Women have been hit hard by the coronavirus labor market: Their story is worse than industry-based data suggest

Key findings:

  • The latest payroll employment data for March show that women were the hardest hit by initial job losses in the COVID-19 labor market; women represented 50.0% of payroll employment in February, but represented 58.8% of job losses in March.
  • If women’s share of new unemployment insurance (UI) claims in recent weeks was driven solely by sector-level differences in gender composition, then they would have accounted for roughly 45% of new UI claims, or about 6.8 million new claims.
  • However, relying solely on the gender composition of sectoral unemployment may lead to an underestimate of new UI claims that were filed by women. Using three states that provide direct estimates of the gender composition of new UI claims shows that the female share of these claims is substantially higher than what we estimate by using only the sectoral composition of employment by gender.
  • We estimate that once the overrepresentation of women in sectors with new layoffs is corrected for, between 7.8 and 8.4 million women filed for unemployment insurance in the three weeks ending April 4.

Since March 15, 15.1 million workers in the United States have filed for unemployment insurance. Tomorrow, the latest initial unemployment insurance claims will be released by the Department of Labor for the week ending April 11, and estimates suggest that there could be another 4.5 million initial claims reported. These top-line numbers are vital for understanding what is going on in the economy and the extent of the economic insecurity millions of workers and their families are experiencing. But what is less clear is who these workers are and where they work. While national statistics that directly report the demographic characteristics of UI claimants will not be available for months, we use national employment data from March and preliminary state UI reports through April to begin to answer those questions. We find that job losses and furloughs have disproportionately affected women. This is the result of two factors: Women are more concentrated in sectors that experienced more job loss, and women also tended to see more job loss than men within these sectors.

We begin by using the Current Employment Statistics to examine initial job losses by sector and gender. Next we use UI claims data from selected states to estimate the share of UI claims filed by women. Eleven states, which represent about 20% of the U.S. workforce, have reported valuable sector-level data on their unemployment insurance claims. We pair this with gender-, state-, and sector-level data from the Current Population Survey (CPS) to estimate the number of women who likely filed initial unemployment insurance claims since March 15. In conducting our study, we found that women are likely overrepresented in initial job losses as compared with results from the three states with reported gender-specific UI claims data. This overrepresentation could be due to the types of jobs women hold within each sector—partially the result of occupation segregation—or it could be due to other phenomena (for instance, labor market discrimination) that have led to disproportionate shares of women being in jobs within each sector that are more subject to job loss. Either way, it is clear that using sector-level data to estimate the number of women who filed unemployment insurance claims in the last three weeks underestimates the actual number of women experiencing job loss and filing for unemployment.

Let’s start with the latest Employment Situation report from the Bureau of Labor Statistics, providing data for just the beginning of the COVID-19 job market losses. The data for March represented the payroll period including March 12, before the first spike in initial unemployment insurance claims. Payroll employment still fell by 701,000 jobs in March. Using February 2020 as a benchmark, Table 1 shows job shares by gender as well as job losses that occurred in March by gender for selected sectors. While jobs filled by women represent 50.0% of payroll employment in February, jobs held by women represented 58.8% of job losses in March. The vast majority of the job losses occurred in leisure and hospitality, with a loss of 459,000 jobs. In this sector, jobs held by women represented 53.3% of overall employment and 56.9% of job losses. Disproportionate job losses were found in nearly all the largest sectors that experienced significant job losses in March. Women were overrepresented in job losses experienced in retail trade and professional and business services. In education and health services, jobs held by men actually increased in March, while jobs held by women fell by 86,000.

Table 1

Establishment-level data suggests women are overrepresented in job losses: Current Employment Statistics, selected sectors, by gender, February and March 2020

Industry Gender Shares of total employment Job losses in March (thousands) Shares of March job losses
Nonfarm payrolls Men 50.0% -289 41.2%
Women 50.0% -412 58.8%
Construction Men 87.0% -28 96.6%
Women 13.0% -1 3.4%
Manufacturing Men 71.3% -12 66.7%
Women 28.7% -6 33.3%
Retail trade Men 50.5% -11 23.4%
Women 49.5% -35 76.6%
Transportation and warehousing Men 74.1% 2 -46.9%
Women 25.9% -7 146.9%
Leisure and hospitality Men 46.7% -198 43.1%
Women 53.3% -261 56.9%
Professional and business services Men 54.1% -23 44.2%
Women 45.9% -29 55.8%
Education and health services Men 22.6% 10 -13.2%
Women 77.4% -86 113.2%

Source: EPI analysis of Bureau of Labor Statistics Current Employment Statistics public data series

Copy the code below to embed this chart on your website.

Overall, the establishment-level data show that women were overrepresented in job losses, at least in the initial weeks of the COVID-19 labor market.

Now, let’s turn to the state-level unemployment insurance claims data available. Eleven states, which represent about 20% of the U.S. workforce, have reported valuable sector-level data on their unemployment insurance claims, and three states—Minnesota, North Dakota, and Nevada—have released gender-specific breakdowns of initial UI claims. Using these data, we estimate that women likely filed more UI claims than men, especially during the earlier weeks of the pandemic. This is the result of two factors—women are more concentrated in sectors that saw more job loss, and women also tended to see more job loss than men within sectors.

The first column in Table 2 shows that most UI initial claims in Minnesota were filed by women, with women filing larger shares of claims in the early weeks of the pandemic. The second column of Table 2 shows what share of UI claims one would expect should come from women, using a prediction of industry-specific job losses and assuming industry-specific UI claims are distributed according to the share of women who work in those sectors.1 What is immediately apparent upon comparison is that our estimate based solely on the sectors impacted significantly understates the number of women who filed for unemployment insurance in Minnesota. The pattern holds for North Dakota and Nevada, the two other states that published gender-specific UI data.

