- Issue: During the COVID-19 pandemic, most states issued lockdown orders that closed many workplaces. The ensuing job losses may have left millions of workers without employer health coverage.
- Goal: To estimate how many workers lost jobs that came with employer-sponsored insurance (ESI) — by industry, age, and gender — during the pandemic.
- Methods: Health insurance coverage data were used to generate the proportion of workers with ESI, by various characteristics. Data on unemployment benefit recipients were used to generate the proportion of workers who lost jobs because of the pandemic. We apply the proportion of workers with ESI to the number of workers who lost jobs to obtain an estimate of jobs with ESI coverage that were lost. We also determine the number of dependents of these workers who potentially lost coverage.
- Key Findings and Conclusion: We estimate that as many as 7.7 million workers lost jobs with ESI as of June 2020 because of the pandemic-induced recession. The ESI of these workers covered 6.9 million of their dependents, for a total of 14.6 million affected individuals. Only with time will we know how many job losses are ultimately permanent, resulting in loss of ESI for workers and their dependents.
In response to the COVID-19 pandemic, most states issued lockdown orders that closed many workplaces and dramatically slowed U.S. economic activity in the spring of 2020. The result was a massive increase in unemployment, which peaked in April at 14.7 percent. During the 15 weeks from mid-March to the end of June, Americans filed nearly 49 million new claims for unemployment benefits.1
The strong link between employment and health insurance coverage has important implications for Americans’ insurance coverage and access to health care. Employer-sponsored health insurance (ESI) is the most common form of health insurance in the United States. In March 2019, 69 percent of the 152 million workers age 16 and older had ESI, meaning that 175 million workers and their dependents had coverage.2 But if millions of workers and their dependents have lost ESI during the pandemic, we would expect increased enrollment in COBRA (Consolidated Omnibus Budget Reconciliation Act) continuous coverage, Affordable Care Act (ACA) marketplace plans, and Medicaid, as well as an increase in the number of uninsured.
The path from loss of a job with ESI to loss of health insurance is not simple and depends on three factors:
- Is the job loss temporary or permanent, and if temporary, does the employer continue ESI coverage until the worker is called back to work?
- To what extent will temporary layoffs without loss of ESI become permanent layoffs with loss of ESI?
- When workers do lose ESI, either at the time of layoff or when a temporary layoff becomes permanent, how many will obtain coverage through other family members, COBRA, the ACA marketplace, or Medicaid?
We examine these factors in more detail in the following sections.
Maintaining Coverage During Temporary Job Loss
A Commonwealth Fund survey of nearly 2,300 adults age 18 and older found that, of workers who had lost jobs with ESI during the pandemic, more than half had been temporarily laid off or furloughed and still had ESI through the employer.3 This makes sense because temporarily laid-off workers remain attached to their employer, and the employer has an interest in the continued health and future productivity of these workers.
Losing ESI When Layoffs Become Permanent
At the outset of the recession in April, nearly 18.2 million of the 23.1 million workers who were unemployed were temporarily laid off or furloughed and expected to be recalled to their previous employer.4 By August, the number of temporarily laid-off workers had fallen to 6.2 million as many workers returned to their jobs. But during the same time, the number of permanently laid-off workers increased from 2.6 million to 4.1 million, and the number of workers unemployed for 15 or more weeks increased from 1.8 million to 8.1 million. Although improvements in the labor market since April have been good news for many laid-off workers, a large minority has already lost or remains at risk of losing ESI.
Obtaining Coverage After Loss of ESI
The Urban Institute estimates that, during the last three quarters of 2020, on average only about a third of those who lose ESI coverage through pandemic-related job loss will become uninsured.5 About a third will obtain coverage through another family member’s ESI, just over a quarter will become covered by Medicaid or the Children’s Health Insurance Program (CHIP), and a small percentage will obtain nongroup insurance. This implies that, even for those who lose ESI following the loss of a job with ESI, only a minority will lose health coverage.
All three of these factors involve much uncertainty, and any estimate of losses of ESI or any health insurance coverage resulting from the pandemic recession must be understood as provisional.
In this issue brief, we use a combination of health insurance coverage data and state-level data on the characteristics of unemployment benefit recipients to estimate how many workers have lost jobs with ESI following the recent surge of unemployment, and break that number down by industry, age, and gender. We also examine how many of these workers’ dependents may have lost coverage.
