discrimination, nonpayment of wages, and other forms of corporate wrongdoing. By delegating dispute resolution to arbitration, the Court now permits corporations to write the rules that will govern their relationships with their workers and customers and design the procedures used to interpret and apply those rules when disputes arise. Moreover, the Court permits corporations to couple mandatory arbitration with a ban on class actions, thereby preventing consumers or employees from joining together to challenge systemic corporate wrongdoing. As one judge opined, these trends give corporations a “get out of jail free” card for all potential transgressions. These trends are undermining decades of progress in consumer and labor rights.

This report tracks these developments and presents the most recent research findings, summarized here:

• It is common for employees to be presented with terms of employment that include both a clause that obligates them to arbitrate all disputes they might have with their employer and one that prohibits them from pursuing their claims in a class or collective action in court.
• Employees subject to mandatory arbitration can no longer sue for violations of many important employment laws, including rights to minimum wages and overtime pay, rest breaks, protections against discrimination and unjust dismissal, privacy protection, family leave, and a host of other state and federal employment rights.
• On average, employees and consumers win less often and receive much lower damages in arbitration than they do in court.
• Employers tend to win cases more often when they appear before the same arbitrator in multiple cases, indicating that they have a repeat-player advantage over employees from regular involvement in arbitration.

Introduction: The problem

Over the past 25 years, it has become increasingly commonplace for corporations to insert arbitration clauses into their contracts with customers and employees. These clauses appear to be innocuous, or even beneficial, to consumers and employees, but they pack a powerful punch. They prevent customers and employees from going to court if they have a dispute. Instead, when there is an arbitration clause, consumers and employees are required to take their complaints to a privatized, invisible, and often inferior forum in which they are less likely to prevail—and if they do, they are less likely to recover their due. Moreover, once a dispute is decided by an arbitrator, there is no effective right of appeal.

At the time of contracting, most consumers and employees do not object to having an arbitration clause in their contracts. After all, who thinks they will have a dispute with their employer or their bank? Who would risk a valuable job opportunity or an important consumer financial transaction over an obscure procedural provision? And if a dispute should arise, who wants to go to court to resolve a dispute over a faulty product or nonpayment of overtime pay? Courts are slow, excessively technical, and intimidating to most people. To hire a lawyer to handle the case would usually cost more than most disputes are worth. Yet despite the seeming benefits of arbitration, there are serious pitfalls.

As the research cited in this report shows, consumers and employees often find it more difficult to win their cases in arbitration than in court. For one thing, arbitration may not provide parties with the same extent of discovery that a court would. In certain types of cases, such as employment discrimination claims, it is practically impossible to win without the right to use extensive discovery to find out how others have been treated. In addition, while some arbitration agreements include due-process protections, others shorten statutes of limitations, alter the burdens of proof, limit the amount of time a party has to present his or her case, or otherwise impose constrictive procedural rules. In practice it is the corporation not the consumer or employee that gets to decide whether to include fairness protections in the arbitration procedure. Although a consumer or employee can try to challenge enforcement of unfair rules in court, the ability to challenge arbitration agreements has been substantially limited by the courts. Moreover, arbitrators are often reluctant to award generous damages to prevailing parties, and their awards are not appealable. On average, employees and consumers win less often and receive much lower damages in arbitration than they do in court. And in a new development, some arbitration agreements are requiring that the losing party pay all the arbitration fees, including the other side’s attorney fees. The loser-pays clauses provide a powerful deterrent to workers or consumers asserting any claims.

The trend toward increasing use of arbitration in consumer and employment relationships threatens to undermine decades of achievements in worker and consumer rights. Over the past few decades, the courts have expanded the scope of arbitration, reduced the ability of individuals to avoid arbitrating their disputes, and narrowed the possibility of obtaining judicial review. They have adopted such sweeping pro-arbitration doctrines that arbitration clauses are almost always upheld when challenged in the courts, even when individuals can show that an arbitration clause was buried in fine print or incorporated by reference to an obscure and inaccessible source. Courts also uphold clauses even when an individual can show that an arbitration system is too expensive for him or her to use. The result has been that many important employment rights can no longer be brought to a court by employees subject to mandatory arbitration. These rights include rights to minimum wages and overtime pay, rest breaks, protections against discrimination and unjust dismissal, privacy protection, family leave, and a host of other state and federal employment rights.

The most pernicious development in arbitration involves the coupling of arbitration with class-action waivers. Major corporations began to insert class-action prohibitions into arbitration clauses for consumer transactions in the late 1990s. Indeed, in 1999, the 10 major banks that issue credit cards—including American Express, Citibank, First USA, Capital One, Chase, and Discover—formed a group called “the Arbitration Coalition” to promote the use of arbitration clauses that bar class actions. This group also funded and jointly drafted amici curiae briefs to convince the Supreme Court to uphold these clauses.1 In part as a result of their efforts, courts generally permit arbitration to be coupled with prohibitions on class-action lawsuits, both for consumer and employment class actions. Thus today it is common for employees to be presented with terms of employment that include both a clause that obligates them to arbitrate all disputes they might have with their employer and one that prohibits them from pursuing their claims in a class or collective action. The legal developments have de facto stripped employees of many of the legal rights and protections that they have fought long and hard to obtain.

A quick primer on arbitration

Arbitration clauses are frequently included in the fine print that an individual is required to click through when making an online purchase.  Arbitration clauses are also often included in the company orientation and personnel materials a worker receives when beginning a new job. Because these arbitration clauses are usually buried in a sea of boilerplate, many people who are subject to them do not realize that they exist or understand their impact. These terms are called mandatory or forced arbitration because if the employee or consumer does not agree to arbitration, he or she will be denied employment or the ability to purchase the product or service. The employee or consumer has no real choice or ability to negotiate the terms of the arbitration clause.  Mandatory arbitration in the consumer and employment setting is very different from arbitration clauses in contracts between two businesses or a company and a union; in those cases, the parties have voluntarily negotiated as equals and knowingly agreed to arbitrate disputes between them.

Unlike a court proceeding, there is no one form of arbitration.  It is a term that describes a wide range of procedures that parties can design however they choose.  In practice, however, arbitration typically takes place in a conference room, where parties are seated around a large table.  Witnesses may or may not be in the room.  Parties may or may not have lawyers.  The arbitrator sits at the head of the table.  He or she is not a judge and does not wear a judicial robe or other ceremonial garb.  Rather, the arbitrator can be any person the parties have designated, although they frequently are lawyers. There is no court reporter or jury.

The arbitrator convenes the hearing and usually begins by explaining that it is an informal proceeding not subject to formal rules of evidence or procedure.  Rather, he or she explains that the arbitrator’s role is to hear any evidence that either side wants to submit and then render a binding decision.  Instead of excluding inadmissible evidence based on objections from lawyers, the arbitrator will generally hear all the evidence and then decide how much weight to give it in reaching a decision. Witnesses are sworn in by the arbitrator and the proceeding begins. During the hearing, the party who initiated the proceeding tells his or her story and presents any documents or witnesses that support it.  The other side has an opportunity to cross-examine.  Then the defending party presents its case, also subject to cross-examination.  The arbitrator may also ask questions of the witnesses.  After the close of the hearing, the arbitrator considers the evidence presented and issues an award.  Often the award takes the form of a simple statement of who won, and the amount of the recovery, if any.  Sometimes the arbitrator issues a written decision explaining the outcome.  Once the arbitrator has ruled, there is no realistic possibility for appeal.

The greater flexibility and informality of arbitration compared with court proceedings means that the parties are relying much more on the neutrality, expertise, and fairness of the arbitrator in reaching a just outcome. This can work well when two equal parties come together to design an arbitration procedure and choose an arbitrator who they both trust. However, for consumers or employees who are required to enter into mandatory arbitration with a large corporation in order to buy a product or service or to get a job, removing these formal protections leaves them vulnerable to unfair procedures and unjust outcomes.

An example of arbitration

One recent case illustrates the difficulties employees now face when trying to enforce their rights under basic employment statutes. In 2008, Stephanie Sutherland was hired by Ernst & Young to work as a “staff/assistant.”2 Her work involved relatively routine, low-level clerical work, for which she was paid a fixed salary of $55,000 per year. She routinely worked 45 to 50 hours per week, but because she was classified by her employer as exempt from overtime, she did not receive any additional compensation for overtime. By the time Ms. Sutherland was terminated in 2009, she had worked 151 hours of overtime, for which she should have been paid about $1,867, had the Fair Labor Standards Act (FLSA)3 and New York state labor laws been observed. She filed a class-action lawsuit seeking to recover overtime pay for her work in excess of 40 hours a week and for other current and former nonlicensed Staff 1 and Staff 2 employees of the firm who worked overtime.

When Ms. Sutherland was hired, she was given an offer letter that also provided that “if an employment related dispute arises between you and the firm, it will be subject to mandatory mediation/arbitration under the terms of the firm’s alternative dispute-resolution program, known as the Common Ground Program, a copy of which is attached.” The arbitration agreement specified that claims arising under state and federal labor statutes, including the federal Fair Labor Standards Act, were subject to the arbitration program. It further specified that any dispute must be brought to arbitration and not to a court, and that all disputes must be brought on an individual basis.

In her lawsuit, Ms. Sutherland attempted to enforce her rights under state and federal minimum-wage and overtime laws. The federal Fair Labor Standards Act has a provision that expressly permits lawsuits for minimum-wage and overtime violations to be brought on a collective basis. Mr. Sutherland sought to use that provision, but to do so, she had to avoid the force of the arbitration clause that said she could only bring a case on an individual basis. To this end, she argued that if she had to arbitrate her claim on an individual basis, it would cost her $160,000 in attorney fees, more than $6,000 in other costs, and more than $25,000 in expert testimony. Overall, she claimed, she would have to spend nearly $200,000 to recover less than $2,000 in unpaid overtime. She argued that because she was unemployed and had substantial college debt, she could not afford to arbitrate on an individual basis, and thus should not be subject to the arbitration provision or the class-action waiver because together they operated to deprive her of rights under the FLSA.

The lower court was sympathetic to Ms. Sutherland’s arguments, and held that the class-action waiver did not apply because it would prevent her from vindicating her rights under the Fair Labor Standards Act. However, the U.S. Court of Appeals reversed, relying on the 2013 Supreme Court decision in American Express Co. v. Italian Colors, 133 U.S. 2304, an antitrust case, in which the Supreme Court held that a class-action waiver in an arbitration clause was enforceable despite the high cost of bringing an individual action. In that case, Justice Scalia, speaking for the majority, wrote that “the fact that it is not worth the expense involved in proving a statutory remedy does not constitute the elimination of the right to pursue that remedy.” On the basis of this precedent, the Court of Appeals denied Ms. Sutherland’s right to bring her dispute to a court or arbitration on a collective basis, thereby effectively eliminating her right to overtime pay under the federal statute.

This case is not an anomaly. Rather, it reflects the current law of arbitration and illustrates the difficulties that ordinary workers face when they try to enforce their statutory employment rights. Below we map out the current law of arbitration and then present data on the extent of use of arbitration and the impact of arbitration on the ability of workers and consumers to enforce their rights.

