The Federal Reserve’s mandate to pursue maximum employment requires that all Americans can access employment and earnings opportunities. Over the years, Fed officials have articulated this. In 2017, for example, Fed Governor Lael Brainard argued that understanding “labor market disparities is central to the mission of the Federal Reserve because it may help us better assess full employment, where resources may be underutilized, and the likely evolution of the labor market and overall economic activity.”
Business cycles rapidly shuffle earnings and employment for hundreds of thousands of workers, and economists share a sense that recessions exacerbate inequality while recoveries, particularly long or strong ones, have the potential to reduce it. Firms might lay off lower-wage or hourly workers when conditions worsen, for instance. They may also loosen hiring criteria as they expand in good times, providing new opportunities for workers who typically have a hard time getting a job.
But this sense stops short of a firm consensus, in part because high-quality data on inequality is hard to come by. Public sources tell us about changes in inequality in broad strokes, but can be unreliable when it comes to inequality in the places where Americans live and the demographic groups they are part of.
In this article, we use a new source, the Opportunity & Inclusive Growth Institute’s Income Distributions and Dynamics in America data, to investigate what better data can tell us about how inequality across demographic groups evolved during and after the Great Recession. We find that earnings gaps between non-Hispanic White men and other groups narrowed in many cases, but most gains were modest. Moreover, the earnings gains relative to non-Hispanic White men occurred in both the recession and the recovery periods. But there were important exceptions. Asian earners demonstrated large relative earnings movements and Black earners experienced relative earnings losses in the recovery.
This short analysis is across racial and ethnic groups has been of interest to Fed policymakers for some time. The Income Distributions and Dynamics in America statistics allow detailed analysis of a range of measures of inequality, which future articles will continue to explore.
One group at a time
To understand how inequality between groups changed over the 2007 to 2019 business cycle, it’s useful to start by looking at inequality levels over time. Figure 1 shows how median earnings for women of different racial and ethnic groups have evolved compared with the median earnings of non-Hispanic White men.1 This is called “relative earnings” because it reports how much women in each group earned as a percent of the reference group’s earnings.
To illustrate this concept, consider the dark blue line in Figure 1, which plots relative earnings each year for Hispanic women. In 2005, the median earnings of a Hispanic woman was $21,470 (in nominal dollars), about 50 percent of the median earnings of a non-Hispanic White man, so the line starts at about 50. By 2019, however, this gap had shrunk somewhat. At the median of the income distribution, working Hispanic women made $33,140 while the median earnings for working White men was $56,410. The rise in the line to about 59 percent reflects this slight convergence.
The other lines in Figure 1 show that the earnings gap narrowed for all groups of women included in the analysis compared with the reference group of non-Hispanic White men. However, experiences differed, sometimes considerably. Hispanic women, for example, saw no relative gains at all from 2010 to 2014, after which their earnings did see some convergence. Black women and Native Hawaiian or Pacific Islander women lost ground in the first half of the recovery.
Loading chart 1...
Contractions and convergence?
Next, we take the type of information plotted over time in Figure 1 and use it to examine whether groups defined by race and gender moved closer together or further apart over the Great Recession and subsequent recovery. We look at how relative earnings changed in two periods: contraction from 2007 to 2009 and recovery from 2009 to 2019.
Figure 2 uses a scatterplot to investigate whether there is a relationship between a group’s relative earnings at the beginning of the economic contraction in 2007, and how that gap changed between 2007 and 2009. For example, Figure 1 shows relative earnings for Hispanic women in 2007—the onset of the recession—at 52 percent. In Figure 2, this marks their position on the x-axis. The gain Hispanic women made during the recession is reflected in their positive value on the y-axis.
If the Great Recession reduced cross-group inequality, we would expect groups that started off further behind would experience more positive change in their relative earnings than other groups. In this figure, this would mean that points to the left would lie above zero and higher than points to the right, indicating that groups with earnings further below the reference group in 2007 saw their earnings grow more over the next two years than groups who were already earning something closer to non-Hispanic White men.
Overall, gaps in median earnings across racial and ethnic groups were stark before the Great Recession and they were almost as stark immediately after.
So looking across groups, was the Great Recession an equalizer? Figure 2 suggests that it did equalize median earnings across demographic groups, but not by very much. No group gained more than about 4 percentage points relative to White men, but no group lost significant ground either. The lowest-earning groups in 2007 did not gain much more than groups who started off closer to parity, such as Asian women. Asian men, whose median income exceeded that of White men in 2007, actually pulled a little further away. Overall, gaps in median earnings across racial and ethnic groups were stark before the Great Recession and they were almost as stark immediately after.
