Got work? The highly responsive labor supply of low-income, prime-age workers.

October 2nd, 2019 at 8:02 pm

[Note: this is draft of a forthcoming paper for CBPP’s Full Employment Project. I posted it here first as I will be referencing its findings at a Brookings inflation conference on Thurs, Oct 3.]

By Jared Bernstein and Keith Bentele[i]


The benefits to running a hot labor market continue to be evident both in the data and in anecdotal accounts. In our last paper, we examined the monetary policy rationale for allowing high-pressure labor markets to continue to flourish.[ii] We also focused on the benefits of persistently low unemployment to lower income workers, through both higher real pay and more hours of work. In this short paper, we turn back to this evidence, with a closer focus on the benefits of high-pressure labor markets to the labor supply of lower-paid workers.

The most basic labor market theories generally lack the necessary nuance to shed much light on this question. The textbook 101 model assumes full employment and an equilibrium wage where employers’ demands’ and undifferentiated workers’ supply perfectly match. A wage set too high will lead to more job seekers than jobs; a wage set too low will cause the opposite problem: too few workers willing to fill available slots. In the real world, however, there are of course periods of slack labor markets, along with factors such as racial and gender discrimination. Some particularly disadvantaged workers may face uniquely high barriers to labor market entry. Also, recent research has identified large sectors in our economy, like retail, tech, and health care, where few employers dominate. In such markets, employers can become wage makers, not wage takers, i.e., they can use their dominance to set wages below the theoretical equilibrium.

In this note, we ask a simple, specific question related to this more realistic version of the labor market: do low-wage workers respond to high-pressure labor markets by increasing their labor supply? What evidence is there that tight job markets pull in such workers?

We find highly cyclical responses to both the extensive and intensive margins of labor supply for low-income, prime-age persons, especially for first-quintile African Americans and for women. A simple decomposition finds, for example, that the earnings of low-income Black people doubled in the high-pressure labor market of the 1990s, with gains in their share working (extensive margin) explaining half of the increase. Conversely, under low-pressure conditions, the decline in working shares dominates sharp income losses for these groups of people. We also find the extensive margin to be particularly important for low-income women, and a simple simulation suggests that in the hot labor market of the 1990s, most of the gains in the gender gap were due to women’s relative (to men) gains along the extensive margin. We offer policy implications of these findings on both macro and micro levels.

Previous Literature

Arthur Okun is widely credited with pioneering research into the benefits of very low unemployment for marginalized groups. In the context of a “high-pressure economy” he hypothesized that employers are more likely to lower formal hiring standards in order to fill vacancies, and that this would benefit  less-advantaged job seekers in the labor market[iii]. In a 1973 paper Okun found that in such periods, women and teenagers experienced disproportionately large increases in employment. Since this initial confirmation of a rather straightforward hypothesis, many studies have reinforced this finding. The uniquely strong economy in the late-1990s prompted a body of work evaluating the impacts of these conditions. Roberts and Rodgers (2000) examined the impacts of low unemployment on earnings and employment in metro labor markets and found that less educated men, and young African American men in particular, experienced the greatest improvements[iv]. Similarly, Wilson (2015) explored the positive impacts of strong economic growth in the late-1990s on both the employment and earnings of African American people[v]. Katz and Krueger (1999) found that the tight labor market of the late-1990s contributed to a significant increase in the both the incomes of lower-income families and falling poverty rates in those years[vi]. Jargowsky’s (2003) research captured a 24% decline in the number of people living in high-poverty neighborhoods, census tracts where 40% or more of residents are in poverty, between 1990 and 2000[vii]. This study found that the strong economy reduced the concentration of poverty for all racial and ethnic groups, but had a particularly pronounced effect on African American communities. The share of poor African American people living in high poverty neighborhoods fell from 30% in 1990 to 19% in 2000. Such impacts have informed William J Wilson’s assertion that the “ideal solution” to addressing a root cause of concentrated poverty, inner-city joblessness, “would be economic policies that produce tight labor markets” (Wilson 2008:568)[viii].