Exhibit 1 shows that, between February and June 2020, the number of unemployed individuals increased by 15.9 million (from 2.0 million to 17.9 million). This is a reasonable estimate of the job loss resulting from the pandemic, and it represents 10 percent of prepandemic employment (Exhibit 2).6
We estimate that, of these newly unemployed workers, 7.7 million lost jobs with ESI, and these job losers had 6.9 million dependents who were covered by ESI. As a result, 14.6 million individuals in total either lost a job with ESI or were the covered dependent of a job loser with ESI.7
The pandemic-related lockdowns severely affected certain industries and groups of workers, while leaving others virtually untouched. As a result, we would expect the number of people losing jobs with ESI to vary greatly by industry, and possibly by other characteristics, such as age and gender.
ESI Coverage Loss by Industry
The seven industries shown in Exhibit 3 accounted for two-thirds of prepandemic employment, 69 percent of unemployed workers in June, and 68 percent of lost jobs with ESI. The disparities among these industries in total job loss versus loss of jobs with ESI are striking. For example, total job losses in manufacturing were roughly proportional to employment — manufacturing accounted for 10 percent of prepandemic employment and 12 percent of unemployed workers in June. But because manufacturing has one of the highest rates of ESI coverage (66%),8 it accounted for a greater proportion of loss of jobs with ESI (18% of lost jobs with ESI and 19% of potential ESI coverage loss when dependents are included).
In contrast, job losses in accommodation and food services were far out of proportion to employment. Although accounting for only 7 percent of prepandemic employment, 20 percent of unemployed workers in June were former accommodation and food service workers. But because only 25 percent of these workers had ESI prepandemic, they accounted for only 11 percent of lost jobs with ESI and 10 percent of potential ESI coverage loss when dependents are included.
Similarly, retail trade accounted for 10 percent of prepandemic employment and 14 percent of unemployed workers in June. But because only 40 percent of workers in retail trade had ESI prepandemic, these workers accounted for only 12 percent of lost jobs with ESI and 11 percent of potential ESI coverage loss including dependents.
Exhibit 4 shows further detail on the pandemic-related increase in unemployment by industry, along with the percentage of workers in each industry who became unemployed, the percentage of workers with prepandemic ESI coverage, and the percentage of those workers who lost jobs with ESI.
Nearly 3.3 million workers in accommodation and food services (30% of the industry’s workforce) became unemployed between February and June 2020. However, only 25 percent of workers in that industry had ESI prepandemic, so only 7 percent of accommodation and food service workers lost jobs with ESI as a result of shutdowns. In contrast, far fewer manufacturing workers lost jobs following the pandemic (about 1.6 million, or 10% of employment in the industry). But because two-thirds of manufacturing workers had ESI coverage prepandemic, the same proportion (7%) of all manufacturing workers lost jobs with ESI as a result of the shutdowns.
ESI Coverage Loss by Age
Exhibit 5 shows that workers ages 35 to 44 and 45 to 54 bore the brunt of ESI-covered job losses. They accounted for 17 percent to 19 percent of workers who lost jobs, but 22 percent to 27 percent of potentially affected individuals (workers plus their dependents) because workers in these age groups were the most likely to be covering spouses and other dependents. In contrast, workers younger than age 25 and age 65 and older accounted for disproportionately small shares of lost jobs with ESI because of their relatively low rates of ESI coverage.
The case of workers younger than 25 is striking because, although they accounted for 12 percent of prepandemic employment and for 16 percent of unemployed workers, they represented only 7 percent of lost jobs with ESI, and only 5 percent of potential ESI coverage loss including dependents. Only one in five of these younger workers had ESI through their job, and only 14 percent provided coverage for a dependent.9
Workers age 65 and older accounted for only 6 percent of unemployed workers, and for only 4 percent of lost jobs with ESI because most older workers have Medicare as their primary coverage. Only 35 percent had coverage from an employer, and some of these workers have retiree health coverage through a former employer.