Women also clearly gained ground, since all points for women show positive changes in relative earnings. This is consistent with the pattern highlighted in Figure 1, but Figure 2 helps show that these gains were modest.
Loading chart 2...
Expansions and inclusion?
We next do the same exercise for the recovery period from 2009 to 2019, shown in Figure 3. Again, if groups with the lowest relative earnings in 2009 saw more income growth during the recovery than groups with higher earnings did, the points in Figure 3 would be higher on the left and lower on the right.
Figure 3 does show a number of points above the zero line, indicating those groups experienced relative earnings growth, but the overall pattern is fairly erratic. For most groups, the gains or losses were small. Asian men and women were notable exceptions. Figure 3 shows that Asian men earned more than White men at the start of the recovery, and their earnings continued to grow relative to this group over the recovery. Asian women earned 80 percent of White male earnings at the onset of the recovery, and their relative gain of almost 20 percentage points over the recovery essentially closed that gap. Other notable gains were made among Hispanic men, who saw their earnings gap with White men narrow by 7 percentage points. Hispanic women and Native Hawaiian or Pacific Islander men saw gains approaching 5 percentage points.
Loading chart 3...
We put these analyses together in our final figure. Figure 4 marks the change in relative earnings (again in percentage points) in the recession with a point for each group, reproducing Figure 2. The subsequent change for the same group over the recovery (the height of the points in Figure 3) is indicated with an arrow. The endpoint of each arrow shows how relative earnings changed for each group over the recession and recovery from 2007 to 2019. The figure shows whether a group gained or lost relative to White men in the contraction and then whether the group went on to gain or lose in the expansion.
Most groups chipped away at earnings disparities over the full recession and recovery period, but the sizes of their relative earnings increases are not large. For example, the endpoint of the arrow for men who are American Indian or Alaska Native indicates relative earnings growth of 2 percentage points over the 12-year period. Thus in 2019, median earnings for American Indian or Alaska Native men was $38,130, compared with $56,410 for White men—more than a 30 percent disparity. The gap in median earnings for Black women relative to White men was the same in 2019 as it was in 2007, and the disparity for Black men was slightly larger in 2019. While many groups saw little change in their relative earnings, Black men were the only group that saw their earnings meaningfully eroded.
By contrast, Asian men and women showed large and sustained increases in relative earnings during this time, particularly in the recovery. To put that growth in perspective, we can compare this with the rising earnings gap between college and high school graduates, a long-term trend that has received much attention from researchers. That gap widened 15 percentage points between 1979, when baby boomers were young workers, and 2013, when millennials were. By comparison, the change in earnings disparities at the median for Asian earners outstripped that difference in fewer years. Hispanic men and women also saw meaningful relative earnings gains.
Loading chart 4...
Income Distributions and Dynamics in America data on median earnings do not suggest that cross-group inequality changed very much during or after the Great Recession. Median earnings for most groups held steady at the level of inequality that prevailed in 2007, though rising earnings for Asian workers was an exception and Black workers lost some ground.
But other research suggests that recessions and recoveries do have disparate impacts across demographic groups. In a 2019 Brookings paper, Stephanie Aaronson, Mary Daly, William Wascher, and David Wilcox find that racial gaps in average individual incomes (by gender) widen when unemployment rates are high and shrink when unemployment rates are low. In a series of articles in the early 1970s, Arthur Okun made the argument that economic expansions help workers move up to better jobs, creating lower-level openings that firms fill from the pool of unemployed workers, many of whom are from marginalized groups.
There are a few reasons why this introductory analysis of national changes might not uncover big changes in cross-group inequality during the Great Recession, even if that really was an important feature of the American economy.
First, this analysis only looks at national changes over one business cycle, so we cannot separate out labor market trends specific to this time—like the rise of gig work or the growing visibility of sexual harassment in the workplace that accompanied the #MeToo movement—from general cyclical patterns. Another factor is our measure. In this analysis we look at medians, but if the Great Recession primarily affected earnings between, say, the 20th and 30th percentiles, then median earnings would stay the same even though income changed meaningfully for some workers. Finally, our statistics apply to people who worked at least 13 weeks at minimum wage, but since lower-wage employment expands more during boom times, the composition of who is working varies over the business cycle.
The solution to these problems is, once again, better data. Fortunately, researchers can now use different parts of the Institute’s Income Distributions and Dynamics in America data to generate a richer answer to exactly the kinds of questions that the Fed’s dual mandate poses.