The specific findings of greater cyclical variation in wages, employment, and labor market participation for economically vulnerable groups have been widely documented. Hoynes (2000) found this to be the case for the employment and earnings of less educated workers, people of color, and women with lower skill levels. All experienced more variation over the course of the business cycle relative to higher skilled men, a finding that was particularly pronounced in the context of one’s probability of being employed full-time year round[ix]. Jefferson (2008) found that trends in the employment-to-population ratio for workers with less education are substantially more volatile compared to those with more educated workers [x]. And more recent research, including that which examines the current strong economy, has only further bolstered these findings. For example, Aaronson et al. (2019:3) state that their research,

“reaffirm[s] the earlier finding of other authors that the labor market outcomes of blacks, Hispanics, and those with less education are more cyclically sensitive than the outcomes of whites and those with more education.”[xi]

And consistent with the work of Roberts and Rodgers (2000), they find that cyclical variability in labor market outcomes is particularly pronounced for young African American workers. Further, Aaronson et al. (2019) find suggestive evidence that further strengthening in the context of an already very strong economy is particularly beneficial to some disadvantaged groups.

The flip side of higher cyclical variability is that recessions hit low-wage workers and members of marginalized groups particularly hard. Decreases in earnings and employment during recessions are consistently disproportionately larger for low-wage earners, people of color, and low-income female-headed households[xii]. Kenworthy (2011) has stressed the devastating impact of recessions on hours worked by very low-income households, an issue compounded when followed by a weak recovery[xiii]. While the evidence is admittedly thin at this point, high-pressure labor markets may potentially play a small protective role in regards to this cyclical vulnerability. Aaronson et al. (2019) find that the gains made during high-pressure periods for Black people and women are “somewhat persistent”. Similarly, Hotchkiss and Moore (2018) find that high-pressure periods lead to higher employment, wages, and earnings in subsequent downturns for young men and Black people. However, they stress that these specific positive benefits are largely confined to these groups and may be short lived.


For a full description of the data used below, see our earlier paper. We use the same procedures and inclusion criteria to generate the estimates used here with only two important changes. First, previously we used household income from all sources to determine quintile thresholds, here we have used a measure of total household income adjusted for household size (household income/number of household members). Second, our estimates of annual hours worked are based on the average of individual-level data on hours worked (hours per week*weeks per year), as opposed to a quintile-level estimate of annual hours. We found that our inclusion of individuals with $0 earnings in our estimates of quintile level hours and weeks produced an underestimation of annual hours that was not present in the averaged individual-level estimate of annual hours.


While it is common for labor analysts to look at employment rates, such rates are generally overall averages, often by gender or race. Because our data is broken out by income fifth (and by gender and race), we can look at the share of prime-age people with positive annual work hours, meaning any reported paid work last year, by quintile.[xiv] As Figure 1 shows, consistent with prior research, the series for persons in the lowest fifth is much more cyclical than that for the middle or top fifth.

Figure 1.

In fact, simply using the unemployment rate (logged, with one lag) and the lagged first-quintile series in the above figure, we can derive a dynamic prediction that tracks the series well (See Figure 2).[xv] The share working exceeded the forecast in the latter 1990s, but this was a period when many “pro-work” policy changes affected low-income workers, including work requirements in TANF, a large expansion in the Earned Income Tax Credit, and an increase in the minimum wage. But it was also a high-pressure labor market period. Using the 2018 unemployment rate of 3.9 percent, the model forecasts a jump of about 5 percentage points in 2018.[xvi]

Figure 2.

Figure 3 examines the share of working African American people. Because of the smaller sample size, we plot the bottom 40 percent. These workers appear to be particularly elastic to high-pressure labor markets, with large employment gains in the 1990s and in the current recovery.

Figure 3.

In the appendix, we include similar figures for prime-aged men and women in the first quintile of adjusted household incomes. The cyclical patterns are roughly similar to the total figure above (the simple model tracks the series well), though as earlier research has shown, the women’s series tends to trend up while that of men trends down due in part to structural challenges like the loss of production jobs.

Because we have earnings data and annual hours, we can decompose the earnings of low-income people into the share working (the “extensive margin”), annual hours among workers, and hourly wages (see Table 1).[xvii] By decomposing these changes over different time periods, we can observe how this group has fared in both high- and low-pressure labor markets and which factors have the greatest impact on real earnings growth.

Table 1. Real Earnings Contributions in High- and Low-Pressure Labor Markets

The first few panels look at high-pressure labor markets. Between 1993-2000, when the unemployment rate fell from 6.9 to 4 percent, the real, annual earnings for all low-income persons rose just under 50 percent (log points), from about $8,200 to $13,400. Importantly, we include those with zero hours, and thus zero earnings, in these calculations, so as to capture the impact on earnings of their crossing the extensive margin (going from zero to positive hours worked). As the first column shows, the share working increase 23 percent, and since this is an additive decomposition (the first three columns of the row “ln change” sum to the fourth column), this added labor supply explains almost half (23/49) of the earnings gain over this high-pressure period.

