Report | Education

Labor market left behind

Briefing Paper #142

Download PDF

Labor market left behind

Evidence shows that post-recession economy has not turned into a recovery for workers

by Jared Bernstein and Lawrence Mishel

Although the recent recession was officially declared over as of November 2001, on Labor Day 2003 the job market remains decidedly weak. Unemployment is high and, instead of coming down in the nascent recovery, it has climbed from 5.6% at the recession’s end to 6.2% in July 2003 (the most recent data available). Tracking the nation’s payrolls reveals the worst hiring slump since the Great Depression. And the weak labor market is not just a problem for those without jobs—wages have been growing more slowly for most workers and even falling in real terms for some.

How could it be that the nation’s economy is supposedly in recovery yet the job market is much weaker now than when the recession ended? The explanation has something to do with the criteria used to determine when a recession ends (see Appendix A), but the main point is that, although the economy is expanding, it is doing so at too slow a rate to quickly lower unemployment or generate the needed job growth.

A close look at current labor market conditions shows:

• In terms of employment growth, the current recovery is the worst on record since the Bureau of Labor Statistics began tracking employment in 1939. Employment is down over one million since the recovery began. Adding in the job losses from the actual recession, payrolls are down by 2.7 million overall—3.2 million in the private sector—making this the worst hiring slump since the Great Depression.

• Even if predictions of a strong rebound in growth in the second half of 2003 are accurate, unemployment will remain close to 6.0% through the rest of this year and most of 2004.

• Since the recovery began, the unemployment rate has gone up by 0.6 percentage points. The increase for African Americans has been 1.3 percentage points.

• There is only one other recovery since World War II in which unemployment had yet to fall this many months after the recession’s end.

• The downturn and weak recovery has been broad-based demographically, and the decline in employment opportunities has actually been greater for college graduates than for high-school dropouts.

• The decline in the growth of the labor force has partially suppressed the growth of unemployment, as an estimated two million persons have given up the search for work and are thus not counted among active jobseekers. Had the labor force’s growth kept pace, unemployment would likely be closer to 7.0%.

• While the unemployment rate was 6.2% in July 2003 (most recent available data), a broader indicator of labor market distress—underemployment—was in double digits, at 10.2%

• After growing at an inflation-adjusted rate of 2% per year through 2001, the real wages of the typical (i.e., median) worker stopped growing entirely in 2002.

• Over the past year, real wages have been falling about 1% for low- and high-wage workers, and wages have been stagnant for others.

• Since the median female wage has remained flat while that of males has fallen, the gender wage gap has closed recently, with female/male medians reaching 81.3%, a historical high.

While these sobering facts characterize the current labor market, forecasters remain confident that economic performance will be much improved in the second half of this year. In large part, this outlook stems from the considerable fiscal boost underway both in terms of tax cuts and spending. As for the spending boost, note that 70% of the 2.4% growth of GDP in the second quarter is attributable to increases in defense spending. At the same time, the tax cuts will be pumping close to $100 billion into the economy over the course of the year, all in the second half.1

Many economists agree that, from the perspective of stimulating growth, these tax cuts were poorly structured, with too much emphasis on reducing taxes on the wealthy and too little emphasis on state fiscal relief.2 Still, the cuts in tandem with the new spending are of such a magnitude that there is little doubt that these fiscal stimuli will generate faster growth, which will lead to some job creation and even some shrinking of the unemployment rate. But the data presented in this report, particularly recent wage trends, raise the question of whether this expected faster growth is sustainable. Most families depend on their paychecks, and with employment down and wage growth slowing, a self-sustaining, robust recovery is not just around the corner. While some of the trends examined here will likely improve over the next few quarters, it will be a long wait before the economy returns to the tight job market and broad-based wage gains that workers benefited from just a few years ago at the end of the 1990s.

Job growth

The worst job slump since the Depression

Figures 1A and 1B examine the growth or loss in private sector jobs over two relevant time periods: 28 months out from the end of the prior business cycle’s peak, and 20 months into a recovery. These time periods are chosen because we are now 28 months out from when the most recent recession began in March 2001 and 20 months out from when it ended (November 2001). The focus here is on private sector employment, excluding government, because it is the most useful for measuring the impact of economic conditions on job creation; government employment is less sensitive to the business cycle. Note, however, that total employment figures still present a similar picture.