Where did the arbitration epidemic come from?

The current arbitration epidemic is a result of judicial developments that began in the 1980s, when the U.S. Supreme Court reinterpreted a little-known federal law enacted in 1925 called the Federal Arbitration Act (FAA). The FAA provides that when a dispute involves a contract that has a written arbitration clause, a court must, upon motion, stay litigation so that the dispute can go to arbitration.4 And after an arbitration proceeding is complete, the FAA gives courts extremely limited power to review arbitral awards, no matter how erroneous they might be. Under the statute, an award can only be set aside on four grounds: it was procured by fraud, the arbitrator was biased, the arbitrator refused to hear relevant evidence, or the arbitrator exceeded his or her power as set out in the parties’ arbitration agreement. Each of these has been interpreted exceptionally narrowly. There is no provision for overturning an award based on errors of fact, contract interpretation, or law.

Initially, the drafters, commentators, and the courts assumed that the FAA applied only to a narrow range of commercial disputes—those brought in a federal court pursuant to its power to decide issues arising under federal law. However, in the 1980s the U.S. Supreme Court radically expanded the scope of the statute. Today courts interpret the statute to apply to disputes of all types, whether brought in a federal or a state court. Moreover, the Supreme Court has held that the FAA overrides any state law that runs counter to the pro-arbitration policies of the FAA. It is important to recount the path by which this transformation occurred because it shows how entrenched the current interpretation has become and how overwhelming are the obstacles to change under the statute as currently interpreted. This, in turn, explains why new congressional action is necessary.

The FAA from 1925 to the mid-1980s

Under the common law as it stood in the early 20th century, arbitration agreements were not specifically enforceable, so it was easy for a reluctant party to an arbitration agreement to avoid arbitrating a dispute. To get this changed and make arbitration agreements enforceable, the New York Chamber of Commerce and the American Bar Association’s Committee on Commerce, Trade, and Commercial Law mounted a multipronged campaign to overturn the anti-arbitration policies of the common law. They drafted and successfully enacted the New York Arbitration Act of 1920. They then turned to Congress, and drafted the 1925 Federal Arbitration Act and lobbied intensely for its enactment. Their main ally in the battle for the federal statute was the Secretary of Commerce, Herbert Hoover, who saw the bill as fitting into his larger vision of promoting business self-regulation.

The stated purpose of both the New York and the federal statutes was to make written agreements to arbitrate enforceable. The key provision of the federal law, copied from the New York statute, was Section 2, which made written agreements to arbitrate in contracts involving commerce “valid, irrevocable, and enforceable, save on such grounds as exist in law or in equity for the revocation of any contract.”6 Other sections of the statute included a mandatory stay of judicial proceedings and the requirement that courts order parties to arbitrate when disputing parties have a written agreement to arbitrate. The FAA also provided for judicial enforcement of arbitration awards and specified extremely narrow grounds for a court to refuse to do so.

The drafters, legislators, and advocates of the FAA assumed that the statute applied only to business disputes. It was drafted with an eye toward trade association arbitration, not employment or consumer disputes. Indeed, the statute contains a specific exemption for “contracts of employment.” Consistent with this understanding, between 1925 and the 1980s, courts interpreted the FAA as applying to a narrow set of cases—commercial cases involving federal law that were brought in federal courts on an independent federal ground. But in the 1980s, the U.S. Supreme Court turned the FAA upside-down through a series of surprising decisions. These decisions set in motion a major overhaul of the civil justice system. It is no exaggeration to call the Supreme Court’s arbitration decisions in the 1980s the hidden revolution of the Reagan Court.

The expanding reach of the FAA after 1985

Between 1985 and 2015, there were more than two dozen Supreme Court decisions in arbitration cases, virtually all of which expanded the scope of the FAA and restricted the ability of states to maintain laws to protect consumers and employees and the ability of individuals to resist costly and unfair arbitration systems. In light of these decisions, the ability of a party to challenge an arbitration clause on the basis of state law has shrunken to a vanishing point.

First, in the 1980s, the Supreme Court adopted a presumption in favor of arbitration to use when deciding cases involving the FAA. It ruled in Moses H. Cone Memorial Hospital v. Mercury Construction Corp., 460 U.S. 1 (1983), that when deciding whether a particular dispute comes within an arbitration clause, courts should resolve all doubts in favor of arbitration. It said that such a presumption furthered the “liberal federal policy favoring arbitration agreements, notwithstanding any state substantive or procedural policies to the contrary.” This declaration of federal policy has served as a fixture of arbitration law and provided a rationale for the extraordinary expansion of the FAA that followed.

Then, in 1984, in Southland Corp. v. Keating, 465 U.S. 1 (1984), the high court rejected the view that the FAA only applied to cases in federal courts. Rather, the Court held that the FAA also applied to disputes over contracts that were brought in state courts, so long as the dispute involved interstate commerce. The Southland decision was a major expansion of the scope of the statute. Moreover, despite direct evidence in the FAA’s legislative history to the contrary, and despite language in Section 2 of the FAA preserving the role of state law to regulate arbitration, the Supreme Court majority held that the statute preempted any state laws with which it conflicted. Thereafter, any state efforts to regulate arbitration would be subject to preemption by the FAA.7

A third development of the 1980s concerned the types of disputes that were subject to the FAA. Whereas previously the FAA had been found to apply only to contractual disputes, in 1985, in Mitsubishi Motors v. Soler Chrysler-Plymouth, 473 U.S. 614 (1985), the Supreme Court held that the FAA also compelled arbitration of statutory disputes. Mitsubishi involved a business dispute in which one party alleged a violation of antitrust laws. Two years later, in Shearson/American Express v. McMahon, 482 U.S. 220 (1987), the Supreme Court expanded on its holding to conclude that a dispute involving alleged violations of the anti-racketeering RICO statute (formally called the Racketeer Influenced and Corrupt Organizations Act) and federal securities laws was also subject to an ordinary boilerplate arbitration clause. 

The Southland decision on preemption and the Mitsubishi decision on the arbitration of statutory claims in the 1980s vastly expanded the scope of the FAA. In 1991, the Court further expanded the range of statutes whose provisions were subject to arbitration by holding, in Gilmer v. Interstate/Johnson Lane Corp. 500 U.S. 20 (1991), that an employee’s allegations that he had been subject to age discrimination in violation of civil rights laws had to be taken to arbitration. Thenceforth, most claims arising under federal statutes would be subject to arbitration. In the decades that followed, the Supreme Court further expanded the scope of the FAA in order to promote the liberal policy in favor of arbitration that it read into the 1925 statute.

At the same time, the Court repeatedly rebuffed attempts by states to enact legislation that would protect consumers and employees from unfair arbitration agreements. Beginning in the late 1980s and through the 1990s the Court struck down legislative efforts by states to protect consumers and employees from oppressive arbitration agreements. One case involved a 1985 Montana law requiring that arbitration agreements in consumer contracts appear on the first page of the contract in reasonable-sized type (Mont. Code Ann. § 27-5-114 (1993)). The purpose of the statute was to ensure that consumers knew that they were consenting to arbitration when they entered into a contractual relationship with a large corporation. In 1992, a Subway franchise owner and his wife in Montana sued, claiming that Subway had defrauded them by refusing to give them the preferred location they had been promised, causing their business to fail and their loan collateral—in this instance, their life savings—to be forfeited. Their franchise agreement with Subway had an arbitration clause that said all disputes must be arbitrated in Connecticut, far from Montana. To travel there and hire a Connecticut lawyer would have been exceedingly costly for the nearly bankrupt Casarottos. Moreover, the arbitration clause did not comply with the requirements of the Montana statutory notice provision: Rather than appearing prominently in the contract, it had been buried in small type. The Montana Supreme Court refused to enforce the arbitration clause, but the U.S. Supreme Court reversed, holding in Doctor’s Associates, Inc. v. Casarotto, 517 U.S. 681 (1996) that the law was restrictive of arbitration and therefore preempted.

The Supreme Court has also made it difficult for consumers or workers to avoid arbitration on the grounds that it would be prohibitively costly for them to take their cases to arbitration. In 2000, in Green Tree Financial Corp.-Ala. .v Randolph, 531 U.S. 79, an individual who borrowed money to purchase a mobile home and who was subsequently saddled with exorbitant finance charges sued, claiming that the lender had violated the Truth in Lending Act—a statute intended to protect consumer borrowers from misleading terms in loans. Her loan agreement had a clause requiring an arbitration tribunal that would have imposed costs far beyond her ability to pay. The Supreme Court nonetheless enforced the arbitration clause, despite acknowledging that the projected costs of the arbitration would probably preclude Ms. Randolph from bringing her case at all. The Court said that a party who opposes arbitration on the grounds that it is too expensive to proceed to arbitration had the burden of showing that the costs of arbitration would be prohibitive.

The Court has also further cut back on the ability of consumers and employees to avoid arbitration on the grounds that a contract is illegal, unconscionable, or otherwise not enforceable. One might think that if a contract is unenforceable, a party cannot be required to arbitrate under it because the arbitration clause is part of the unenforceable contract. That was the law until 1967. But in 1967 the Supreme Court held, in Prima Paint Corp. v. Flood & Conklin Mfg. Co., 388 U.S. 395, that when a party claimed that a contract it had signed was induced by fraud, that party had to assert its claim in arbitration. That is, even if the entire contract (in that case, a commercial lease) was invalid, the arbitration clause survived because, the Court found, the promise to arbitrate was separable from the rest of the contract. This holding is called the “separability doctrine.”

In 2006, the Supreme Court in Buckeye Check Cashing, Inc. v. Cardegna, 546 U.S. 440, extended the separability doctrine to illegal contracts, even though doing so meant that a party had to arbitrate an alleged violation even when the underlying contract that contained the arbitration agreement was entirely void. The only exception the Court recognized was when a party claimed that there was illegality, fraud, or some other recognized contractual defense in the arbitration clause itself.

One of the most frequently raised objections to arbitration clauses is that they are unconscionable. Unconscionability is a well-established contract-law doctrine that says that when a contract is grossly unfair in its terms and/or in the manner in which it was procured, it will not be enforced. Each state has developed its own definition of unconscionability over time. In 2010, in Rent-A-Center West v. Jackson, 561 U.S. 63, the Court expanded the separability doctrine in a way that eliminated many unconscionability challenges to arbitration clauses. In that case, the Court held that a party who claimed that the arbitration clause in his employment contract was unconscionable under his state law had to bring that claim to arbitration because the aspect of the arbitration clause he alleged was unconscionable was not the same aspect to which he objected. As Justice Stevens explained in dissent:

Prima Paint and its progeny allow a court to pluck from a potentially invalid contract a potentially valid arbitration agreement. Today the Court adds a new layer of severability—something akin to Russian nesting dolls—into the mix: Courts may now pluck from a potentially invalid arbitration agreement even narrower provisions that refer particular arbitrability disputes to an arbitrator [emphasis in original].

