Skip to main content

Occupational segregation left workers of color especially vulnerable to COVID-19 job losses

Latino/a workers experienced large employment losses in spring 2020 due to occupational segregation

February 7, 2022


Tyler Boesch Data Scientist, Community Development and Engagement
Ryan Nunn Assistant Vice President, Community Development and Engagement
Alene Tchourumoff Senior Vice President, Community Development and Engagement
Occupational segregation left workers of color especially vulnerable to COVID-19 job losses, key image
Reza Estakhrian/Getty Images

Article Highlights

  • Workers of color disproportionately lost jobs at the start of the pandemic
  • Pre-existing occupational segregation was a major factor, especially for Latino/a workers
  • Workers of color had not fully recovered to their pre-pandemic employment rates by late 2021
Occupational segregation left workers of color especially vulnerable to COVID-19 job losses

The economic hardships Americans have experienced since the beginning of the COVID-19 pandemic have fallen disproportionately on people of color. To better understand why the imbalance exists, we examined how labor market conditions prior to the pandemic laid a foundation for these disparate experiences during the pandemic. Our analysis indicates that the segregation of workers of color into hard-hit occupations was a major cause of the disproportionate job losses, especially for Latino/a workers.

Disproportionate losses for workers of color

Comparing Bureau of Labor Statistics data on the change in each racial and ethnic group’s employment-to-population ratio1 between a pre-pandemic baseline period of December 2019–February 2020 and one of the worst months of the pandemic recession, May 2020, shows that employment declines were substantial for all groups but larger for workers of color:

  • White employment fell by 7.5 percentage points;
  • Black employment fell by 9.5 percentage points;
  • Native American employment fell by 10.1 percentage points;2
  • Asian employment fell by 10.6 percentage points; and
  • Latino/a employment fell by 11.4 percentage points.

By subtracting the overall decline in employment—7.9 percentage points—from each group, we can determine each group’s “excess” employment decline.3 In Figure 1, we plot these excess employment declines to show each group’s experience relative to the overall workforce over time.

Comparing these excess declines reveals that as the labor market recovered into 2021, workers of color regained much of their lost ground—both in absolute terms and relative to the overall labor market. By October 2021, Black and Latino/a employment rate declines remained 0.8 and 0.4 percentage points greater, respectively, than the overall employment decline from just before the pandemic to October 2021 (1.6 percentage points). Meanwhile, Asian workers returned to pre-pandemic employment rates in Summer 2021. Not shown in the figure: The excess decline for Native American workers, by contrast, had disappeared, though their employment rate level remained persistently below that of most other groups.

Loading chart 1...

The role of occupational segregation

Looking back on the disproportionately difficult pandemic labor market experience of workers of color, occupational segregation stands out as one factor deserving of attention. To what extent were workers of color more vulnerable to pandemic-related job loss because of pre-existing segregation into certain types of work?

Workers of different groups are not proportionally found in each occupation; rather, some occupations have systematically larger shares of workers of color and some have smaller shares (del Rio and Alonso-Villar 2015). This matters for the quality of jobs that workers obtain and can render workers who are concentrated in certain jobs more exposed to economic shocks. This was the case during the pandemic recession. Workers providing customer-facing services, such as many restaurant employees, were especially likely to lose their jobs (Wardrip and Tranfaglia 2020). By contrast, many workers in the professional and technical industries have the ability to work remotely and have been relatively insulated from layoffs.

To better understand how occupational segregation shaped the employment trajectories of different racial and ethnic groups, we conducted an analytical exercise that isolates the contribution of occupational segregation. Specifically, we calculated the share of each occupation’s employment4 that a group held from December 2019 to February 2020, holding it constant before and after that date.5 As the pandemic caused employment to decline in the spring of 2020, some occupations were hit harder than others. Figure 2 shows how each group’s excess employment rate (i.e., the amount by which the pandemic change in a group’s rate exceeds the overall change) would have evolved over time if the group maintained its share of each occupation’s employment but suffered the loss of occupation-level employment that was actually observed during the pandemic. For example, the large decline observed for Latino/a workers indicates that their employment would have disproportionately declined in spring 2020 even if Latino/a workers in each occupation had suffered the same degree of job loss as their counterparts.

Figure 2 shows the effect of occupational segregation—i.e., the disproportionate employment losses accounted for by overrepresentation of workers of color in occupations with larger employment losses.

Loading chart 2...

Conversely, it is helpful to see what remains after the contribution of occupational segregation is removed from the equation. In Figure 3, we essentially subtract Figure 2 from Figure 1. If a group’s share of employment within an occupation is falling, this is reflected in a decline in Figure 3—the group is losing even more employment than would be expected from its pre-pandemic distribution across occupations (i.e., more than would have been expected from occupational segregation alone).

Loading chart 3...

Putting it all together

The most striking employment decline—attributable to occupational segregation—is evident for Latino/a workers, whose pre-pandemic pattern of employment is associated with a 2.8 percentage point excess decline through May 2020. That is, Latino/a workers tended to be in occupations—like chefs and waiters—that experienced unusually large employment declines. A gradual recovery brought Latino workers to 0.4 percentage points below baseline in October 2021. Occupational segregation is the principal driver of employment decline and recovery for Latino/a workers; Figure 3 shows only a small contribution from other factors.