Figure 1A

Figure 1B

Figure 1A reveals that almost two-and-a-half years after the recession began, private sector employment is down 2.9%, or 3.2 million jobs. In prior recessions/recoveries, employment had generally recovered by this point or had grown significantly, as in the two earliest recessions shown.3 The exception is the early 1990s recession, which was also followed by a jobless recovery. But job losses in this recession/recovery far surpass the losses in the early 1990s.

Figure 1A reveals the weakness of the current recovery, which began in November 2001. Figure 1B shows that, as would be expected, every post-WWII recovery until this one had added jobs by this point, including the early 1990s recovery, which also took a long while to get underway. Yet the nation’s private sector payrolls are now 1.1% lower (down 1.2 million jobs) than when the recession ended, making this the weakest employment recovery on record (total payrolls are down 0.8%, or one million jobs).

Job losses in a few sectors have been particularly notable. Manufacturing has been hit the hardest: employment is down by 2.3 million since the recession began—a 14% decline, with about half of those losses coming during the recovery. As a share of total employment, manufacturing is down from 13% to 11% since the recession began, a remarkable shift in the composition of employment over such a short period. Employment in the information sector has fallen almost as much in percentage terms, down 12% (439,000) since March 2001.

Unemployment past and present

A lower level, but a larger increase

While the unemployment rate has not been as high as in past weak labor markets, it has at this point risen more than in many comparable periods in the past. This stems largely from the fact that the jobless rate started the recession from its lowest level in over three decades, hitting 4.0%, on average, in 2000. Since late 2000, the rate ha
s generally crept up, despite the official ending of the recession in late 2001. When the recession began, 6.1 million persons were unemployed; by July 2003, that number had climbed to 9.1 million.

Figure 2 shows the jobless rate for various groups in March 2001 (the peak of the last business cycle), in November 2001 (the month the recession ended), and in July 2003. Overall, unemployment rose 1.4 points over the recession, from 4.2% to 5.6%, and then continued its upward trend through July of this year. This pattern prevailed for each group, with minority unemployment growing faster than the average since the recovery began (African American unemployment is up 1.3 points since November 2001, compared to 0.6 points for all workers).

Figure 2

While it is widely held that the labor market recovery lags that of the overall economy, how typical is this current pattern of rising unemployment in a recovery? Figure 3 shows that it is, in fact, quite unusual at this point in a post-recession period for unemployment to be higher than when the expansion began. Figure 3 shows the percentage-point changes in unemployment over the first 20 months in each post-war recovery (as in Figure 1B, the choice of 20 months is driven by the fact that July 2003 is 20 months into the expansion). While there are a few cases in which unemployment was only slightly lower by this time relative to when the recession ended, only in this and the last recovery was it still higher 20 months out.

Figure 3

Table 1 compares the increase in unemployment over the past four recessions/recoveries by race and by age. Each value in the table is the percentage-point increase in unemployment from the peak of the business cycle to nine quarters after the peak. (It is preferable to switch to quarterly data here since this measure is less volatile for smaller sub-groups; thus, the most recent data point is the second quarter of 2003.) For whites, the increase in unemployment is about the same as over the last jobless recovery, but for African Americans, the increase of 3.2 percentage points is greater than that of the past two similar periods. Hispanic unemployment is up less in this period relative to past ones, but as shown below, that is partly due to a large decline in the labor force participation of Hispanics, and is thus not necessarily indicative of a stronger labor market for this group.

Table 1

Relative to the past two recessions/recoveries, increases in unemployment have been greater in this period for each age group, especially for younger workers age 16-24.

The persistent weakness in the labor market—with job growth contracting for almost two-and-a-half years—has led to an increase in long-term unemployment, as more job seekers compete for ever fewer job slots.4 The absence of job creation is making the job search for the unemployed much more difficult than July’s 6.2% unemployment rate would suggest.