This suggests that any estimate based on shares of women in each sector may undercount the number of women who filed for unemployment insurance in the last four weeks. Recall that the establishment survey suggested that 58.8% of job losses were among jobs held by women, and disproportionately even within each sector. The outsized job losses among women weren’t simply because of the overall sectors where job losses occurred, but the occupations or particular jobs within those sectors were skewed toward more losses among women workers. The results below affirm those findings and suggest that using sector-level data from each state alone underreports gender losses by 11% to 24% depending on the state or specific week. Also, in all three states, the data suggest that women make up a slightly higher share of all initial claimants. In Minnesota and North Dakota, it appears that the share of initial claims by women for the week ending March 21 was higher than in subsequent weeks, supporting the prior findings from the establishment survey that women made up a larger share of initial job losses.

Table 2

Sector-level gender shares underestimate state-reported unemployment insurance data: State-reported UI claims data and imputed data, by gender, March 15 through April 4, 2020

Minnesota North Dakota Nevada
State UI data as reported Sector-based UI prediction State UI data as reported Sector-based UI prediction State UI data as reported Sector-based UI prediction
Week ending March 21 62.8% 47.9% 55.1% 45.6% NA 45.7%
Week ending March 28 54.2% 44.1% 53.2% 41.3% 50.6% 43.4%
Week ending April 4 50.5% 44.7% 51.0% 41.2% 51.4% 43.3%
Three-week average 55.9% 45.6% 52.5% 42.0% 51.1% 44.2%

Note: Sector-based UI estimates use the gender shares within each sector for each state and apply those directly to the sectors with UI claims.

Source: EPI analysis of state-level gender-specific UI initial claims data for Minnesota, North Dakota, and Nevada; sector-specific initial UI claims data for Alabama, Kansas, Maine, Massachusetts, Michigan, Nebraska, North Dakota, Nevada, New York, Oregon, and Washington; and EPI Current Population Survey Extracts, Version 1.0 Version 1.0 (2020), https://microdata.epi.org

Copy the code below to embed this chart on your website.

Using what we’ve learned from these comparisons, we now turn to our estimates of unemployment insurance claims for women nationally and across states. We sum our state- and sector-specific claims predictions across all states to get our initial estimate for the share and number of women filing for UI claims as shown in the first two rows of Table 3. Next, we use the fact that sector-specific predictions underreport actual gender shares of UI claims in the three states for which we have data, as discussed re Table 2 above. The underreporting ranges from 6.9 to 10.5 percentage points. We use this range to provide lower- and upper-bound estimates of the share and level of women filing UI claims since March 15. According to our analysis, women represent 52.2% to 55.8% of overall claims and 7.8 to 8.4 million claimants.

Women may continue to be overrepresented among claimants, more so than their sector averages would suggest. This may be true for the reasons already stated—women hold jobs within sectors that may be materially different from the jobs men hold. For instance, if women are less likely to be managers within an establishment and all but the high-level managers are laid off, women will be more likely to be laid off. In addition, the CARES Act greatly expanded eligibility for unemployment insurance, including provisions to allow workers to claim benefits if workers had to leave for a variety of reasons related to COVID-19 or state-level social distancing requirements. Workers are eligible for UI because of caregiving responsibilities, that is, they can receive benefits if they need to quit to care for a child whose child care or school is closed. Considering caregiving responsibilities are often disproportionately borne by women, this could mean that they continue to shoulder greater job losses and a larger share of unemployment insurance claims.

We will continue to examine the data as it becomes available and analyze as many demographic characteristics as possible to better understand who across the United States is bearing the brunt of the job losses.

Table 3

Estimating the share and level of women’s UI claims using available data: Sector-based UI claims adjusted by predicted gender shares, March 15–April 4, 2020

Women with UI claims
Sector-based share of UI claims 45.3%
Sector-based level of UI claims 6,798,836
Lower bound, predicted share 52.2%
Upper bound, predicted share 55.8%
Lower bound, adjusted level 7,834,420
Upper bound, adjusted level 8,374,725

Source: EPI analysis of Bureau of Labor Quarterly Census of Employment and Wages data series; EPI Current Population Survey Extracts, Version 1.0 (2020), https://microdata.epi.org; state-level gender-specific UI initial claims data for Minnesota, North Dakota, and Nevada; and sector-specific initial UI claims data for Alabama, Kansas, Maine, Massachusetts, Michigan, Nebraska, North Dakota, Nevada, New York, Oregon, and Washington, retrieved from each state's employment office

Copy the code below to embed this chart on your website.


1. To create this prediction, we first take industry-specific UI claims from the 11 states that have reported them and calculate average industry-specific UI claim shares of 2019Q3 QCEW employment. Then, we apply these shares to Minnesota’s industry-specific employment totals, proportionally scaling the resulting industry-specific estimates of Minnesota UI claims so that the total matches Minnesota’s reported initial UI claims. Finally, we allocate the resulting predicted industry-specific UI claims to men and women by assuming the UI claims are distributed by gender in the same way as in Minnesota’s sector-specific employment, as measured by state-sector-specific female shares of employment in the 2017–2019 basic monthly Current Population Survey. The 11 states that reported relatively complete industry-specific weekly UI initial claims at the 2-digit NAICS level found in state-level employment offices are Alabama, Kansas, Maine, Massachusetts, Michigan, Nebraska, North Dakota, Nevada, New York, Oregon, and Washington.