ESI Coverage Loss by Gender
Exhibit 6 shows that the adverse effects of the pandemic recession fell disproportionately on women. Although women made up 47 percent of prepandemic employment, they accounted for 55 percent of total job losses. But because women were somewhat less likely than men to have health coverage through their own job, and less likely to have family coverage, they and their dependents accounted for slightly more than half of all lost jobs with ESI and potential ESI coverage loss when dependents were included.
Between February and June 2020, the number of unemployed workers increased by 15.9 million. We estimate that about 7.7 million (48%) of these workers lost jobs with ESI. And because 6.9 million dependents were covered by ESI through a job loser, a total of 14.6 million individuals potentially lost ESI as of June because of the pandemic. If all these individuals lost ESI coverage, it would represent an 8 percent reduction in total ESI coverage.
The ESI coverage rate of individuals who lost jobs because of the pandemic (48%) is somewhat less than the ESI coverage rate of workers overall (51%) (Exhibit 4). This is because workers employed in two of the three industries most affected by lockdowns — accommodation and food services, and retail trade — were less likely than average to have ESI coverage (Exhibit 3). Workers in manufacturing — the third industry heavily affected by the lockdowns — were more likely than average to have ESI coverage.
These estimates of lost jobs with ESI and dependents covered by the job losers do not imply that 14.6 million individuals have lost ESI or become uninsured during the pandemic. Rather, they represent an upper-bound estimate of ESI losses because the only available estimate suggests that roughly half of workers who have lost jobs with ESI have been furloughed or temporarily laid off and have continued to be covered by ESI.10 Estimating the number of individuals who have lost ESI because of the COVID-19 recession requires estimates of the number of job losers, the number of job losers with ESI, and the number of job losers with ESI whose coverage was not continued by their employer. Considerable uncertainty surrounds all these estimates.
Uncertainly also surrounds estimates of the number of individuals who have lost any health insurance coverage as a result of lost ESI. As discussed earlier, an Urban Institute study estimated that roughly a third of those who lose ESI coverage through pandemic-related job loss will become uninsured. We caution, however, about the difficulties in making such estimates.11
As previously discussed, the path from loss of a job with ESI to loss of health insurance is not simple. It follows that the evolution of ESI coverage during the pandemic will be far from straightforward and will depend on many unpredictable circumstances. Only with time will we know how many job losses are ultimately permanent and result in loss of ESI. In the interim, it will be important to monitor key labor market statistics, including the number of workers on furlough or temporary layoff who become permanent job losers, and the number of job losers who have been unemployed for 15 weeks or more and are unlikely to be attached to an employer and covered by ESI.
How We Conducted This Study
The data for this study come from two sources. Data on health insurance coverage come from the 2019 Annual Social and Economic Supplement to the Current Population Survey (CPS). The CPS is a primary source of labor force statistics for the U.S. population and is cosponsored by the U.S. Census Bureau and the U.S. Bureau of Labor Statistics. In March 2019, the CPS interviewed nearly 95,000 households, covering 180,000 individuals. The March CPS collects information on health insurance coverage for both the prior calendar year and at the time the survey is fielded. Data on whether workers had employee-only, employee-plus-one, or family coverage are also available.
Data on unemployed workers come from “Characteristics of the Unemployment Insurance Claimants,” a monthly report published by the Employment and Training Administration of the U.S. Department of Labor.12 These monthly characteristics reports have rarely if ever been used in research and should not be confused with the familiar “Unemployment Insurance Claims Report” (or “weekly claims report”). The weekly claims report is a bare-bones report that provides simple weekly counts of the number of initial and continued unemployment insurance claimants in each state. In contrast, the characteristics report is a detailed monthly report of the industry, occupation, and demographic characteristics of unemployment recipients in each state.13
The characteristics report is based on random samples of “continued claimants” that each state draws during the week containing the 19th of the month. These continued claimants are unemployed workers who have filed and been determined eligible for unemployment benefits, have experienced at least one week of unemployment, and have filed a continued claim for unemployment benefits in a subsequent week. We refer to these individuals simply as unemployed workers or job losers, although they do not include jobless workers who are ineligible for benefits or newly unemployed workers who are eligible.