Figure 4 compares the trend in average weeks spent job hunting to that of the unemployment rate. The figure shows that long-term joblessness is at levels associated with much higher rates of unemployment, leading to a historically large gap between the two levels. This gap largely evolved over the last 12 months, as unemployment has hovered around 6% while the average spell of unemployment has lengthened by 2.7 weeks (from 16.6 to 19.3).

Figure 4

Since April 2003, unemployment spells have lasted an average of 19.5 weeks. The last time unemployment lengths were above 19 weeks for at least four months was in late 1983 when the unemployment rate was around 9%. Thus, these historically high levels of long-term joblessness are another reminder that the relatively low rate of unemployment masks the full extent of labor market distress.

Employment and unemployment by education level

A broad-based recession

Unlike most past downturns, this recession and jobless recovery have been broadly felt by working families throughout the income scale. The burden of past recessions has been most acute for workers in blue-collar or low-end service jobs, while white-collar workers were typically more insulated from the vicissitudes of market forces. But in this recession, although blue-collar workers in manufacturing have been hit hard by job losses, those with higher levels of education have also been affected.

Figure 5 shows the trend in employment rates at two ends of the education spectrum: for those with less than high school (LTHS) educations and for those who are college graduates.5 Employment rates—the ratio of employment to population—are a standard measure of labor demand. (Note that, due to the great disparity in the levels of their employment rates, a separate axis is used for each educational group.)

Figure 5

Interestingly, in terms of employment rates, high school dropouts did much worse in the past two downturns for which we have data relative to the most recent one. After climbing steeply in the latter 1990s, when the quickly growing economy yielded strong job growth for high school dropouts, their employment rates flattened but did not fall much in the recession/recovery. College graduates, on the other hand, experienced a clear and sizable decline in demand in this recession, and one that was uniquely negative in the 24-year period covered by the available data.

Figure 6 presents the changes in employment rates by education level for the three most recent recessions/recoveries, nine quarters after the one in which the business cycle peaked. The difference in the pattern of employment rate changes in this relative to these two earlier periods is striking. In the recessions/recoveries of the early 1980s and early 1990s, the weak job market was most costly in terms of job opportunities to those with the least education, and it was less severe for those with more years of schooling. In the early 1980s, for example, employment rates fell by 1.8 percentage points for high-school dropouts, but fell progressively less for those with higher education levels. Though the losses were greater across the board in the early 1990s, a similar pattern prevailed, with the smallest losses among the most highly educated.

Figure 6

But in this recession/recovery, the pattern reverses quite sharply and the largest losses are for college graduates, whose employment rates fell by 2.3 percentage points, while the rates of those with less education fell less. In fact, employment rates for dropouts showed little change (-0.4 points). By this measure, job opportunities have weakened the most among college graduates.

Figure 7 follows the same approach but looks at changes in unemployment rates. In unemployment rates there is a very clear downward pattern by education level in the first two recession/recovery periods and a more equal pattern in the last period. While unemployment fell the least among college graduates in the current recession/recovery period, the rise in their jobless rates—1.4 points—was larger than in either of the past two periods. The increase in the jobless rate of the least educated was significantly less.

Figure 7

One explanation for these trends relates to the business sectors hit hardest by the recession. The seeds for this recession/weak recovery were planted in some sectors&mda
sh;primarily information technology and financial services—that employ many highly educated workers. Of course, other sectors, such as manufacturing, have also been severely affected, but this is much more the norm in downturns. The broad-based nature of labor market distress in this recession is somewhat unique.

Missing labor force

Fewer persons in the labor force suppress the rise in unemployment

One reason unemployment rose less in this relative to prior recessions was that the labor force grew more slowly. Given the weakness of the labor market, fewer persons than usual sought work, and this kept unemployment from rising as much as it might have had labor force growth kept pace with population growth.

