Racial And Ethnic Disparities In Pandemic-Era Unemployment Insurance Access: Implications For Health And Well-Being

Abstract

Research demonstrates that receiving unemployment insurance decreases mental health problems. But researchers have also found racial and ethnic disparities in unemployment insurance receipt resulting from differences in work history and location. We examined a population disproportionately affected by job loss and unemployment insurance exclusions, using a survey of service workers from a single city who were parents of young children and who overwhelmingly had eligible work histories. During the COVID-19 pandemic, workers not identifying as White non-Hispanic in our sample were more likely to get laid off than White workers. Among those who were laid off, these workers and White workers experienced similar increases in material and mental health difficulties and similar gains when they received unemployment insurance. However, these workers were less likely than White workers to receive unemployment insurance at all. These results indicate that unemployment insurance has unrealized potential to reduce material and health disparities. Policies should be implemented to make this coverage more effective and equitable through increased access.

The COVID-19 pandemic brought economic disruption, including mass layoffs and unemployment. This disruption contributed to worldwide financial hardship and food insecurity, which are major factors in the physical health of populations. It also likely contributed to the dramatic increase in mental health disorders during the pandemic,^1,2 given that a large, robust literature links job loss and unemployment to worse mental health.^3–5 Job displacement and hardship during the pandemic have, moreover, been unevenly distributed, with people who identify as other than non-Hispanic White and lower-wage workers bearing the brunt of them.^6–9

In response to concern about unemployment and economic hardship caused by the pandemic, US policy makers expanded unemployment insurance, a federal and state partnership that provides partial wage replacement to individuals who have lost their jobs and is the main public policy for ameliorating the economic effects of unemployment. These expansions included Pandemic Unemployment Assistance and Pandemic Emergency Unemployment Compensation; for brevity, we refer to all such expansions under the umbrella term “unemployment insurance.” The pandemic-related changes to unemployment insurance were intended to make it both more accessible (to, among others, those whose low earnings would traditionally render them ineligible) and more generous during this public health crisis, with the potential to mitigate health-related harms and widening health disparities by reducing economic hardship. At the same time, however, the pandemic created barriers to program access as offices closed and many telephone lines were overwhelmed, meaning that on net, accessibility did not necessarily increase.

An emerging body of research conducted in the US both before and during the COVID-19 pandemic has focused on whether and how government support for displaced workers, and in particular unemployment insurance, alleviates the harms caused by unemployment, including impacts on physical and mental health.^10–12 Evidence from before the pandemic suggests that such support can promote health by reducing poverty and economic hardship, that receipt is related to better physical and mental health, and that receipt can buffer the negative effects of unemployment on health.^13–15 Studies of the pandemic period specifically have shown that the expanded unemployment insurance reduced material hardship and mental health problems, with evidence that the $600 unemployment insurance supplement provided in summer 2020 had particularly strong effects on material hardship.^11,16

Government support through unemployment insurance has had unequal reach, however. Because of restrictive minimum earnings requirements, bureaucratic red tape, and disincentives for employer cooperation, traditional unemployment insurance in the US has primarily helped higher-earning, full-time workers who are disproportionately White while underserving lower-earning workers who are disproportionately Black, Latinx, and people who identify as other than White non-Hispanic.^17–19 For this reason, to the extent that government support has protected recipients’ health, lack of access for marginalized populations may have contributed to resource and mental health disparities—a topic that has yet to be explored in this developing research area.

In this article we describe a study of unemployment insurance receipt in a population of low-wage service workers with young children, a group that consists primarily of people who do not identify as White non-Hispanic. Their challenges are of substantial public health and policy concern, as low-wage workers, displaced workers, and parents of young children are at the nexus of negative resource and mental health impacts of the pandemic, and because racial disparities in government supports could substantially hinder policy efforts to aid these hard-hit groups.

Study Data And Methods

SAMPLE RECRUITMENT

Details of the sample recruitment are available elsewhere.³ Briefly, people were eligible if they worked in an hourly position in a retail, food service, or hotel business in the city of Philadelphia, Pennsylvania; had a child ages 2–7; and had a mobile phone that could send and receive basic SMS text messages. All participants reported working for the establishments at which they were recruited, with the exception of a small number who reported being employed in food service or retail by a larger entity (for example, a university or hospital) in which their establishment was located. The sample was originally recruited between August and November 2019 for a study examining parents’ work schedule unpredictability and family well-being, using a venue-based sampling approach—a commonly used technique for producing generalizable samples of hard-to-reach, unrostered populations.²⁰

DATA COLLECTION

When first recruited for the original study purpose, participants were asked to complete thirty daily surveys as well as a one-time survey about demographics, household characteristics, and mental health. All aspects of this study received human subjects approval from the Duke University Institutional Review Board.

Additional surveys were conducted throughout the pandemic. The current study used data from four waves collected during the following periods: May–June 2020, October 2020-January 2021, March–June 2021, and July–November 2021. At each wave, all sample members (who had not withdrawn from the study) were invited to participate in a one-time survey. Data were collected via basic SMS text messages, with an option to instead complete the survey online via Qualtrics. Study materials were available in English and Spanish.

Our analytical sample included 749 people who responded to at least one of the surveys (79 percent of the sample responded in May–June 2020, 70 percent in October 2020-January 2021, 84 percent in March–June 2021, and 75 percent in July–November 2021). We used 2,174 total observations (person-waves) for analysis.