To understand the monthly characteristics report, it is important to keep in mind the distinction between initial and continued claimants. When a worker is first laid off, he or she files an initial claim, which establishes a benefit year lasting for the next 52 weeks. The initial claim is filed only once, and the claimant becomes a continued claimant if he or she is eligible for benefits and files a continued claim (usually by phone or by logging into the unemployment insurance agency’s website) every other week. So initial claims represent the inflow of potential new recipients, whereas continued claims represent the stock of currently eligible claimants who have been unemployed for at least one week.
We use the CPS and Labor Department data together to estimate the number of workers and their dependents who lost jobs with employer-sponsored insurance (ESI) following the wave of unemployment that started during the week ending March 14. We first use the CPS to generate the proportion of workers and dependents with ESI in March 2019, by industry, age, and gender. We assume that the likelihood of having ESI in March 2020 was the same as it was in March 2019. We then estimate the number of workers who lost jobs by comparing the Labor Department’s monthly characteristics reports for February and June, by the same characteristics. We apply the proportion of workers with ESI to the number of workers who lost jobs to obtain an estimate of the number of workers who lost a job with ESI.
Finally, for each worker with health insurance, we determine the number of dependents who were covered by the ESI of a worker who lost a job with ESI. We do this by examining whether the worker had employee-only, employee-plus-one, or family coverage. Among workers with family coverage, the average family size was 3.89 individuals.
Appendix. Strengths, Limitations, and Comparisons with Other Studies
We believe that use of the Labor Department’s monthly “Characteristics of the Unemployment Insurance Claimants” report is a strength of the analysis because it allows us to reliably observe the characteristics of workers who have lost their jobs since the start of the pandemic-related lockdowns. The monthly unemployment insurance characteristics report is preferable to the monthly Current Population Survey (CPS) household sample because it is so much larger and because it is based on administrative records maintained by the states. The CPS interviews about 60,000 households per month, though in some months the sample is much larger. The data must be weighted to obtain national estimates. These estimates may include substantial sampling errors, so differences less than 500,000 are not considered statistically significant. However, this is not a concern with the unemployment insurance characteristics reports, which are based on large samples of state administrative data.
Also, the administrative data that are the basis for the Labor Department’s monthly characteristics report are generally viewed as more reliable than the CPS’s household data, which relies on self-reported employment status. The usual practice of the Bureau of Labor Statistics (BLS) is to accept answers as recorded and not reclassify survey responses. An example of the problems raised by self-reports occurred in early 2020, when there was a large increase in the number of people who were classified as “employed but absent from work.” BLS determined that many of these workers should have been classified as “unemployed on temporary layoff,” in which case the unemployment rate would have been 1 percentage point higher.
Like any empirical analysis, this one has limitations, and two are worth emphasizing. First, it has long been known that take-up of unemployment insurance is incomplete — many job losers who could receive benefits do not claim them. This is usually because they believe they will be reemployed quickly, and the time-costs of claiming outweigh the benefits. We believe this is not likely to be a major problem in this context because the rates of claiming have been extremely high during the pandemic. The reasons for the unusually high claim rates include the additional $600 per week provided to unemployment insurance recipients under the CARES Act through July 25, 2020, and the likelihood that recently laid-off workers are lower-wage workers who are more likely to claim benefits, even absent the $600 per week CARES add-on.
Second, although the CPS provides the proportion of ESI-covered workers by industry and by demographic group, our analysis assumes that the proportion of workers who lost jobs in each industry and demographic group is the same. Job losers usually are disproportionately less experienced and have less education than workers who remain employed during a recession, so it is likely that the job losers in each industry and demographic group were disproportionately not covered by employer-sponsored health insurance (ESI). This is one reason the estimates we obtain of lost jobs with ESI coverage should be viewed as an upper bound.
One manifestation of this second limitation is that using different characteristics (for example, gender rather than industry) can produce somewhat different estimates of the total number of workers who have lost ESI because of the pandemic. For example, when we use industry classification, we estimate that about 7.4 million workers lost jobs with ESI, whereas when we use gender classification, the estimate rises to nearly 8.4 million. (These are the lowest and highest estimates we obtain.) Because of this limitation, we use the average estimate across the variables examined in this brief.
Comparisons with Other Studies
A recent Urban Institute article has reviewed and reconciled the wide range of estimates from recent studies of the relationship between the COVID-19 recession and ESI losses.14 Accordingly, we discuss these earlier studies only briefly.