Figure 8 shows the gap between the actual labor force growth and what would have prevailed if labor force growth had stayed on trend. To represent the gap, the change in labor force participation rates over the last five downturns/recoveries are averaged, going 28 months out from the business cycle peak (omitting, once again, the first dip in the early 1980s). On average, excluding the current period, participation rates were unchanged over prior 28-month periods following a peak. Between March 2001 and July 2003, however, labor participation rates fell by 0.9 percentage points. The line in the graph simulates what the participation rates would look like if they had behaved as they did in past downturns (i.e., what participation rates would look like if they had held constant at the March 2001 level).

Figure 8

Had this historical pattern prevailed, by July 2003 the labor force would have been larger by two million persons. If half of these missing labor force members were unable to find jobs (a conservative estimate), the unemployment rate in that month would have been 6.8% instead of 6.2%.

Figure 9 shows the average change in labor force participation rates by race and gender over prior periods compared to the current change (quarterly data are used here to reduce the volatility in measures for smaller groups). In the quarterly data, the decline in the total labor force nine quarters out from the last cyclical peak was 0.8 points, compared to 0.1 point on average in five prior recessions/recoveries. The decline in the male participation rate was similar to that in prior periods, but other groups all show much larger drops in the most recent nine quarters. The participation rate for women has historically risen by 0.8 points, but in this period, it fell by 0.3 points. Each racial group also shows lower participation rates, especially Hispanics.6

Figure 9

The contraction in the labor force has two important implications. First, as noted, it lessened the growth of the unemployment rate, since some of the jobless never actually sought work and were thus not counted among the unemployed. This has the effect of holding down the rise in unemployment that would have occurred had the labor force continued to expand, as suggested by Figure 8. Second, once these persons come back into the labor market, competition for jobs will become more heated. If this return happens before the economy begins to see solid job growth, the unemployment rate might increase, or at least be stubborn to fall.


A more comprehensive measure

When the labor market is this weak, the unemployment rate does not provide a full picture of the extent of labor market distress. Along with active job seekers, there are others affected by the downturn who go uncounted in the official rate. In July 2003, for example, there were 4.6 million part-time workers who wanted full-time jobs but were unable to find them. There were also 470,000 discouraged workers (persons who sought work within the past year but have given up due to lack of prospects), and another, larger group (just below one million) who face some barrier between them and the labor force, such as child care or transportation constraints.

The underemployment rate, shown in Figure 10, includes these groups of persons.7 Two points about the trend are notable. First, the underemployment rate is of course significantly higher than the unemployment rate (also shown in the figure). But less obvious is the fact that the gap between the two expands in hard times. When unemployment was at 3.9% in December 2000, underemployment was 7.0%, a gap of 3.1 points. But in July 2003, when underemployment was 10.2%, unemployment was 6.2%, a gap of 4.0 points. Thus, weak labor markets led to a greater rise in this more comprehensive measure of labor market distress.

Figure 10

Figure 11 shows the underemployment rate by race and ethnicity in the first quarter of the recession and the first quarter of this year ( data for the second quarter are not seasonally adjusted, so the comparison is limited to comparable quarters). Unemployment is much higher for minorities, and their rates have grown faster than whites as well. Black underemployment was 16.4% in 2003:q1, 5.5 percentage points above the black unemployment rate (nonseasonally adjusted) in that quarter.

Figure 11

Wage trends

Jobless recovery leads to much slower earnings growth

A central concern about the protracted jobless recovery is its impact on wage growth. The increase in unemployment and the lack of job creation has dramatically altered the beneficial dynamic that prevailed in the latter half of the 1990s, when tight labor markets and faster productivity growth fueled broad-based wage gains. In those years, wages and incomes were on the rise throughout the wage distribution; in other words, increases did not solely go to the most highly paid, but they went to low- and middle-wage workers as well.

Figure 12 reveals that, while real wages rose through 2001, by the beginning of 2002 the absence of tight labor markets began to take a toll on wage growth. The figure shows inflation-adjusted median hourly wages for men, women, and both combined.8

Figure 12

For established reasons, wage series tend to develop their own momentum. Employers are often reluctant to cut the nominal wages of their workforce, even in bad times, and the decline in new wage offers takes a while to show up in the aggregate data. Thus, it takes some time, even in a weak labor market, for wage growth to reverse course. Figure 12 shows that median wages rose throughout the recession, growing 2% annually between the first quarter of 1998 and the end of the recession in the fourth quarter of 2001 (by way of comparison, at a similar point in the late 1980s, real median wages were flat for women and falling for men). In the recovery, however, the overall median wage has been flat, with real declines for men and little gain for women.