The second panel show the more recent period, 2011-2017, when unemployment fell from 8.9 percent to 4.4 percent. Real earnings were up about 30 percent over this period, meaning that on an annualized basis, real earnings grew 2 points faster in the 1990s high-pressure labor market (7 versus 5 percent per year) through 2017. The extensive and intensive margins, entry into employment and increased hours respectively, explain about one-third each, as does the growth in real hourly wages.

The next panel drills down into the experience of first-quintile African Americans in the high-pressure labor market of the 1990s, when Black unemployment fell from 13 percent to 7.6 percent and their real, annual earnings remarkably doubled, for an annual real growth rate of just under 15 percent per year. Fully half of that gain was due to an almost 50 percent increase in the share working, with the other half split between more annual hours by workers and higher real hourly wages.

It is important to turn to low-pressure labor markets to see how this process shifts into reverse. From 2007-11, the jobless rate rose from 4.6 to 8.9 percent, and first quintile real earnings fell almost 30 percent, with half the decline attributable to the fall in the share of workers from 67 to 58 percent. Compare this fall to the absence of change (not shown) for the top quintile over these years. In both 2007 and 2011, their prime-age share at work was 95 percent, a stark reminder of who bears the brunt of recessions.

African American unemployment about doubled, 2007-11, from about 8 to 16 percent. Black people in the first quintile lost almost half of their real earnings in this downturn, with three-fifths of the decline coming from the decline in the share working.

Turning back to the top fifth, the bottom panel of Table 1 shows how affluent, prime-age people are extremely inelastic regarding changes in labor supply. In fact, they’re largely topped out in terms of annual hours worked and share working. Their average share working is 95 percent and the standard deviation around that mean over these years is  three-fifths of a percentage point, compared to a standard deviation of 6 points for the first quintile of all prime-age people (10 times that of the top fifth) and 10 points for the African American first quintile. These are profound differences in cyclical variability in employment.

Table 2 presents the same decomposition for first quintile, prime-age people by gender, looking at the high-pressure 1993-2000 period, the Great Recession, and the slow recovery of 2007-11. As shown in the appendix figures on share working by gender, low-income women are more cyclically responsive in these data, at least until around 2000, when both genders appear somewhat more elastic to the business cycle. The findings show that low-income women’s extensive margin gains were particularly important in the 1990s, explaining 60 percent of their significant real earnings’ gains (45/75). For men, however, that margin contributed little compared to more hours worked and higher real wages.

Table 2. Real Earnings Contributions in High- and Low-Pressure Labor Markets, by Gender


Table 3 does a simple simulation to parse out the role of crossing the extensive margin in reducing the gender wage gap in the high-pressure labor market of the 1990s. Note that the gender gap compressed by 23 percentage points in these years, from 39 percent in 1993 to 62 percent in 2000. A simple simulation (see footnote and table note) suggests that almost all of that closure was a function of the increase in women’s share working.[xviii]

Table 3. Gender Inequality (Q1) and the Extensive Margin

Note: The 2000 simulated value (in bold) is the product of the 1993 share of women working and their 2000 hours and hourly wage.

In the low-pressure labor market panel, real earnings fell sharply for both genders, and in this case, the extensive margin dominates in both cases.


Our findings and those of other researchers, in tandem with extensive recent anecdotes from the media regarding new opportunities for left-behind workers, show that the labor supply of low-income workers, especially women and Black people, is highly elastic to labor market conditions, while that of high-income workers is much less so. In the high-pressure labor market of the latter 1990s, real annual earnings grew by half for all first quintile workers (including “zeros”) and doubled for African American people. In both cases, the increase in the share working explained about half these gains.

The policy implications of these findings invoke both macro and micro policies. At the macro level, as we stressed in our earlier paper, running the labor market hotter-for-longer returns economically significant benefits to those who need them the most. From the perspective of the Federal Reserve’s monetary policy, the additional fact that inflation has proven to be quite insensitive to low unemployment seals the deal: as long as inflation remains “well-anchored,” for vulnerable workers to get ahead, we can and should pursue full employment.