By contrast, Black workers were not as exposed to the recession by their pre-pandemic occupational distribution, and segregation only contributed to a 1.0 percentage point excess decline through May 2020. But as shown in Figure 3, Black workers experienced large losses within occupations, leading them to have a much worse experience than occupational segregation alone would have predicted. For example, Black workers lost substantial ground in occupations like vehicle and equipment cleaners and file clerks. By October 2021, however, excess Black employment loss had recovered to 0.8 percentage points below the overall decline, roughly in line with what was expected due to pre-pandemic occupational segregation.

Small sample sizes force us to implement these calculations somewhat differently for Native American workers, and as a consequence we do not show them in the figures. But when we examine changes from the second quarter of 2019 to the second quarter of 2020, we find a substantial role for occupational segregation. Had Native American workers maintained their pre-pandemic occupation employment shares, they would have experienced a larger employment rate decline (4.0 percentage points) than actually occurred (2.3 percentage points). Native American workers actually gained employment share within occupations, partially offsetting their exposure to hard-hit professions. By the second quarter of 2021, that outperformance continued, such that Native American workers’ excess decline had shrunk to zero despite a continued disadvantage (3.1 percentage points) implied by pre-pandemic occupational segregation.

Asian workers had an experience that was more like that of Black workers than that of Latino/a workers. Occupational segregation made a relatively small contribution to their very large excess employment declines through the spring of 2020. But Asian workers suffered employment declines within occupations, losing ground within occupations like insurance underwriters and construction and building inspectors. However, by October 2021, this ground was recovered, and Asian employment was improving more rapidly than overall employment. Interestingly, in contrast to the beginning of the pandemic, their rising share of within-occupation employment is driving recovery in Asian workers’ employment.

Finally, the experience of White workers was quite similar to that of the labor force overall—unsurprisingly, given their large share of overall employment. White workers gained occupation share early in the pandemic, then lost those gains by October 2021.

As policymakers look to address the persistent racial inequities in our labor market, as well as the disproportionate economic harms suffered by workers of color during the pandemic, it is important to have a better understanding of how the two phenomena are related. Occupational segregation is an important part of that relationship. Though not the only source of labor market disadvantage, it is a substantial contributor that merits more attention by researchers and policymakers.

The authors thank Sara Chaganti and Janelle Williams for their valuable contributions to the development of this article.


1 The employment rate indicates simply whether a respondent has worked at all for pay or not in the last week. It does not capture variation in number of hours worked, hourly wage, or other characteristics of the job. It therefore provides a limited perspective on the disparate labor market impacts of the pandemic recession. When calculating the employed population, we include those who were at work during the week preceding their survey interview and those with a job who were not at work in the previous week. During the early months of the pandemic, there was some degree of miscoding of the unemployed as “employed but not at work” in the previous week. However, given the longer time horizon of our analysis, we maintain the traditional definition.

2 Small sample sizes—in combination with strong, distinctive seasonal employment patterns—led us to implement this calculation somewhat differently. For Native workers, we calculated the excess employment rate decline from April–June 2019 to April–June 2020.

3 We include all individuals 16 years of age and older. These estimates are not seasonally adjusted. However, our subsequent analysis focuses on “excess” employment rate changes (i.e., changes for a particular group relative to the overall change), which are likely less affected by seasonality.

4 Our analysis uses occupational codes from the U.S. Census Bureau, which categorizes all occupations into approximately 400 groupings.

5 More specifically, we estimate a counterfactual employed population for each racial group by multiplying each detailed occupation’s total employment by each racial group’s average share of that occupation in December 2019–February 2020. Dividing each racial group’s counterfactual employed population by its actual 16+ population for a given month yields the racial group’s counterfactual employment rate. We subtract this counterfactual rate (in any given month) from the employment rate during the baseline period (December 2019–February 2020). The population-wide change in employment rate—from the baseline period to any given month—is then subtracted from each racial group’s counterfactual change to determine the “excess” change, which is plotted in Figure 3.

Tyler Boesch
Data Scientist, Community Development and Engagement
Tyler Boesch analyzes data, develops visualizations, and creates statistical models to help the Community Development and Engagement team understand issues affecting low- and moderate-income communities. Before joining the Bank, he was a graduate research assistant with the University of Minnesota Center for Urban and Regional Affairs.
Ryan Nunn
Assistant Vice President, Community Development and Engagement
Ryan Nunn is an assistant vice president in the Minneapolis Fed’s Community Development and Engagement Department. Leading the Bank’s applied research function, Ryan works to improve outcomes for low- and moderate-income communities with the help of better evidence and analysis.
Alene Tchourumoff
Senior Vice President, Community Development and Engagement
Alene Tchourumoff, senior vice president of Community Development and Engagement and the Center for Indian Country Development, leads the Bank’s efforts to promote economic opportunity and access to credit for low- and moderate-income people throughout the Ninth District and for residents of Indian Country.