In addition to expanding the scope of the FAA, the Court has narrowed the standard of review of arbitral awards, thus restricting the ability of parties to appeal an arbitral decision in court. In 2008, in Hall Street Associates, L.L.C. v. Mattel, Inc., 552 U.S. 576, the Court held that parties cannot agree to have a court review the decisions of their arbitration tribunals. In that case, the parties to a commercial lease had an arbitration agreement that called for arbitration of all disputes but also specified that a court should vacate any award that was not supported by the facts or was based on an erroneous conclusion of law. Although arbitration is said to be a creature of the parties’ contract, and the parties are supposed to be able to craft arbitration systems however they like, the Supreme Court refused to enforce the parties’ agreement about the scope of review. Rather, it held that the national liberal policy favoring arbitration required limiting judicial review to the specific grounds enumerated in the FAA itself. In dicta, the Supreme Court also disparaged the long-settled principle that courts could refuse to enforce arbitration awards that were “in manifest disregard of the law.” Thus, after Hall Street, the grounds for attacking an arbitral award have become extremely narrow.

Supreme Court decisions on arbitration of employment contracts

The arbitration of employment disputes has its own history, although one that parallels the general trends described above. The FAA contains a clause that appears to exclude employment disputes from the statute’s coverage. Section 1 of the statute provides that “nothing herein contained shall apply to contracts of employment of seamen, railroad employees, or any other class of workers engaged in foreign or interstate commerce.” Despite this language, in 1991, in Gilmer v. Interstate/Johnson Lane Corp., 500 U.S. 20, the Supreme Court applied the FAA to an employment case, ruling that an employee was required to bring his age discrimination complaint to arbitration rather than to a court. The decision was ambiguous about the effect of the statutory exclusion for contracts of employment because, in that case, the arbitration clause was not in a contract between an employee and an employer, but rather was in a contract between an employee and the agency with which the employee was required to register to get the job. The Supreme Court clarified the ambiguity in 2001 in Circuit City Stores, Inc. v. Adams, 532 U.S. 105, interpreting the exemption for “contracts of employment” exceedingly narrowly. It ruled that the statute applied to all contracts of employment except those involving workers who, like seamen and railroad workers, were engaged in transportation that crossed state lines. Since then, courts have applied the FAA to numerous employment cases.

Legal issues in arbitration today: Arbitration and class-action waivers

The most controversial issue in arbitration law today grows out of the interaction between arbitration and class actions. Composite arbitration–class-action waivers have become common in contracts offered by credit card companies, banks, cell phone providers, and providers of other common services.8 They are also used with increasing frequency in employment contracts.9 Consumers and employees have challenged composite arbitration–class-action waivers on two grounds—that such composite clauses are unconscionable or that they make it impossible to vindicate statutory rights. Some state courts and lower federal courts have refused to enforce these composite clauses on both grounds, but recent decisions by the Supreme Court are calling these decisions into question.

The Supreme Court has addressed the issue of composite arbitration–class-action waivers several times in recent years. In 2011, in AT&T Mobility LLC v. Concepcion, 563 U.S. 333 (2011), it held that a California law making class-action waivers in most consumer cases unconscionable was invalid because it was preempted by the FAA. In 2013, in American Express Co. v. Italian Colors Restaurant, the Court enforced a class-action waiver even though the plaintiffs had shown that without a class action, it would be impossible for them to vindicate their legal rights. Although Italian Colors was not a labor case, it has significant ramifications for employees’ rights under the labor laws. Both these cases will be discussed below.10

Preemption, unconscionability, and class-action waivers

In 2011, in AT&T Mobility LLC v. Concepcion,11 the Supreme Court upheld a class-action waiver in a consumer contract against a challenge that the waiver was unconscionable under California state law. In that case, an AT&T customer brought a class action alleging that the company had engaged in fraudulent practices by charging sales taxes—approximately $15 per phone—to customers promised free cell phones in exchange for a two-year service contract. AT&T’s customer agreement included an arbitration clause that also banned class actions and classwide arbitration. The plaintiffs wanted to bring their case as a class action, so they argued that the class-action waiver was unconscionable.

The Ninth Circuit applied California’s three-pronged test, which determines that a class-action waiver in a consumer contract is unenforceable if (1) the agreement is a contract of adhesion—i.e., a form contract presented by a powerful party to a weaker party on a take-it-or-leave-it basis, (2) the dispute is likely to involve small amounts of damages, and (3) the party with superior bargaining power carried out a scheme to deliberately cheat large numbers of consumers out of individually small sums of money. The Ninth Circuit found all three prongs of the test satisfied, and therefore denied AT&T’s motion to compel arbitration on an individual basis.12

The Supreme Court reversed, holding that the California rule was preempted because it interfered with arbitration. Justice Scalia, writing for the majority, also disparaged the use of class arbitration. He enumerated the reasons he found class arbitration to be an unsatisfactory procedure. He stated that class arbitration would undermine the informality, efficiency, and speed that are the raison d’être for arbitration in the first place. He also stated that in class arbitration, an arbitrator would have to devise a method to afford absent class members notice, an opportunity to be heard, and a right to opt out. He then stated that class arbitration could impose great risks on defendants, who could receive a devastating judgment when numerous small claims were aggregated and yet would lose their right to interlocutory appeals or judicial review. For these reasons, he concluded that “[a]rbitration is poorly suited to the higher stakes of class litigation.”13

Some lower courts initially limited the Concepcion decision to the consumer setting and refused to extend it to the employment cases, but over time, most courts have extended it.14 Moreover, although the Concepcion case was about preemption of a specific state law, many courts have read it more broadly to disallow all unconscionability challenges to class-action waivers.15

The effective-vindication doctrine

Even though the Concepcion decision has been read to preclude most unconscionability challenges to arbitration in the employment setting, there is another line of argument some have used to invalidate waivers of the right to bring collective or class actions. That is the argument that a ban on class litigation would abrogate plaintiffs’ substantive statutory rights.

The U.S. Supreme Court has long maintained that arbitration is only appropriate when it entails no loss of substantive statutory rights. The Court first expressed this principle in 1985 in Mitsubishi Motors v. Soler Chrysler-Plymouth discussed above, in which the Court held that a party was required to arbitrate a claim arising under the Sherman Antitrust Act.16 In justifying its decision in Mitsubishi, the Court stated that arbitration could be ordered only if the litigant “may vindicate its statutory cause of action in the arbitral forum.”17 The Court further explained that “[b]y agreeing to arbitrate a statutory claim, a party does not forgo the substantive rights afforded by the statute.”18

The effective-vindication-of-substantive-rights principle is essential if courts are to justify closing the courthouse door to otherwise qualified litigants. In a number of consumer and employment cases, plaintiffs have asserted that the enforcement of class-action waivers would force litigants to forgo their substantive rights, and hence that arbitration should not be required.19 These cases were not controlled by Concepcion because, as explained above, the Concepcion decision involved a conflict between the FAA and state law, and the Court found the state law to be preempted. In contrast, the effective-vindication doctrine is of primary importance when there is a potential conflict between the FAA and a federal law.

Consumers have raised effective-vindication arguments against arbitration in cases in which it would be prohibitively expensive for them to arbitrate their claims. As we saw above, the Supreme Court has not been sympathetic to these arguments. Employees have raised effective-vindication arguments when arbitration combined with a ban on class actions would extinguish their substantive rights to engage in collective action.

Many effective-vindication cases arise under the Fair Labor Standards Act—a statute that explicitly provides that aggrieved employees can bring a “collective action.”20 Often these cases involved allegations of misclassification—for example, whether employees were improperly termed supervisors and thus improperly determined to be ineligible for overtime payments. In deciding FLSA class-action waiver cases, lower courts have to decide whether the provision in the FLSA statute for bringing “collective actions” is a procedural right or a substantive right. If it is a substantive right, then under Mitsubishi, it cannot be waived. Most courts that have considered this issue have held that the right to proceed in a collective action under the FLSA is procedural, and thus the composite arbitration and class-action waiver was required.21

While it might be reasonable to see the right to engage in a collective action to be a procedural right in the FLSA context, the same argument cannot be made concerning class-action waivers in claims arising under the National Labor Relations Act (NLRA). In the NLRA, the right to engage in collective and concerted action is the core right that the statute protects. Yet there is currently an open question as to whether a composite arbitration and class-action waiver clause would deprive workers of their substantive right to engage in collective action under the National Labor Relations Act. In D.R. Horton, Inc., 357 NLRB No. 184 (2012), the National Labor Relations Board took the position that a mandatory arbitration clause in an employment contract that required all actions to be brought on an individual basis interfered with the employee’s rights to engage in concerted activity under the labor laws. The D.R. Horton decision was overturned by the Fifth Circuit. There are several other similar cases pending in other circuits, and the issue may reach the U.S. Supreme Court.

Although the question raised by D.R. Horton has not yet been addressed by the Supreme Court, there is another recent Supreme Court case that bears ominously on the issue. In June 2013, the Supreme Court decided American Express Co. v. Italian Colors Restaurant.22 The case arose when a group of merchants brought a class action alleging that American Express (AmEx) imposed on them an illegal tying arrangement that violated the Sherman Antitrust law. Each of the merchant’s contracts with AmEx contained a clause that prohibited the merchant from bringing any dispute to a forum other than arbitration, and required that all disputes be arbitrated on an individual basis. AmEx moved to compel arbitration, and the district court granted the motion. The merchants contended that arbitration of the antitrust claim on an individual basis would cost hundreds of thousands of dollars, whereas the average recovery would be only $5,000. Hence, they claimed, without the ability to bring a class or collective action, they would lose their substantive rights. The Second Circuit agreed.23

The Second Circuit decision was overturned by the Supreme Court in June 2013. The Supreme Court upheld the class-action waiver despite irrefutable evidence that the cost of bringing an antitrust case was so high that without the ability to proceed as a class action, the case could not be brought. In doing so, Justice Scalia, writing for the majority, cast doubt on the effective-vindication-of-substantive-rights principle. He called the principle mere “dicta,” and stated that, at most, it might apply to “filing and administrative fees attached to arbitration that are so high as to make access to the forum impracticable.”24 He wrote, cryptically, “[T]he fact that it is not worth the expense involved in proving a statutory remedy does not constitute the elimination of the right to pursue that remedy.”25

Justice Kagan delivered a strong dissent in Italian Colors. The overall effect of the opinion, she explained, is that “[t]he monopolist gets to use its monopoly power to insist on a contract effectively depriving its victims of all legal recourse.”26 She argued that the effective-vindication rule was essential to prevent stronger parties from using these and other kinds of means to eviscerate statutory protections. As she explained, “The effective-vindication rule [ensures that] arbitration remains a real, not a faux, method of dispute resolution. With the rule, companies have good reason to adopt arbitral procedures that facilitate efficient and accurate handling of complaints. Without it, companies have every incentive to draft their agreements to extract backdoor waivers of statutory rights.”27

Although the Italian Colors case itself involved a dispute brought by merchants, the majority’s decision has important consequences for employment cases. By narrowing the effective-vindication doctrine, the Court has potentially undermined challenges to class-action waivers in arbitration clauses. That is, just as AT&T Mobility knocked out most unconscionability challenges to unfair arbitration agreements on preemption grounds, Italian Colors threatens to eliminate most challenges brought on the basis of the effective-vindication doctrine. And in doing so, Italian Colors suggests that the arbitration law trends may signal the destruction of the legal protection for collective action that has been at the heart of labor laws for over 60 years.28

Other current issues: Severance, interpretation of arbitration agreements, and private attorney general actions

There are two arbitration cases that will be decided by the Supreme Court this term. One, MHN Government Services, Inc. v. Zaborowski, concerns whether a court, when presented with an arbitration agreement that is unconscionable in several respects, can invalidate an entire arbitration agreement or whether it must simply sever the unconscionable elements and enforce the rest.29 The California courts have taken the position that when there are multiple unconscionable aspects to an arbitration clause, it can invalidate the clause in its entirety. This principle is important because it disincentivizes powerful parties from writing arbitration clauses with unduly harsh provisions. If a court would simply sever any unconscionable provision and enforce the rest of an arbitration clause, a powerful party might be tempted to include numerous harsh elements, knowing that even if some are deemed unenforceable, they can still require the counterparty to arbitrate. The principle is being challenged on the grounds that it is an arbitration-specific rule that disfavors arbitration and is therefore preempted by the FAA.