Using data on new unemployment insurance claimants, the Economic Policy Institute estimates that 16.2 million workers lost ESI as of early May 2020.15 This study did not estimate the number of dependents losing coverage.
Other studies estimate both the number of workers losing ESI and the number of dependents. A Kaiser Family Foundation study estimates that as of May 2020, nearly 27 million workers and dependents lost ESI and became uninsured.16 Another 19 million lost ESI but were able to switch to a job-based plan through a family member.
Health Management Associates estimates that between 12 million and 35 million people could lose ESI, although as the Urban Institute review points out, the data and methods behind this range of estimates are undocumented.17 In contrast, a well-documented Urban Institute study estimated that, on average from April 2020 to December 2020, 7.3 million workers and dependents will lose employer-based coverage as a result of the recession.18 Families USA released a related report, which found that 5.4 million laid-off workers became uninsured; however, this report did not provide an estimate of the number of workers losing ESI.19
The authors thank the Commonwealth Fund’s Sara Collins, David Blumenthal, Elizabeth Fowler, and Christopher Hollander for valuable feedback.
1. For comparison, between mid-March and the end of June 2019, there were 3.3 million new claims for unemployment benefits; and during the first 15 weeks of the 2007–2009 recession, there were 5.2 million new benefit claims.
2. Authors’ estimates of the 2019 Annual Social and Economic Supplement to the Current Population Survey.
3. Sara R. Collins et al., An Early Look at the Potential Implications of the COVID-19 Pandemic for Health Insurance Coverage (Commonwealth Fund, June 2020).
4. The labor market data in this paragraph come from “The Employment Situation — August 2020,” news release, U.S. Department of Labor, Bureau of Labor Statistics, Sept. 4, 2020. See Tables A-1, A-10, and A-11.
5. Jessica Banthin et al., Changes in Health Insurance Coverage Due to the COVID-19 Recession (Urban Institute, July 2020).
6. Our findings also mirror the estimates of reduced employment (and increased unemployment) from the CPS household survey and the Current Employment Statistics (CES) establishment survey. Specifically, for February to June 2020, the CPS shows a drop in employment of 16.6 million (from 158.8 million to 142.2 million) and an increase in unemployment of 12.0 million (from 5.8 million to 17.8 million). The CES establishment survey shows a drop in total nonfarm payroll employment of 14.7 million (from 152.5 million to 137.8 million). For a discussion of why these estimates differ, see Tomaz Cajner et al., Reconciling Unemployment Claims with Job Losses in the First Months of the COVID-19 Crisis, Finance and Economics Discussion Series 2020-055 (Board of Governors of the Federal Reserve System, July 2020).
7. These estimates do not include anyone who took COBRA. We expect COBRA take-up to be very small at this point. The CARES Act, passed in March 2020, extended the time period for COBRA elections. Specifically, qualified individuals have until 60 days after the end of the national emergency to elect COBRA coverage under the CARES Act. Our estimates also do not account for the fact that individuals who lost employment-based health insurance may be able to sign up for it through another family member.
10. Collins et al., An Early Look, 2020.
11. Banthin et al., Changes in Health, 2020.
12. “Characteristics of the Unemployment Insurance Claimants,” U.S. Department of Labor, Employment and Training Administration, Aug. 27, 2020.
13. As of September 8, 2020, Michigan had not reported for June. Therefore, we use May 2020 claimants for Michigan as a placeholder for June 2020.
14. Jessica Banthin and John Holahan, Making Sense of Competing Estimates: The COVID-19 Recession’s Effects on Health Insurance Coverage (Urban Institute, Aug. 2020).
15. Ben Zipperer and Josh Bivens, “16.2 Million Workers Have Likely Lost Employer-Provided Health Insurance Since the Coronavirus Shock Began,” Working Economics Blog, Economic Policy Institute, May 14, 2020.
16. Rachel Garfield et al., Eligibility for ACA Health Coverage Following Job Loss (Henry J. Kaiser Family Foundation, May 13, 2020).
17. Health Management Associates, COVID-19 Impact on Medicaid, Marketplace, and the Uninsured, by State (HMA, April 3, 2020).
19. Stan Dorn, The COVID-19 Pandemic and Resulting Economic Crash Have Caused the Greatest Health Insurance Losses in American History (Families USA, July 2020).