Figure 13 compares real wage growth at different percentiles in the wage scale over the first halves of the past three years, so as to examine whether this same pattern can be observed for workers at different wage levels. The reversal of real hourly wage gains between 2002 and 2003 is particularly clear in these data, as real wages fell about 1% for those in the bottom 30% and those at the 90th percentile and were stagnant elsewhere in the wage scale.9 So, instead of real wages growing 2-3%, as in 2002, they were stagnant or falling, representing a 3-4% reversal.

Figure 13

Table 2 displays wage trends since 1989 by educat
ion level and gender. Such data help reveal the impact of the recession/recovery on the wage growth of workers at different education levels. The data in this table provide further evidence of the broad-based impact of the weak labor market as wage losses are not concentrated among any particular education group.

Table 2

The 1990s business cycle also got off to a weak start, as wages fell for non-college-educated men and were stagnant for similarly educated women. Note that such a pattern led to increased wage inequality as the pay of more highly educated workers increases relative to those with less education. The tight labor market of the latter 1990s reversed these negative trends, as the low unemployment rate led to gains throughout the education scale (though gains were still faster among college graduates).

Despite the increase in unemployment as the recovery ended, these positive trends persisted through 2001. However, data through the first half of this year reveal recent real losses among most male workers—only those with advanced degrees escaped the trend, and their wage growth slowed by half, from 1.6% per year to 0.8%. Women’s wage growth also slowed for each educational group. Interestingly, it is difficult to see an obvious pattern of wage losses by education level. For example, wages for men with college education fell a bit faster than wages for high-school-educated men; for women, wage deceleration was somewhat uniform regardless of education attainment. As these trends only recently reversed course, however, some time is needed to determine how wage trends will evolve in the recovery.

Wage data by gender show men’s earnings falling relative to that of women. Figure 14—the ratio of female-to-male median hourly wages—shows that this pattern has led to a narrowing of the gender wage gap. In fact, the gender gap at the end of this data series (2003:q2) was 81.3%, the narrowest gap since these data first began to be collected in 1979. Of course, as Figure 12 shows, the closing of the gap is largely attributable to male wage losses, not female wage gains. The explanation for the different gender wage trends is likely to be found in specific sectors, such as manufacturing, that are both male dominated and most adversely affected by the downturn. Conversely, some disproportionately female sectors of the labor market, such as health care, have held up relatively well thus far.

Figure 14

While these wage data all come from the same source—the Bureau of Labor Statistics (BLS) Current Population Survey—virtually all of the government’s wage series show the same result: weak labor markets leading to slower wage growth. Average nominal hourly wages in the BLS monthly data on production, non-supervisory workers—a group that saw wage growth of about 4% from the latter 1990s through the first half of 2001—has slowed to an annual rate of about 3%. Weekly earnings from that data series, which combine the impact of slower wage growth and fewer hours per week, grew at an annual rate of just 1.3% in the second quarter of 2003, compared to an average of 3.5% in 2000. Another data series—the private sector portion of the Employment Cost Index—shows wages and salaries grew 2.6% from 2002:q2 to 2003:q2, the slowest growth rate on record since data collection first began for this series in 1976.

Of course, the fact that inflation has been low recently means that even these low growth rates of nominal wages represent real gains for some workers. Nevertheless, both the living standards of working families and the strength of the recovery are threatened by this reversal in the strong wage growth that characterized the latter half of the 1990s.

The future course of unemployment

Even optimistic forecasts show unemployment well above 5.0%

Figure 15 shows the results of three unemployment forecasts: the Blue Chip Consensus (an average of 53 independent forecasters’ predictions), the Blue Chip Top Ten (an average of the 10 most pessimistic of the Blue Chip forecasts), and the Bush Administration’s Office of Management and Budget’s (OMB) most recent forecast. Note that each of these includes forecasters’ beliefs about the impact of the $350 billion tax cut passed earlier this year.