The other arbitration case currently before the Supreme Court involves a state court’s ability to interpret arbitration clauses. It has generally been assumed that contract law is a matter of state law, and that it is for state courts, not federal courts, to interpret contracts. In a consumer arbitration class-action waiver case called DIRECTTV, Inc. v. Imburgia, an arbitration clause provided that, notwithstanding the arbitration clause, “If, however, the law of your state would find the agreement to dispense with class arbitration procedures unenforceable, then [the entire section requiring arbitration] is unenforceable.”30 The case arose in California at a time when class-action waivers in consumer contracts of the sort in that contract were held to be unenforceable. Accordingly, the state court refused to enforce the class-action waiver. The Supreme Court has accepted review in order to determine whether the state’s own interpretation of the contract conflicts with the FAA and hence should be overturned.

Another issue that is likely to come to the Supreme Court soon involves the waiver of rights under statutes that permit individuals to enforce laws enacted for the public benefit. In 2004, California enacted a statute called the Private Attorney General Act, or PAGA law, to assist in the enforcement of its Labor Code.31 The purpose of the statute was to permit aggrieved employees to enforce the California Labor Code because the public enforcement agency lacked the resources to achieve maximum compliance with state labor laws.

In 2014, in Iskanian v. CLS Transport, a truck driver brought a class-action suit alleging failure to pay overtime and provide rest breaks.32 In that case, the employee was subject to an employment agreement that contained both an arbitration clause and a waiver of class or representative action. The California Supreme Court found that the waiver was not enforceable as applied to PAGA claims. Relying on the settled proposition that “[a]nyone may waive the advantage of a law intended solely for his benefit. But a law established for a public reason cannot be contravened by a private agreement,” it found that PAGA actions were representative actions and thus the right to bring the suit to a court could not be waived.

The lower federal courts in California have been inconsistent in their willingness to follow Iskanian and prevent a compelled waiver of employment PAGA actions. Some lower courts have done so, but many others have rejected Iskanian on the grounds that its reasoning and result are inconsistent with Concepcion.33 However, on September 30, 2015, the Ninth Circuit, in a divided opinion, affirmed the result in Iskanian and rejected a class-action waiver of PAGA claims.34 It is likely that this issue will go to the Supreme Court.

The future of arbitration law

Arbitration law is a dynamic area of law. Because the Supreme Court decisions have made arbitration the only forum available for resolving disputes in many cases, the particular details of arbitration procedures need to be resolved. Hence the number of cases continues to grow, and new issues are continually arising. However, the trends are clear: Courts will not permit states to constrict arbitration, and they will enforce arbitration agreements in all but the rarest circumstances, no matter how much advantage they give to the stronger parties. In light of these rulings, it is not surprising that the use of arbitration by private-sector businesses and employers has grown enormously.

How mandatory arbitration works

The prevalence of mandatory arbitration and class-action waivers

Arbitration in employment

Until the 1990s, arbitration in employment was almost exclusively a creature of the labor contracts of unionized workplaces. In the unionized setting, labor arbitration provides a jointly established mechanism for enforcing the provisions of collective-bargaining agreements and providing industrial justice in the workplace. Labor arbitration has been one of the most enduring and successful features of the American industrial relations system because it has served the interests of both unions and management, and both parties are equally involved in establishing and administering the system. These arbitration cases are decided by a well-established cadre of professional neutral labor arbitrators whom both parties must consider fair and neutral to be selected to decide cases. By contrast, prior to the 1990s, arbitration was only rarely used in nonunion workplaces precisely because there was no union present to play the institutional role as the bilateral partner to the employer in establishing arbitration.

The picture of arbitration as a creature of the unionized workplace started to shift as the Supreme Court began allowing statutory employment rights to be subject to arbitration agreements in its 1991 Gilmer decision, discussed above. Beyond simply providing for arbitration of statutory claims, Gilmer gave the green light to employers to require employees to sign arbitration agreements as a mandatory term and condition of employment. The case and its progeny allowed employers to unilaterally introduce arbitration procedures to cover statutory employment rights and make these procedures mandatory in the sense that the employer would refuse to hire a job applicant who would not sign the arbitration agreement.

Since 1991, arbitration has grown rapidly in nonunion workplaces.35 Many major corporations now use mandatory arbitration procedures, including Anheuser-Busch InBev, Citigroup, Darden Restaurants, Haliburton, J.C. Penney, Lowes, Oracle, Rent-A-Center, Securitas, Sysco, United Healthcare, and Wells Fargo.36 As this list suggests, mandatory arbitration now covers a wide range of employees in many different industries.

How many employees are covered by mandatory arbitration procedures? This is a surprisingly difficult question to answer, in part because of the private nature of these arbitration procedures. There is no requirement that employers who require their employees to sign mandatory arbitration agreements report this to a government agency such as the Bureau of Labor Statistics (BLS). Nor are data on the incidence of mandatory arbitration gathered in any of the official government surveys of employers. As a result, while the BLS releases detailed data annually on the extent of union membership and representation, there is no official government estimate of the extent of mandatory arbitration.

In the absence of official government statistics on the extent of mandatory arbitration, our best estimates come from academic surveys that have looked at aspects of this question. The picture they show is one of substantial growth over the 1990s and 2000s. These studies are summarized below.

A 1992 survey of corporate use of dispute-resolution procedures found that only 2.1 percent of the employers surveyed used mandatory arbitration.37 By comparison, a 1995 GAO survey of 1,448 establishments subject to Office of Federal Contract Compliance Programs (OFCCP) reporting requirements found that 7.6 percent of them had adopted mandatory arbitration procedures covering their employees.38 More recently, a 2003 survey of 291 employers in the telecommunications industry that one of us (Colvin) conducted found that 14.1 percent had adopted mandatory arbitration procedures.39 However, since the adopting employers tended to be the larger organizations, 22.7 percent of the nonunion employees in the organizations surveyed were covered by mandatory arbitration procedures. In that survey the focus was on procedures covering typical lower-level employees in the industry, such as customer service workers or technicians.

An important feature of the proliferation of mandatory arbitration procedures is that it has encompassed a broad range of lower-level employees. For example, use of mandatory arbitration is widespread in the retail industry, including in chains such as Macy’s and Target. It is also used by many restaurant chains, such as Hooters, the Olive Garden, and Waffle House. If the growth trends have continued since that 2003 survey, it is reasonable to estimate that today, a quarter or more of all employees in nonunion workplaces are subject to mandatory arbitration agreements. Put differently, it is likely that the share of American workers who are subject to employer-initiated mandatory arbitration procedures is twice the rate of the now only 11.1 percent who are union members.40

Arbitration in consumer contracts

Arbitration has become even more common in consumer transactions than in employment. The most comprehensive and recent study of the prevalence of arbitration in consumer transactions was conducted by the Consumer Financial Protection Bureau (CFPB). The Dodd–Frank Wall Street Reform and Consumer Protection Act (Dodd–Frank) that established the CFPB also mandated that it conduct a study of the use of mandatory arbitration clauses in consumer financial contracts. In addition, it empowered the CFPB to issue regulations governing the use of mandatory arbitration in these contracts based on the results of this study.

The CFPB began its study in 2012, released preliminary findings in December 2013, and issued its final report in March 2015. The CFPB’s Arbitration Study report documents that mandatory arbitration in consumer financial contracts is widespread and that mandatory arbitration clauses are included in a majority of contracts in many areas of consumer finance. The CFPB study found that credit card issuers representing 53 percent of the total credit card market include mandatory arbitration clauses. For prepaid cards, which tend to be used more by lower-income individuals, 92 percent of agreements include mandatory arbitration clauses. In student loans, 86 percent of the largest private lenders use mandatory arbitration clauses. The study found that in California and Texas over 99 percent of payday loan agreements include mandatory arbitration. Even among checking accounts, where use is lower, banks and credit cards that use mandatory arbitration represent 44 percent of insured deposits. In addition, the rate of use of mandatory arbitration in credit card agreements is likely to be temporarily depressed because the settlement of an antitrust lawsuit required four large banks to cease using mandatory arbitration for three-and-a-half years. Although these banks had not resumed using mandatory arbitration at the time of the study, which immediately followed the expiry of the settlement, if they were to resume using mandatory arbitration, this would raise the usage rate to over 90 percent for credit cards.

Regarding the content of these mandatory arbitration procedures, the most important finding of the CFPB study is that over 90 percent of them expressly prohibit class actions. Given the relatively small amounts of many consumer financial transactions and the similarity across claims, the availability of class actions is a crucial element in providing access to justice for consumer financial claims.

Another important finding of the CFPB study is that most consumers are unaware that they had entered into mandatory arbitration agreements. Three-fourths of the consumers surveyed in the study did not know that their credit card agreement included an arbitration clause. Misunderstandings were also widespread. Fewer than 7 percent of the consumers were aware that they were covered by an arbitration agreement that kept them from suing in court.

The CFPB study makes it clear that arbitration has largely displaced the civil justice system for most of the major transactions of ordinary people. A further question is whether this is a good thing. There is debate among researchers about whether consumers fare better in arbitration than in the courts. Some claim that consumers do better, and some claim the contrary.41 The evidence involves individual claims, each with its own merits.