At the micro level, our findings speak strongly against the notion that receipt of anti-poverty benefits should be conditional on the work requirements that have recently surfaced in various states, for example, through federal waivers to the Medicaid program. Low-income workers have already been highly responsive to job opportunities; the problem is that those opportunities either don’t exist in slack labor markets, or they face internal (skill deficits) or external (discrimination) barriers to getting into the job market. Hassling them off of the benefit rolls by making them jump through administration hoops will not lead them to work more, but it will surely diminish their income and their health.


Table A.1

Table A.2



[i] We thank Jesse Rothstein for comments and Kathleen Bryant for technical assistance. Any mistakes are our own.

[ii]  Jared Bernstein and Keith Bentele, “The Increasing Benefits and Diminished Costs of Running a High-Pressure Labor Market,” Center on Budget and Policy Priorities: Full Employment Project, May 15, 2019,

[iii] Okun, Arthur M. 1973.“Upward Mobility in a High-pressure Economy.” Brookings Papers on Economic Activity. 1:207-261.

[iv] Cherry, Robert and William M. Rodgers III (eds.) 2000. Prosperity for All? The Economic Boom and African Americans. New York: Russell Sage Foundation.

[v] Wilson, Valerie. 2015. “The Impact of Full Employment on African American Employment and

Wages,” Economic Policy Institute.

[vi] Katz, Lawrence, and Alan Krueger. 1999. “The High-Pressure U.S. Labor Market of the 1990s.” Brookings Papers on Economic Activity. 1:1-87.

[vii] Paul Jargowsky. 2003.  “Stunning Progress, Hidden Problems: The Dramatic Decline of Concentrated Poverty in the 1990s.” The Brookings Institution.

[viii] William Julius Wilson. 2008-09. “The Political and Economic Forces Shaping Concentrated Poverty.” Political Science Quarterly. 123(4):555-571.

[ix] Hoynes, Hilary. 2000. “The Employment and Earnings of Less Skilled Workers Over the Business Cycle.” in Finding Jobs: Work and Welfare Reform, edited by Rebecca Blank and David Card. New York: Russell Sage Foundation.

[x] Jefferson, Philip N. 2008. “Educational Attainment and the Cyclical Sensitivity of

Employment.” Journal of Business and Economic Statistics 26(4):526-35.3

[xi] Aaronson, Stephanie R., Mary C. Daly, William Wascher &  David W. Wilcox. 2019. “Okun Revisited: Who Benefits Most From a Strong Economy?” Brookings Papers on Economic Activity. BPEA Conference Drafts.

[xii] Aaronson, Stephanie R., Mary C. Daly, William Wascher &  David W. Wilcox. 2019. “Okun Revisited: Who Benefits Most From a Strong Economy?” Brookings Papers on Economic Activity. BPEA Conference Drafts.

Bentele, Keith G. 2012. “Evaluating the Performance of the U.S. Social Safety Net in the Great Recession.” Center for Social Policy Publications. Paper 62.

Cajner, Tomaz, Tyler Radler, David Ratner, and Ivan Vidangos. 2017. “Racial Gaps in Labor Market Outcomes in the Last Four Decades and over the Business Cycle.” Working Paper 2017-071. Finance and Economics Discussion Series. Washington, D.C.: Federal Reserve Board.

Zavodny, Madeline, and Tao Zha. 2000. “Monetary Policy and Racial Unemployment Rates.” Federal Reserve Bank of Atlanta Economic Review 85(4):1–59.

[xiii] Kenworthy, Lane. 2011. Progress for the Poor. New York: Oxford University Press.

[xiv] Note that this is a somewhat different metric than the oft-cited prime-age employment rate from BLS. The measure asks if people are working in the reference week of the month; this one asks if people had any positive work hours over the course of the prior year. The latter tends to have a higher level, but the trends are roughly similar.

[xv] “Dynamic” in this context means the lagged dependent variable is predicted (versus plugging in the actual value) for each observation in the predicted series.

[xvi] We plan to shortly update this analysis with the recently released 2018 data.

[xvii] Annual earnings for all prime-age persons in the quintile (assigning zeros to non-workers) is the product of the
share working*annual hours*hourly wage. Taking log changes facilitates the decomposition.

[xviii] We make this calculation by simulating women’s 2000 earnings using their actual hours and wage but using their 1993 share working (45%). The difference between this simulated gender gap and the actual gender gap can thus be assigned to their large increase in share working.

Print Friendly, PDF & Email