Figure 15

The Blue Chip Consensus calls for unemployment to remain at or above 6% for the rest of this year, despite the fact that these same forecasters predict the economy to grow a good deal faster in the second half of the year due in part to the stimulus from the tax cut and faster government spending.

There are three reasons why unemployment is likely to remain in the 6% range this year and next. First, the economy must grow by at least 3.5% for the unemployment rate to come down. This important benchmark—the sum of productivity (trending at about 2.5%) and labor force growth (which is getting back to around 1%)—is the tipping point, that is, the growth rate at which unemployment would neither rise nor fall. In other words, 3.5% growth keeps unemployment at the status quo. With faster growth, the economy can expect unemployment to fall; this is what drives the declines predicted by the Blue Chip Consensus and OMB forecasts for next year. But even based on optimistic growth forecasts for the second half of 2003, unemployment will remain high.

Second, even these small gains come with a delay. It takes a few quarters (three to six months) before growth fully filters through to the labor market (that is why unemployment is often called a lagging indicator).

Third, those who have dropped out of the labor force are likely to return when either job creation finally begins or when those who have been sitting out the weak labor market can no longer afford to do so. If the rate of job growth is not of the magnitude necessary to provide job slots for the new and returning entrants, unemployment will either rise further or fall more slowly than would otherwise be the case.

Finally, we include the more pessimistic Blue Chip predictions because forecasters have consistently overpredicted growth (and thus, underestimated the unemployment levels) for at least the past year. These less optimistic projections may offer a more realistic assessment of where the unemployment rate is likely to stand by year’s end.


Today’s weak labor market was borne of a recession that resulted largely from the bursting of bubbles in the stock market and technology sector and was exacerbated by a liberal dose of corporate malfeasance. Over the past few years, the Bush Administration and Congress have passed numerous tax cuts purportedly intended to stimulate the weak economy and restart the engine of job creation that was delivering 241,000 jobs per month, on average, between 1995 and 2000, a stark contrast to an average loss of 93,000 jobs a month since March 2001. Thus far, these tax cuts have failed to lower unemployment, much less stimulate private sector job growth. In fact, hundreds of economists, including 10 Nobel laureates, argued that these cuts were poorly structured as a stimulus.10
As noted before, recent fiscal boosts, both in terms of spending increases and tax cuts, will almost surely lead to faster near-term growth. This in turn will probably reverse the recent pattern of monthly job losses and perhaps result in a lower unemployment rate. But, as the evidence presented here suggests, the weaknesses in the labor market, particularly slow wage growth, will likely be a strong countervailing factor.

Absent more broad-based real wage gains, it will be difficult to generate a self-sustaining recovery anytime soon. The most optimistic predictions, t
hose by the Bush Administration, call for unemployment of 5.5% by the end of 2004. More realistic assessments are closer to 6.0% (see Figure 15). On this Labor Day 2003, we are stuck with a weak labor market, and the actions of policy makers have thus far been insufficient to alter this reality.

—September 2003

Appendix A

Questions & answers regarding the current labor market

Q: Was the National Bureau of Economic Research (NBER) wrong in dating the recession to have ended in November 2001?

A: No, at least not in the sense that they were consistently following a set of well-established rules for recession dating. The NBER committee that dates recessions uses a variety of measures to determine whether the economy is expanding or contracting, of which employment is but one. Most of the other measures, including retail sales, aggregate income (minus government transfers, such as welfare benefits and unemployment insurance), and real GDP, have all been growing (another variable the NBER considers, industrial production, has not reversed its negative trend). Of course, as stressed throughout, the expansion is too weak to generate jobs—a point the NBER committee acknowledged.

Q: Thus far in the recession and jobless recovery the unemployment rate has not gone above 6.4%. Isn’t this low in historical terms?