The CFPB found that most arbitration agreements in consumer transactions include a class-action waiver. This finding reinforces a 2007 survey that found that the most prominent firms in the telecommunications, credit, and financial service industries routinely insert arbitration clauses into their contracts with consumers (76.9 percent), but rarely use them in their other commercial agreements. (6.1 percent).42 The authors of the survey opined that corporations preferred to have arbitration clauses in contracts with consumers because the clauses could be coupled with bans on class actions. In a similar vein, a survey conducted in 2009 by one of the authors of this report, Katherine Stone, found that arbitration was a mandatory term in the service agreements of all four of the largest cell phone companies, five of the eight largest cable companies, six of the nine major credit card companies, and three of four large national retail banks, and that all of the arbitration clauses were accompanied by a ban on class actions.43 Thus the detrimental impact of arbitration clauses on the ability of consumers to band together to pursue low-value claims seems undeniable. And it is only through collective efforts that consumer and employment rights can truly be protected.44

The arbitration process

Arbitration processes in general involve some form of private tribunal that adjudicates the issue in dispute. Arbitration procedures are typically a simpler, more informal version of court procedures, for example relaxing the formal rules of evidence. Underneath these generalizations, however, there is a great deal of variation in arbitration procedures. Different arbitration procedures vary considerably in their degrees of formality, similarity to court procedures, and amount of due process provided to the participants.

The arbitration agreement itself is the primary source of the rules governing the arbitration process. The parties to this private agreement are generally allowed to write into the arbitration clause whatever rules they wish to govern how disputes will be resolved. In practice this means that the corporation that chooses to make arbitration mandatory for its workers or consumers will write the rules of the procedure, and the worker or consumer will have no choice but to assent if they want to enter into an employment or consumer transaction.

Although corporations are free to craft whatever rules they wish for arbitration, many choose to incorporate by reference the rules of an established arbitration service provider. These arbitration service providers, such as the American Arbitration Association (AAA) or JAMS, will administer the arbitration, providing lists of arbitrators for the parties to select from, hearing rooms in which the arbitration can be conducted, and standard rules or procedures to be followed. Organizations such as the AAA and JAMS are important actors in the arbitration system. While they are established as private nonprofit entities, they are also well-known organizations that are subject to public pressures and provide legitimacy to the arbitration process.

In response to concerns about fairness in mandatory arbitration in the 1990s, a number of interested organizations jointly drafted a Due Process Protocol establishing basic fairness standards to be followed in arbitration. These included such important standards as the right to representation by counsel and disclosure of arbitrator conflicts of interest. However, in many other areas of procedure, such as how much discovery should be provided, the allocation of the arbitrators’ fees, and whether arbitration should be mandatory or voluntary, the Due Process Protocol did not provide clear guidance. Despite its limitations, the Due Process Protocol did provide some degree of fairness protections, which were then incorporated into the procedures of both the AAA and JAMS. In some areas these organizations’ procedures go beyond the protections provided in the Due Process Protocol. For example, whereas the protocol leaves the allocation of fees issue open, the AAA’s employment arbitration rules provide that when arbitration is mandatory (i.e., “employer promulgated”), the employer is required to pay 100 percent of the arbitrator’s fees.

The larger service providers administer many, but not all, mandatory arbitration cases. In a 2014 survey of plaintiff attorneys conducted by one of the authors of this report, Alexander Colvin, and Mark Gough of Penn State University, respondents were asked who had administered the most recent mandatory arbitration case they were involved in. The AAA was the largest service provider, administering 50 percent of cases. JAMS was second with 20 percent of cases. Another 15 percent of cases were administered by other smaller service providers, which have not been subject to the same scrutiny or research attention as AAA or JAMS. Meanwhile a further 15 percent of cases were run on an ad hoc basis with no administering agency at all. In this latter category of ad hoc cases, it is the mandatory arbitration agreement itself that alone provides the rules establishing the procedures for arbitration. While we can look at the procedures of organizations such as the AAA and JAMS as providing some degree of due process protections for employees or consumers required to arbitrate under mandatory procedures, this research suggests that there is a high degree of variation in arbitration processes. The ability of corporations to set the rules of mandatory arbitration allows them, and not the workers or consumers, to choose whether to adopt the procedures of a reputable organization with due process protections or rules that violate basic principles of fairness.

A major new feature of mandatory arbitration agreements in both the employment and consumer settings is the inclusion of waivers of class-action claims. The Supreme Court’s 2011 decision in AT&T v. Concepcion upholding the enforceability of class-action waivers is fueling the adoption of class-action waivers in arbitration agreements. A corporate-defense law firm recently estimated that the percentage of companies that include arbitration clauses with class-action waivers in their contracts grew from 16 percent in 2012 to 43 percent in 2014.45

Class-action waivers appear to be widely used in employment arbitration agreements. In a 2015 survey of 481 practicing employment arbitrators, Colvin and Gough asked the arbitrators about the provisions of the arbitration agreements in cases they had decided. The respondents reported that class-action waivers were included in 52 percent of the agreements in cases they had decided.46

Procedures provide only part of the story of how arbitration works. Under established arbitration law, if the arbitration agreement does not specify procedures to be used, then the arbitrator has plenary authority to decide how the case is conducted, with very limited grounds for review. As a consequence, the neutrality and fairness of the arbitrator is a central concern in ensuring the fairness of the arbitral process.

Colvin and Gough’s 2015 survey of practicing employment arbitrators provides some insights into who the arbitrators are. Demographic diversity is limited; 74 percent are male and 92 percent are non-Hispanic white. Just under half (49 percent) are full-time neutrals. Most of the part-time neutrals who also serve as arbitrators are practicing attorneys, and these are twice as likely to normally represent employers (61 percent) as employees (30 percent) in their legal practices. Over half (59 percent) of all full- or part-time employment arbitrators had at some point in their career worked as legal counsel representing employers, whereas 36 percent had at some point represented employees or unions. It is certainly possible and indeed often happens that an arbitrator can become a genuine neutral despite having been an advocate representing one side or the other. But it is a major concern that a substantial majority of employment arbitrators come out of backgrounds representing employers.

Outcomes of mandatory arbitration

Mandatory arbitration is not just a theoretical limitation on worker and consumer rights; it has a major practical impact on the ability of workers and consumers to pursue their legal claims and to win their cases.

Impact of arbitration on workers’ success rates and recovery amounts

Arbitration can be an effective alternative mechanism to the courts for resolving many disputes. Whereas the litigation system is often slow and costly, arbitration systems can be faster and cheaper. For example, labor arbitration has a long track record of success in unionized workplaces and is widely accepted as fair and effective by organized labor and employers. However, for workers and consumers, the question is whether mandatory arbitration unilaterally introduced by companies can be as effective as the courts at enforcing their statutory rights.

Investigating the outcomes of mandatory arbitration is challenging for researchers. Ideally we would like to conduct a double blind study in which cases are randomly assigned to either litigation or mandatory arbitration and the outcomes compared. However in practice this would be both impracticable and unethical when dealing with people with real cases. Nonetheless, even if we cannot compare randomly assigned cases under litigation with arbitration, we can get some information by looking generally at the outcomes of cases in the two forums and then analyzing similarities or differences between them.

Table 1 shows the results from a 2011 study comparing overall trial outcomes in mandatory arbitration and litigation. The comparison looks at the outcomes of 1,213 mandatory arbitration cases administered over a five-year period by the American Arbitration Association, the nation’s largest arbitration service provider. These are compared with the outcomes of studies of employment discrimination cases in the federal courts and non–civil rights employment cases in state courts.

This comparison supports the idea that arbitration can avoid some of the delays of the litigation system. Whereas the average time to trial is almost two years in either federal or state court, it is just under a year under mandatory arbitration. However, the differences in the outcomes of trials are also stark.

Employee win rates in mandatory arbitration are much lower than in either federal court or state court, with employees in mandatory arbitration winning only just about a fifth of the time (21.4 percent), which is 59 percent as often as in the federal courts and only 38 percent as often as in state courts. Differences in damages awarded are even greater, with the median or typical award in mandatory arbitration being only 21 percent of the median award in the federal courts and 43 percent of the median award in the state courts. The most comprehensive comparison comes when we look at the mean or average amount recovered in damages across all cases, including those in which the employee loses and zero damages are awarded. When we make this comparison, we find that the average outcome in mandatory arbitration is only 16 percent of that in the federal courts and 7 percent of that in state courts. While there are additional factors to consider in comparing the two systems, at the outset it is important to recognize that in a simple aggregate comparison, mandatory arbitration is massively less favorable to employees than are the courts.

Table 1

Evidence suggests that the picture has not changed much since 2011. A 2015 study of federal court employment discrimination litigation by Theodore Eisenberg found that the employee win rate has dipped in recent years to an average of only 29.7 percent.48 At the same time, another 2015 study found that the employee win rate in employment arbitration had also dipped in recent years, to an average of only 19.1 percent.49 Research has not shown whether a similar dip in employee win rates has occurred in state courts. Whatever the reason for the declining employee success rate in employment cases, these results indicate that while the gap between federal court and arbitration win rates has decreased, it is still the case that the employee win rate in arbitration is 35.7 percent lower than the employee win rate in federal court.

The data presented above only look at overall differences in outcomes. It is reasonable to wonder how much of the mandatory arbitration–litigation outcome gap is due to factors such as the type of cases reaching the trial stage. After all, most cases filed in court settle before they go to trial. So it is possible that settlement patterns could explain part of the difference between trial and arbitration outcomes.

We do not believe that settlement can explain the difference because both court cases and arbitration cases settle prior to trial or hearing in roughly similar proportions. A major study by Nielsen et al. found a 58 percent settlement rate in federal court employment-discrimination litigation,50 while recent research on mandatory arbitration found a 63 percent settlement rate across all employment cases in that forum.51 It may be that there are some differences in which cases settle, but overall it does not appear that differences in the likelihood of settlement before trial can explain the mandatory arbitration–litigation outcome gap.

Another factor that might explain some of the gap between arbitration and court outcomes is differences in pretrial disposition of cases. Many employment litigation cases are resolved through summary judgment motions. The cases that reach trial are often those that survive summary judgment and as a result represent stronger claims. Traditionally, summary judgment was not used frequently in arbitration. However, that picture is increasingly inaccurate, at least as far as mandatory employment arbitration is concerned.

In their 2014 survey, Colvin and Gough asked plaintiffs’ attorneys about their most recent employment cases in litigation and mandatory arbitration.52 In court, summary judgment motions were filed in 77 percent of the cases. However, and surprisingly, summary judgment motions were also filed in nearly half of the arbitration cases (48 percent). While this gap is not insignificant, summary judgment is more common in arbitration than often recognized. One way of looking at how much impact summary judgment has on outcomes is to compare cases across litigation and arbitration where no summary judgment motion was filed. Given the lack of any summary judgment motion in these cases, any differences between the two forums would not be the result of different use of summary judgment. Looking at this subsample of cases in arbitration and litigation where there was no summary judgment motion, Colvin and Gough found that the win rate was 32 percent lower in mandatory arbitration than in litigation. This result indicates that the gap in outcomes cannot be explained away as an effect of greater use of summary judgment motions in litigation.