A: While it is true that unemployment rose more and peaked at higher rates in prior downturns (e.g., 10.8% in December 1982, and 7.8% in June 1992), the increase in unemployment this long into a recovery period (i.e., nine quarters past the last business-cycle peak) has surpassed that of the last two recessions/recoveries (see Table 1). For example, for African Americans the increase in unemployment nine quarters out from the last peak was 3.2 percentage points, compared to 2.5 and 2.1 points in the two earlier comparable periods (see Table 1). Also, as stressed above, the unemployment rate would be higher if labor force growth had kept pace with population growth.11
Q: Many commentators have argued that the biggest problem holding back job growth right now is the fact that productivity is growing quickly. Is that the problem?

A: No, faster productivity growth is not the problem. The fact that we are producing more efficiently now than we were a few years ago does mean that the U.S. workforce can create the same amount of output in fewer hours, but that is a positive development that should lead to higher living standards than would otherwise be the case.12 The economy’s current problem is weak demand by consumers and investors.

Appendix Figure A shows that productivity growth in the recovery has been only slightly (0.3%) above the average of that of the past eight expansions. Employment growth, on the other hand, has been much stronger in past recoveries (3.3% vs. -0.5%). Productivity often accelerates as an expansion gets underway, as output growth is typically faster than growth in hours, and employers slowly expand their workforce to meet the return of demand for their firms’ goods and services. But in this recovery (and in the last), demand has been too weak, outside of a few sectors, such as health care, housing, and autos, to prompt employers to begin hiring. As the last set of bars in Figure A shows, demand (output in the nonfarm sector) grew more than twice as fast in past expansions compared to this one.

Appendix Figure A

Q: Why has the jobless recovery persisted for so long?

A: Again, the main constraint has been the persistence of weak demand for the goods and services the U.S. economy produces, outside of those few sectors noted above. Though consumption in these areas has kept the economy afloat, consumers have not been spending as aggressively as they usually do after a recession, and this in turn has made investors more cautious. This vicious cycle has been reinforced by high debt levels. The result is historically very weak hiring by employers who are reluctant to build up their workforces in such an environment. As stressed in this Briefing Paper’s analysis, all of these trends are undermining wage growth, which further constrains demand.

Appendix B

Data Notes

Table 1: Authors’ analysis of Bureau of Labor Statistics data.

Table 2: Authors’ analysis of CPS Outgoing Rotation Group files. Sample is all workers, age 18-64. See Mishel et al. (2003, Appendix B) for description of methods.

Figures 1A, 1B, 2, 3, and 4: Authors’ analysis of Bureau of Labor Statistics data.

Figure 5: Data for 1979:q1-1991:q4 are from authors’ analysis of CPS monthly data, seasonally adjusted by authors. From 1992 forward, we use published BLS data (note that these data are unpublished prior to 1992). Prior to 1992, the education question in the CPS was coded differently, so we make adjustments as described in Mishel et al. (2003, Appendix B) to make the series comparable over the coding change.

Figure 6-7: Authors’ analysis as described in note for Figure 5.

Figure 8: Authors’ analysis of BLS data. Simulated series is derived by holding labor force participation constant at its March 2001 level. Also, the impact of the new population weights in January 2003 is removed by subtracting the number of persons added to the labor force in January 2003 due to the adjustment of weights (576,000), as shown in Bowler et al (2003, Table 5).

Figure 9: Authors’ analysis of Bureau of Labor Statistics data.

Figure 10: Unemployment data from BLS. Underemployment is reported monthly as measure “U-6” in Table A-12 of the BLS monthly employment report. We seasonally adjust that series.

Figure 11: Authors’ analysis of CPS monthly data, using the BLS definition for “U-6” (see prior note).

Figure 12-14: Authors’ analysis of CPS Outgoing Rotation Group files. Sample is all workers, age 18-64. See Mishel et al. (2003, Appendix B) for description of methods, including smoothing procedure for wage percentiles.

Figure 15: Blue Chip estimates come from the August 10, 2003 publication of Blue Chip Economic Indicators, Vol. 28, No. 8, Table 3. Office of Management and Budget estimates are from the most recent mid-session review, available at:, Table 4.

Figure 16: Authors’ analysis of BLS data. Productivity and output growth are for the nonfarm business sector; employment growth is for total nonfarm payrolls. All rates are annualized.


Bernstein, Jared and Lawrence Mishel. 2003.  Jobless Recovery Catches Up With Wages. EPI Issue Brief #193. Washington, D.C.: Economic Policy Institute.

Bowler, Mary, Randy E. Ilg, Stephen Miller, Ed Robison, and Anne Polivka. 2003. “Revisions to the Current Population Survey Effective in January 2003.” Washington, D.C.: Bureau of Labor Statistics.

Mishel, Lawrence, Jared Bernstein, and Heather Boushey. 2002. The State of Working America, 2002-03. Ithaca, N.Y.: Cornell University Press.


1. The Joint Committee on Taxation reports that the tax cut will deliver $61 billion by the end of this fiscal year (September 2003). For next fiscal year, the total stimulus is $149 billion. Adding one quarter (for Oct.-Dec.) of that to the $61 billion leads to about $100 billion this calendar year.

2. See < > for a statement to the effect signed by 10 Nobel laureates and hundreds of economists.

3. We omit the recession that began in January 1980. That downturn was the first of a “double dip,” so going out 28 months leaves us in the midst of the next downturn. We do include this second recession in this and following analysis.

4. The BLS has recently introduced a new survey of job openings (JOLTS, or Job Openings and Labor Turnover). JOLTS data reveal that in April of this year there were about three million job openings. Labor force data show that there were 8.5 million unemployed in that month (both values are non-seasonally adjusted, as JOLTS data cannot be adjusted yet). Thus, there were close to three unemployed persons for each job opening in the labor market.

5. The sample for these figures is persons 25 years of age and up.

6. Interesting contrasts in labor force participation rate trends are also evident by age group. Labor force participation fell much more in this relative to past recessions/recoveries for 16-19 year olds, and conversely, grew much more for those 55 and older. For persons aged 25-54, participation rates have usually grown slightly in these periods, but in the nine quarters since 2001:q1, they are down about one point.

7. Note that the underemployment series, which corresponds to BLS series U-6 in Table A-12 from the monthly unemployment report, is not seasonally adjusted by the BLS. We seasonally adjust the series ourselves; it is not an official BLS series.

8. The series are seasonally adjusted. The source for these data is the CPS Earnings Files. See Mishel et al. (2003, Appendix B) for discussion of methods.

9. The real values combine the impact of nominal wage trends and inflation. Bernstein and Mishel (2003, Figure C) show that while inflation decelerated considerably over this period, nominal wages grew even more slowly in recent quarters.

10. See < > for a statement to the effect signed by 10 Nobel laureates and hundreds of economists.

11. It is also worth noting that literature on how people experience economic distress suggests that changes in economic variables are felt more acutely than levels. That is, a sharp rise in unemployment appears to lead to more pessimism about the economy than a small rise, even if the latter is from a higher unemployment rate. As a simple test of this notion, we regressed the growth (log changes) of the Conference Board’s consumer confidence index on unemployment levels and, separately, unemployment changes (and a constant). In the levels regression, the coefficient on unemployment was wrong-signed (positive) and statistically insignificant. In the changes regression, the coefficient of the change in unemployment was negative and significant (with an elasticity of about 1). Note also that evaluation statistics improved with the use of log changes in the unemployment rate, suggesting that the increase in unemployment from lower levels—as occurred in this downturn—leads to greater losses in confidence than increases from higher levels.

12. Of course, how the fruits of productivity growth are distributed throughout the economy is wholly another matter.


Thanks to the following foundations who support EPI’s Living Standards Program: The Charles Stewart Mott Foundation, Foundation for Child Development, John D. and Catherine T. MacArthur Foundation, Joyce Foundation, Public Welfare Foundation, and Rockefeller Foundation. The findings presented in this report are those of the authors alone, and do not necessarily reflect the opinions of these foundations.

The authors would also like to thank Sujan Vasavada, Yulia Fungard, and Abhi Nargundkar for their research assistance, and Lee Price for helpful comments.

See related work on Education

See more work by Jared Bernstein and Lawrence Mishel