It could also be argued that the extra time to reach trial might lead to higher damages in the litigation cases. In employment discrimination cases, an employee who is successful in proving discrimination is entitled to collect damages for the economic loss suffered, including back pay and front pay. This would include losses from any period of resulting unemployment, taking into account the duty to mitigate losses by searching for and accepting alternate employment. The key point is that the damages are tied to the period of unemployment caused by the discriminatory employment decision, not to the period from taking a claim to trial. But even considering the possibility of some accumulation of additional damages while awaiting trial, for example due to ongoing psychological distress, the damages under litigation so far outstrip the time to trial that they cannot be explained by the time to trial. According to Table 1, the period to trial in litigation is only about twice as long as in arbitration, whereas the average damages in federal court are nearly four times as large and in state court well over five times as large as in mandatory arbitration.

Overall, the data show a very large gap in outcomes between cases in courts and under mandatory arbitration. The most important measure of overall outcomes is the average damages across all cases, including wins and losses so as to take both win rates and damage rates into account. These are the results reported in the final row of Table 1, which indicate that plaintiffs’ overall economic outcomes are on average 6.1 times better in federal court than in mandatory arbitration ($143,497 versus $23,548) and 13.9 times better in state court than in mandatory arbitration ($328,008 versus $23,548). These are very large differences in outcomes, and attempts to explain away this gap have been largely unsuccessful.

Impact of arbitration on workers’ access to justice and ability to get attorneys

The mandatory arbitration–litigation gap in outcomes has a direct effect on the ability of individual workers to recover compensation for the injuries they have suffered. The gap also reduces the liability exposure of corporations that adopt mandatory arbitration. However, equally important, the mandatory arbitration–litigation gap has a major impact on the ability of workers to make claims in the first place.

To effectively pursue legal claims, most employees rely on finding an attorney willing to take their case. Although individuals can file claims without using an attorney, few are willing to do so, and their success rates are much lower than those who have legal representation. Nielsen et al. found that only 22.5 percent of employees filing employment discrimination cases in the federal courts were unrepresented, and just over a third of those employees eventually obtained representation by legal counsel before the case was completed.53 Some have argued that the greater simplicity and lower cost of arbitration would allow more employees to bring cases in that forum without legal representation. But in practice, we find that only 21.1 percent of employment cases in mandatory arbitration are brought by employees without legal counsel.54

How do employees obtain legal representation? Given that most consumers and low- or middle-income employees lack the financial resources to pay lawyers’ typical hourly rates, the key mechanism for financing representation is the contingency fee, where the plaintiff’s attorney receives 30–40 percent of the damages as a fee if successful, but charges no fee if the employee loses. In their study of plaintiffs’ attorneys in employment cases, Colvin and Gough found that 75 percent typically represented employees under a contingency-fee arrangement, and a further 17 percent used a hybrid arrangement that combined contingency and hourly fees.

The mandatory arbitration–litigation outcome gap has a significant and pernicious effect on the ability to obtain legal counsel under these contingency-fee arrangements. The plaintiffs’ attorney accepting employment cases knows that he or she will lose some of the cases and receive no fee for them, while receiving a fee based on the damages awarded in the successful cases. As a result, attorneys decide whether to accept a case based on their judgment about the likely outcome. But as we have seen, the average outcome is substantially lower in mandatory arbitration than it is for litigation: Damages from arbitration are 16 percent of the average damages from federal court litigation and a mere 7 percent of the average damages in state court. Thus lawyers are reluctant to take cases that are subject to mandatory arbitration. Even if arbitration cases are easier and cheaper to process, the large differences in outcomes can substantially reduce the financial incentive and ability of plaintiffs’ attorneys to accept cases brought by employees covered by mandatory arbitration.

In surveying plaintiffs’ attorneys about their likelihood of accepting potential cases, Colvin and Gough found just such an effect. Whereas on average plaintiffs’ attorneys accepted 15.8 percent of potential cases involving employees who could go to litigation, they accepted about half as many, 8.1 percent, of the potential cases of employees covered by mandatory arbitration. Thus, in addition to producing worse case outcomes than litigation, mandatory arbitration also reduces the likelihood of obtaining the legal representation that will help employees bring a claim in the first place.

Repeat player advantages in arbitration

In dispute resolution, the advantages accruing to repeat players in the system have long been a concern. A business or other organized group that frequently engages in litigation is likely to have advantages over an individual employee or consumer with no previous experience in resolving disputes.55 Repeat players have advantages because they gain familiarity with the system and how to operate effectively in it. They may also be able to lobby for changes to the system that benefit them.

One of the advantages of the traditional labor arbitration system in unionized workplaces is that both the company and the union are repeat players in the system. That means that they are both likely to be involved in future cases, have experience with past cases, and are invested in the development of a fair, effective system of dispute resolution. This balanced bilateral system with repeat players on both sides means that an arbitrator who was not a genuine neutral, and instead began to favor one side, would soon become unacceptable to the other side and not be selected for future cases. This balance between two strong repeat players is a key feature allowing private arbitration systems to function effectively.

In employment and consumer arbitration, the employer is likely to be a repeat player whereas the employee or consumer is likely to be a one-shot player.56 How then can the advantage of the repeat player be balanced? One possibility is that the legal counsel on each side serves as an effective repeat player in the system. A large sophisticated law firm representing the business could be balanced by an aggressive and sophisticated law firm representing the plaintiff. However, in practice legal representation for employees and consumers is much more fractured and of variable quality than that for businesses, which can generally afford to hire large and sophisticated corporate law firms to defend their cases. In a study of lawyers representing parties to employment arbitration, Colvin and Pike found that 76.6 percent of attorneys representing employers listed employment law as a primary practice area, compared with only 56.7 percent of attorneys representing employees.57 Furthermore, in that study, 54.6 percent of employers were represented by a law firm that handled multiple cases in the study population, whereas only 10.7 percent of employees were represented by a law firm handling multiple cases. While attorneys and law firms can provide a type of repeat player in arbitration, this result indicates that it is employers who are far more likely than employees to benefit from representation by this type of repeat player.

Do we find repeat-player advantages in the outcomes of mandatory arbitration cases? In a study of 2,802 mandatory employment arbitration cases decided between 2003 and 2014, Colvin, one of the authors of this report, and Gough looked at the relationship between numbers of cases involving the same employer and outcomes.58 They initially found that as employers were involved in more cases they tended to win more of these cases. This is not surprising and could arise from a range of factors, such as larger employers having better lawyers, more sophisticated human resource (HR) departments, and better internal systems for dealing with workplace conflicts. However, once they controlled for the number of cases involving the employer, they also found a significant effect for the number of cases in which the employer appeared before the same arbitrator. More specifically, the first time an employer appeared before an arbitrator, the employee had a 17.9 percent chance of winning, but after the employer had four cases before the same arbitrator the employee’s chance of winning dropped to 15.3 percent, and after 25 cases before the same arbitrator the employee’s chance of winning dropped to only 4.5 percent.59 The study also found that this negative effect of a long-term employer/arbitrator relationship on an employee’s chances of winning was stronger when the employee was self-represented, i.e., when there was no plaintiff lawyer available to balance the employer’s repeat-player advantage.

What could explain the repeat-player advantage of employers appearing before the same arbitrator multiple times? One possibility is that arbitrators may feel pressure to rule in favor of the employer to be selected in future cases. Although this would go against arbitrator ethical standards and is something that genuinely neutral arbitrators would consciously resist, part-time or more marginal arbitrators without well-established neutral practices could be subject to greater pressures of this nature. While it is difficult to get firm data on this issue, it is noteworthy that some arbitrators in the recent New York Times series on mandatory arbitration admitted that these pressures favor repeat players.60 Even absent any sort of arbitral bias, more sophisticated repeat-player employers may gain an advantage by getting to know particular arbitrators well and developing an understanding of their decision-making patterns and what types of arguments appeal to them. While this alternative explanation might exonerate arbitrators themselves of bias, it would nevertheless suggest that there is a bias in the system that gives employers an advantage over employees as repeat players in the system.

The use of arbitration as part of corporate HR

Mandatory arbitration in employment contracts is spreading as companies adopt it as part of their employment policies. Arbitration has become an important tool in the corporate arsenal to defend against legal claims. But it is also part of the overall human resources strategy of many companies and interacts with other HR policies. Most large companies that adopt mandatory arbitration also have internal dispute-resolution procedures to resolve organizational conflicts before they reach arbitration.

One well-known American company that has introduced this type of internal dispute-resolution procedure is Anheuser-Busch.61 Its dispute-resolution procedure includes mandatory arbitration of employment law disputes. However, the procedure begins with local management review of employee complaints, followed by mediation of any potential legal dispute before the claim proceeds to arbitration. A study of this procedure by Bales and Plowman found that the vast majority of claims are successfully resolved in these earlier stages. From 2003 to 2006, 95 percent of claims were resolved at the initial local review stage. Of the 87 claims that proceeded to mediation over this period, 72, or 83 percent, were successfully resolved at that stage. Ultimately only 15 cases, or 1 percent of the total number of complaints filed under the procedure over the four-year period, reached arbitration. Mandatory arbitration is a part of the Anheuser-Busch procedure, but the overwhelming majority of the claims brought under this system are being effectively resolved through mediation and internal dispute-resolution procedures.

Other companies have adopted more elaborate internal dispute-resolution procedures. The diversified manufacturing company TRW adopted employment arbitration after an upsurge of litigation in the early 1990s.62 However, as part of developing a more comprehensive set of internal dispute-resolution procedures, it also introduced local management complaint procedures, peer review panels (in which peers of the complainant sit on a type of workplace jury to decide complaints), and mediation. The range of dispute-resolution options provided employees with alternative ways of resolving complaints. The result was that cases were resolved early in the process, with only 72 cases reaching mediation over the first three years of the program and only three of these cases reaching arbitration. Furthermore, when cases did reach arbitration, TRW set up the procedure to be binding on the company if they lost, but not binding on the employee if the company won. As a result, employees retained the right to go to court after arbitration. TRW’s procedure is unusual in this respect, but it is a powerful example of the feasibility of resolving employment disputes through effective internal procedures without the necessity of mandatory arbitration procedures that bar employee access to the courts.

These examples show that multipronged dispute-resolution procedures can obviate the need to resort to arbitration under mandatory, binding procedures. However, under current law, the company gets to decide what procedures will be imposed on workers or consumers. The way in which this allows companies to control the legal environment under which they operate was illustrated recently by the conflicts around the ride-sharing company Uber.

There has been a great deal of attention in the courts and the media to the employment status of Uber drivers. The question is, should they be considered employees and thus entitled to the protections of employment law or, as the company alleges, should they be considered independent contractors and not entitled to any employment rights? Despite the publicity, it is less well known that, since 2013, Uber has required its drivers to sign mandatory arbitration agreements. As explained above, the arbitration clause means that a private arbitrator, not a court, will answer the crucial policy question of whether Uber drivers are employees or independent contractors. The question is important not only for Uber drivers, but for other workers in the so-called “gig economy,” who provide on-demand services coordinated by entities that maintain service platforms.

In a recent decision, a California state court judge refused to enforce Uber’s arbitration agreement on the basis that it was unconscionable.63 Among the features rendering the agreement unconscionable was that it required the driver to pay half of any arbitrator’s fees, creating a major barrier to access for low-income drivers. While the agreement did allow drivers to opt out of the arbitration clause within the first 30 days following signing on to drive for Uber, the opt-out language was buried in fine print toward the end of a long contract, leading the judge to describe it as “illusory because it was highly inconspicuous and incredibly onerous to comply with.”64 Although that judge declined to enforce the arbitration agreements used by Uber in 2013 and 2014, the case is under appeal. In practice Uber can easily redraft the mandatory arbitration agreement to correct the specific deficiencies identified by the judge, thereby making its arbitration agreement enforceable.

The Uber mandatory arbitration procedure requires that all claims be brought individually, not as class actions. As explained above, such a clause is allowable and usually enforceable, thereby preventing Uber drivers from banding together to get their legal claims and status determined, whether by an arbitrator or by a court. In the new world of combined arbitration and class-action waivers, an increasing numbers of workers and consumers are, like Uber drivers, trying to band together to protect their legal rights because to proceed solo would be prohibitively expensive. The status of the Uber class-action ban, as well as the Uber arbitration agreement, is currently on appeal.65

What can be done?

Arbitration Fairness Act

The most direct way to address mandatory arbitration would be for Congress to amend the Federal Arbitration Act to exempt consumer and employment arbitration, or to provide more protection for consumer and employee rights in arbitration. Whereas state-level legislative action to this effect would almost certainly be preempted by the FAA, legislation passed by Congress would encounter no such problem.

The most prominent effort to deal with mandatory arbitration at the federal level has been the proposed Arbitration Fairness Act (AFA). Although there have been various versions of the statute, the most recent version would amend the FAA to specify that “…no predispute arbitration agreement shall be valid or enforceable if it requires arbitration of an employment dispute, consumer dispute, antitrust dispute, or civil rights dispute.”66

If enacted, the AFA would effectively eliminate all mandatory arbitration in the employment or consumer realms, as well as in antitrust and civil rights cases. In its statement of congressional findings, the proposed AFA specifically refers to the problems of employees and consumers having little effective choice about entering mandatory arbitration agreements, the deleterious effect on the development of public law, and the lack of judicial review.67

The Arbitration Fairness Act has been repeatedly introduced in Congress, with versions proposed in 2009, 2011, and 2013. Most recently, the AFA was again proposed in 2015 by Sen. Al Franken (D-Minn.) and Rep. Hank Johnson (D-Ga.). However, it has not received a vote, and passage in the current Congress appears unlikely.

The Franken Amendment and the Fair Pay and Safe Workplaces Executive Order

In the absence of general action addressing mandatory arbitration, more progress has been achieved on specific limitations. In 2009, Franken successfully amended the annual Department of Defense Appropriations Act of 2010 to address the use of mandatory arbitration by defense contractors. The specific case motivating the amendment involved serious allegations of sexual assault, harassment, and discrimination of a female employee of Halliburton. The Franken Amendment barred any defense contractor with over $1 million in contracts from enforcing a mandatory arbitration agreement in any case involving claims under Title VII of the Civil Rights Act or tort claims relating to sexual assault or harassment. The Franken Amendment is a substantial restriction on the use of mandatory arbitration by defense contractors, but is limited to that sector and applies only to the limited set of claims specified in the amendment. For example, the amendment does not restrict use of mandatory arbitration for other statutory claims such as wage and hour claims under the Fair Labor Standards Act or any claims based on state employment statutes.

The approach taken in the Franken Amendment was subsequently extended to all federal contracts through the Fair Pay and Safe Workplaces Executive Order of 2014 (the FPSW order). The FPSW applies to all federal contractors with contracts of greater than $1 million. Similar to the Franken Amendment, it bars these contractors from enforcing mandatory arbitration agreements in claims based on Title VII or tort claims involving sexual assault or harassment. Although the FPSW is an important extension of the Franken Amendment to a broader set of employers, it suffers from the same limitation in that it applies only to a limited subset of potential employment cases. A federal contractor subject to the FPSW could continue to require its employees to sign mandatory arbitration agreements and simply decline to enforce the agreement for Title VII and the specified tort claims, while retaining the ability to use mandatory arbitration as a shield against litigation based on FLSA, state laws such as the state antidiscrimination and wage and hour statutes, or other claims. A further limitation of the FPSW order is that it may well be subject to legal challenge on the basis that it contradicts the provisions of the FAA (as a statutory measure, the Franken Amendment would not be subject to this same argument).

Consumer Financial Protection Bureau

As discussed earlier, the Consumer Financial Protection Bureau has conducted a study of mandatory arbitration in the consumer financial industry as required by the Dodd–Frank Wall Street Reform and Consumer Protection Act. In addition to mandating this study, Dodd–Frank also gives the CFPB authority to restrict or ban mandatory arbitration in consumer financial contracts. The CFPB is considering whether to ban class action waivers in mandatory arbitration agreements based on the results of its study. If it does ban the use of mandatory arbitration, this would eliminate the practice in the consumer-finance industry and have a major impact on credit card and other consumer debt contracts.

While the potential action by the CFPB could have a major salutary effect in the consumer-finance contracts field, it is important to recognize the limits of its authority. Action by the CFPB would not extend to employment contracts. Nor would it extend to other types of consumer contracts. So whereas mandatory arbitration clauses might disappear from credit card contracts, they would still exist in restaurant employee contracts, software purchase agreements, medical services contracts, Uber driver agreements, and many other agreements that affect American consumers and workers on a daily basis.

Conclusion

In the past three decades, the Supreme Court has engineered a massive shift in the civil justice system that is having dire consequences for consumers and employees. The Court has enabled large corporations to force customers and employees into arbitration to adjudicate practically all types of alleged violations, including violations of laws to prevent consumer fraud, unsafe products, employment discrimination, nonpayment of wages, and countless other state and federal laws designed to protect citizens against corporate wrongdoing. By delegating dispute resolution to arbitration, the Court now permits corporations to write the rules that will govern their relationships with their workers and customers and design the procedures used to interpret and apply those rules when disputes arise. Moreover, the Court permits corporations to couple mandatory arbitration with a ban on class actions, thereby preventing consumers or employees from joining together to challenge systemic corporate wrongdoing. As one judge opined, these trends give corporations a “get out of jail free” card for all potential transgressions. These trends are undermining decades of progress in consumer and labor rights.

It is difficult to know the practical impact of the courts’ broad delegation of dispute resolution to arbitration because arbitration is private and arbitration decisions are not generally published. However, research suggests that consumers and employees are less likely to win their cases when they are heard in arbitration, and when they do win, the amounts of damage awards are far less than would be forthcoming in a court. Moreover, there is considerable evidence that individuals who have suffered from corporate wrongdoing are deterred from bringing their claims altogether because arbitration can be too expensive and the results too risky for individual consumers or workers to undertake. The ban on class actions in particular makes it unlikely that many claims of corporate wrongdoing—particularly those that involve small sums for each in large groups of individuals—will ever be heard. As Justice Breyer opined, “Only a lunatic or a fanatic sues for $30.”68

In the few years since the Supreme Court upheld the use of class-action bans coupled with arbitration clauses, this type of composite clause has become ubiquitous in the small print governing employment, credit cards, cell phones, bank accounts, Internet providers, and countless other types of everyday transactions. The increase of arbitration clauses that require the losing party to pay the winning party’s costs, including attorney fees, will have an even more profound dampening effect on the ability of ordinary citizens to have their day in court.

What can be done to reverse these trends? Arbitration providers tout their voluntary efforts to ensure that arbitration provides due-process protections and unbiased decision-makers. However, while voluntary efforts by arbitration service providers and corporations to enhance due process in their arbitration procedures are desirable, they do not address the fundamental problem that the current law of arbitration allows the corporation to decide what type of arbitration procedure to impose on its employees or customers. Voluntary measures cannot prevent corporations that want to protect their interests—at the expense of employees and customers—from introducing provisions such as class-action waivers and loser-pay clauses that cut off access to justice. Nor can they adequately police against repeat-player bias.

Some courts and state legislatures have tried to oppose the radical change in the civil justice system, but to little avail. The Supreme Court has stated that the Federal Arbitration Act embodies a liberal federal policy in favor of arbitration, and that the act must be applied by state and federal courts. The Court repeatedly holds that the act overrides any state law or judicial doctrine that obstructs arbitration.

In addition to efforts at the state level, two federal agencies are attempting to curtail the use of arbitration by large corporations to deprive consumers and employees of their legal rights. The Consumer Financial Protection Bureau is considering a ban on class action waivers in mandatory arbitration in consumer financial transactions. By focusing on this issue, the CFPB has attracted a response from the U.S. Chamber of Commerce, which has launched a well-funded campaign to curtail the CFPB’s powers and possibly defund it altogether. At the same time, the National Labor Relations Board is attempting to curtail the use of class-action-barring arbitration agreements in the employment setting on the grounds that such agreements interfere with the core principle of labor law—employees’ rights to engage in concerted action for mutual aid and protection. However, to date, the Courts of Appeals have rejected the NLRB’s reasoning.

Despite the laudable efforts of the Consumer Financial Protection Bureau and the NLRB to protect consumers and employees from arbitrations, the legal trends suggest that agency action on this front will very likely be struck down. As a result, the only way to reverse these trends is to amend the statute itself.

The Arbitration Fairness Act currently before Congress is the best hope for stopping these trends and restoring justice to ordinary citizens. It is crucial that this act get the support of everyone who believes that consumer and employee rights are important and worth protecting.

—Katherine V.W. Stone is the Arjay and Frances Fearing Miller Distinguished Professor of Law at the UCLA School of Law. Alexander J.S. Colvin is the Martin F. Scheinman Professor of Conflict Resolution at Cornell University.

Endnotes

1. The history of the Arbitration Coalition is recounted in Ross v. American Express, 35 F. Supp. 3d 407 (S.D.N.Y. 2014) (failing to find antitrust liability despite the banks’ concerted efforts to promote class action barring arbitration clauses), aff’d sub nom, Ross v. Citigroup., ___ F.3d ___ (Case No. 14-1610, 2d. Cir., Nov. 19. 2015).

2. Sutherland v. Ernst & Young, 726 F.3d 290 (2d Cir. 2013).

3. Pub. L. No. 75-718, 52 Stat. 1060 (1938) (codified as amended at 29 U.S.C. §§ 201–219 (2006)).

4. 9 U.S.C. § 3. In order to come under the FAA, an agreement must involve commerce and include a written arbitration clause. 9 U.S.C. § 2.

5. The history and vision behind the enactment of the FAA is presented in Katherine V.W. Stone, “Rustic Justice: Community and Coercion Under the Federal Arbitration Act,” North Carolina Law Review 77: 931 (1999).

6. 9 U.S.C. Sec. 2.

7. The holding in Southland was reinforced and expanded in Perry v. Thomas in 1987 482 U.S. 483 (1987).

8. See Katherine V.W. Stone, “Procedure, Substance, and Power: Collective Litigation and Arbitration of Employment Rights,” UCLA Law Review Discourse: 61, 164 (2013).

9. See e.g., O’Conner et al. v. Uber Technologies, Inc. et al., (N.D. Calif. 2015); Mohamed v. Uber Technologies, Inc., (N.D. Calif., 2015).

10. There is another controversial issue that arises when parties are precluded from bringing a class action by virtue of an enforceable class-action waiver and they seek to arbitrate their claim on a class-wide basis. Courts agree that parties are free to specify whether their arbitration clause permits a class arbitration proceeding and if they do so, their intent will be controlling. However, in the majority of situations, an arbitration clause doesn’t say anything about the availability of class-wide arbitrations. Courts have been divided on what should be the default rule when a contract is silent about the availability of class arbitration. See, generally, Stolt-Nielsen S.A. v. AnimalFeeds International Corp., 559 U.S. 662 (2010). Courts are also divided on the predicate issue of whether a court or an arbitrator should decide whether the parties’ agreement did or did not intend to permit class arbitration.

11. 563 U.S. 333 (2011).

12. Laster v. AT&T Mobility LLC, 584 F.3d 849, 855 (9th Cir. 2009), rev’d sub nom. Concepcion, 131 S. Ct. 1740.

13. Id. at 1752.

14. See Tory v. First Premier Bank et al., No. 10-C-7326, 2011 WL 4478437 (N.D. Ill. Sept. 26, 2011); Sanchez v. Valencia Holding Co., 132 Cal. Rptr. 3d 517 (Ct. App. 2011); see also Jerett Yan, “A Lunatic’s Guide to Suing for $30: Class Action Arbitration, the Federal Arbitration Act and Unconscionability After AT&T v. Concepcion,” Berkeley Journal of Employment and Labor Law: 32, 551, 559–61 (citing cases).

15. See, e.g., Chen-Oster v. Goldman, Sachs & Co., 785 F.Supp.2d 394 (2011) (S.D.N.Y. 2011); Owen v. Bristol Care, Inc., 702 F.3d 1051 (8^th Cir., 2013).

16. Id. at 640.

17. Id. at 637.

18. Id. at 628.

19. See, e.g., Sutherland v. Ernst & Young LLP, 768 F. Supp. 2d 547, 554 (S.D.N.Y. 2011) (holding that a clause barring class actions was unenforceable because it would require plaintiffs to forgo their substantive rights). But see Banus v. Citigroup Global Mkts., Inc., 757 F. Supp. 2d 394 (S.D.N.Y. 2010) (enforcing class action waiver).

20. See 29 U.S.C. § 216(b).

21. See, e.g., Caley v. Gulfstream Aerospace Corp., 428 F.3d 1359 (11th Cir. 2005); Carter v. Countrywide Credit Industries, Inc., 362 F.3d 294 (5th Cir. 2004); Adkins v. Labor Ready, Inc., 303 F.3d 496 (4th Cir. 2002). But see Raniere v. Citigroup, 827 F. Supp. 2d 294 (D. Conn. 2003) (refusing to enforce waiver of class action in a Fair Labor Standards Act action).

22. 133 S. Ct. 594 (2013).

23. In re Am. Express Merchs.’ Litig., 554 F.3d 300, 320 (2d Cir. 2009), vacated sub nom. Am. Express Co. v. Italian Colors Rest., 130 S. Ct. 2401 (2010), remanded to sub nom. In re Am. Express Merchs.’ Litig, 634 F.3d 187 (2d Cir. 2011), aff’d on reh’g, 667 F.3d 204 (2d Cir. 2012), cert. granted sub nom. Am. Express Co., 133 S. Ct. 594.

24. Slip Op at 6-7 (Opinion of the Court).

25. Id.

26. Slip Op at 1 (J.Kagan dissenting).

27. Id. at 5.

28. See Yan, supra note 14 at 552 (noting that “[t]he unavailability of the class proceedings would have dire ramifications on employees seeking to vindicate their rights”).

29. The lower court opinion is at 936 F. Supp. 2d 1145 (ND Calif. 2013).

30. The lower court opinion is at 225 Cal.App.4th 338 (2014).

31. Stats. 2003, ch. 906, § 1

32. The lower court opinion is at 173 Cal. Rptr. 3d 289.

33. See Nanavati v. Adecco USA, Inc., 2015 WL 1738152 (N.D. Calif. 2015), summarizing California lower court PAGA waiver cases after Iskanian.

34. Sakkab v. Luxottica Retail North America, slip. op. No. 13-55184 (9th Cir., September 28, 2015), reported in National Law Journal, October 5, 2015.

35. See Alexander J.S. Colvin, “Empirical Research on Employment Arbitration: Clarity amidst the Sound and Fury?” Employee Rights and Employment Policy Journal, 11(2): 405–447 (2008).

36. Although there is no public registry listing all the companies that require mandatory arbitration of their employees, the disclosure statements that arbitration service providers are required to make public include the names of the companies involved. The most complete and extensive case disclosures currently available are those provided by the American Arbitration Association: https://www.adr.org/aaa/faces/aoe/gc/consumer.

37. See Peter Feuille and Denise R. Chachere, “Looking Fair and Being Fair: Remedial Voice Procedures in Nonunion Workplaces.” Journal of Management 21: 27–36 at 31 (1995).

38. Although the GAO survey initially found that 9.9 percent of respondents had adopted mandatory arbitration, when the agency followed up with the respondents to obtain copies of their procedures, a number indicated that they did not actually have mandatory employment arbitration and had made errors in their responses. When that correction is made, only 7.6 percent of the respondents remain as mandatory arbitration adopters. See General Accounting Office, Employment Discrimination: Most Private Sector Employers Use Alternative Dispute Resolution, GAO/HEHS 95-150, (1995).

39. Alexander J.S. Colvin, “Empirical Research on Employment Arbitration: Clarity Amidst the Sound and Fury?” Employee Rights and Employment Policy Journal: 11(2): 405–447 (2008).

40. “Union Members–2014,” Bureau of Labor Statistics (2015).

41. See generally, Sarah Randolph Cole, “Of Babies and Bathwater: The Arbitration Fairness Act and the Supreme Court’s Recent Arbitration Jurisprudence.” Houston Law Review, 48(3): 457–506 at 471–476 (2011).

42. Theodore Eisenberg, Geoffrey Miller, and Emily Sherwin, “Arbitration’s Summer Soldiers,” University of Michigan Journal of Law Reform 41(4): 871–896 at 886 (2008).

43. Katherine V.W. Stone, “Signing Away Our Rights,” The American Prospect (April 2011) 20.

44. Katherine V.W. Stone, “Procedure, Substance, and Power: Collective Litigation and Arbitration of Employment Rights,” UCLA Law Review Discourse 61: 164–181 (2013).

45. Carlton Fields Jorden Burt, The 2015 Carlton Fields Jorden Burt Class Action Survey: Best Practices in Reducing Risk and Managing Cost in Class Action Litigation.

46. Alexander J.S. Colvin and Mark D. Gough, Understanding the Professional Practices and Decision-Making of Employment Arbitrators. Report to the National Academy of Arbitrators Research and Education Fund (2015).

47. Notes: All damage amounts are converted to 2005 dollar amounts to facilitate comparison.

The “Colvin” dataset draws on all employment arbitration cases based on employer-promulgated procedures administered by the American Arbitration Association from January 1, 2003, to December 31, 2007. Data are assembled by Colvin from reports filed by the AAA under California Code arbitration service provider reporting requirements. Alexander J.S. Colvin, “An Empirical Study of Employment Arbitration: Case Outcomes and Processes.” Journal of Empirical Legal Studies 8(1): 1–23 at 5 (2011).

The “Eisenberg and Hill” litigation statistics are reported in Eisenberg, Theodore, and Elizabeth Hill “Arbitration and Litigation of Employment Claims: An Empirical Comparison.” Dispute Resolution Journal 58(4): 44–55 (2003).

48. Theodore Eisenberg, “Four Decades of Federal Civil Rights Litigation.” Journal of Empirical Legal Studies 12: 4–28 (2015).

49. Alexander J.S. Colvin and Mark D. Gough, “Individual Employment Rights Arbitration in the United States: Actors and Outcomes.” ILR Review 68(5): 1019–1042 (2015).

50. Laura Beth Nielsen, Robert L. Nelson and Ryon Lancaster, “Individual Justice or Collective Legal Mobilization? Employment Discrimination Litigation in the Post Civil Rights United States,” Journal of Empirical Legal Studies 7(2): 175–201 (2010).

51. Colvin and Gough (2015), supra note 49.

52. Alexander J.S. Colvin and Mark D. Gough, Comparing Mandatory Arbitration and Litigation: Access, Process, and Outcomes. Report to the Robert L. Habush Endowment of the American Association for Justice (2014).

53. Nielsen et al. (2010) supra at 200.

54. Colvin and Gough (2015) supra at 1030.

55. Marc Galanter, “Why the ‘Haves’ Come Out Ahead: Speculations on the Limits of Legal Change.” Law and Society Review 9(1): 95–160 (1974).

56. The problem of repeat-player effects in mandatory arbitration was first raised in a series of studies by Lisa Blomgrem Amsler (formerly Bingham), e.g., Lisa B. Bingham, “Employment Arbitration: The Repeat Player Effect,” Employee Rights and Employment Policy Journal 1(1): 1–38 (1997); Lisa B. Bingham, “On Repeat Players, Adhesive Contracts, and the Use of Statistics in Judicial Review of Employment Arbitration Awards.” McGeorge Law Review 29(2): 223–259 (1998).

57. See Alexander J.S. Colvin and Kelly Pike, “Saturns and Rickshaws Revisited: What Kind of Employment Arbitration System has Developed?” Ohio State Journal on Dispute Resolution 29(1): 59–83 at 70 (2014).

58. See Colvin and Gough (2015), supra note 69.

59. Ibid at 1033–34.

60. Jessica Silver-Greenberg and Michael Corkery, “In Arbitration, a ‘Privatization of the Justice System,’” New York Times, Nov. 1, 2015, p. A1.

61. See: Richard A. Bales and Jason N.W. Plowman, “Compulsory Arbitration as Part of a Broader Dispute Resolution Process: The Anheuser-Busch Example” Hofstra Labor & Employment Law Journal, 26(1): 1–32 (2008).

62. Alexander J.S. Colvin, “Adoption and Use of Dispute Resolution Procedures in the Nonunion Workplace.” Advances in Industrial & Labor Relations, 13: 80–94 (2004).

63. Joel Rosenblatt, “Uber Loses Bid to Force Arbitration on California Driver,” BloombergBusiness, Sept. 21, 2015.

64. Carolyn Said, “Judge Rejects Uber Forced-Arbitration Clause; 2 Cases Proceed,” San Francisco Chronicle, June 10, 2015.

65. O’Connor v. Uber Technologies, Inc. et al.: Mohamed v. Uber Technologies

66. Section 3(a), Proposed “Arbitration Fairness Act of 2015,” H.R. 2087.

67. Section 2, AFA.

68. Justice Breyer, dissenting in AT&T Mobility LLC v. Concepcion.