MEASURES

LAYOFFS AND UNEMPLOYMENT INSURANCE RECEIPT:

The surveys asked respondents whether they were currently working for pay and, if not, whether they had been laid off; a dichotomous indicator was created equal to 1 if the respondent was laid off and 0 otherwise. Respondents who reported layoff were asked about unemployment insurance application and receipt. Although question wording varied slightly between waves (see online appendix 1 for survey instruments),²¹ responses enabled us to create two dichotomous indicators: the first equal to 1 if the respondent reported having applied for unemployment insurance and 0 otherwise, the second equal to 1 if the respondent reported that they were currently receiving unemployment insurance and 0 otherwise. Those who received unemployment insurance while the supplemental Pandemic Emergency Unemployment Compensation unemployment insurance benefits were available were asked whether they had received the extra benefit amount (either $300 or $600 per week, depending on when they were surveyed). For regression analyses, we created a measure equal to 1 if they received $300, 2 if they received $600, and 0 if no supplement was received, meaning that the regression coefficient reflects the effect of an additional $300 per week on the outcome.

MENTAL HEALTH:

The main outcomes of interest were depressive and anxiety symptoms, measured with the Patient Health Questionnaire (PHQ-2)²² and Generalized Anxiety Disorder 2-item questionnaire (GAD-2),²³ respectively. The PHQ-2 and GAD-2 each ask two questions about symptom frequency. Total scores, ranging from 0 to 6, were converted to dichotomized indicators, with a total score of 3 or more indicating likely, clinically meaningful depression or anxiety.^22,23

INCOME LOSS AND MATERIAL HARDSHIP:

Surveys asked how respondents’ total household income, including unemployment insurance, Child Tax Credits, and benefits, had changed since before the pandemic. Answer options were “stayed the same,” “fallen by less than half,” “fallen by more than half,” and “increased.” An income loss indicator was set equal to 1 if income had fallen by any amount and to 0 otherwise. A severe income loss indicator was set equal to 1 if income had fallen by more than half and to 0 otherwise. To measure material hardship, we asked about resources during the prior month, using two questions that have been used in other studies during the pandemic:¹¹ whether they had been unable to pay the current month’s rent or mortgage, and whether the food they bought didn’t last and they didn’t have money to buy more.

ANALYTIC PLAN

First, we calculated descriptive statistics, overall and by respondents’ race and ethnicity. Our sample included people who identified as White, Black, Asian American/Pacific Islander, Native American, other, and multiple racial groups. We separately asked if people identified as Hispanic. Using this information, we created analytic groups of White non-Hispanic, Black non-Hispanic, and Hispanic of any race that were large enough to be analyzed separately. We also, in some contexts, provide analysis of all people identifying as anything other than White, non-Hispanic. Second, to examine the associations between layoffs, unemployment insurance and supplement receipt, and our outcome variables, we ran a set of linear probability models controlling for survey wave, with standard errors clustered at the person level. By controlling for survey wave, we held constant other policy supports available at particular points during the pandemic, such as the Child Tax Credit, which was only provided in the second half of 2021. We chose linear probability models for ease of interpretability.

LIMITATIONS

Although our study demonstrated the relationships among layoffs, unemployment insurance receipt, material hardship, and mental health among a sample of similarly situated workers, the sample population was limited and targeted. Hourly workers in other industries, salaried workers, and childless workers may have been affected differently. Further, our findings are local to a particular major city. Pandemic experiences may differ across communities, based on the course of disease prevalence and governmental and social response.

It is also possible, as with any survey data, that unemployment insurance receipt was misremembered or misreported, leading to recall bias.²⁴ However, our survey asked whether the respondent was currently receiving unemployment insurance and the supplement, not about retrospective receipt (see survey instruments in appendix 1).²¹ Payments were distributed weekly, so the respondent was only a maximum of six days past receiving one when we asked; the supplement, for these workers, represented much more money than the base payment and was highly salient because of its novelty, its receipt by large portions of the population, and heavy media coverage. Also, our survey was clearly fielded by a private university, not the government, which may have reduced respondents’ concerns about surveillance that can affect willingness to report income in many national surveys. Nonetheless, readers should interpret our results while being aware that our data were self-reported, not administrative records of unemployment insurance receipt.

Finally, we did not have access to the data necessary to provide clarity on specific structural and administrative barriers to unemployment insurance receipt among applicants, such as employer contestation, poor mail service, and discriminatory behavior among bureaucrats. These are pressing areas for future research.

Study Results

SAMPLE CHARACTERISTICS

Exhibit 1 presents sample characteristics at the time of recruitment in fall 2019. About half of respondents were Black and about one-fifth were Hispanic, similar to previous reports on central city hourly workers.²⁵ Our sample was majority female, which is consistent with working in the service industry and with having custody of a young child.²⁶ The average recipient was in their early thirties, consistent with being the parent of a young child,²⁶ and the modal education was twelve years, consistent with hourly service employment.²⁷ Just under 60 percent lived with another adult in the household. Most respondents worked one job, had been in that job for several years, and were earning around $12 an hour.

EXHIBIT 1.

Descriptive statistics of characteristics of low-wage service workers with young children in Philadelphia, Pennsylvania, at the time of study enrollment, overall and by race and ethnicity, August-November 2019

Characteristics	All respondents (N = 749)	BIPOC (n = 590)	White (n = 131)	Black (n = 370)	Hispanic (n = 149)
Demographics
Female, %	82.8	82.3	85.5	80.5	87.9
Mean age, years (SE)	31.0 (0.275)	30.4 (0.279)	33.0 (0.585)	30.3 (0.362)	29.4 (0.496)
Mean no. of children	2.2	2.3	2.1	2.3	2.1
Living with romantic partner, %	45.3	41.8	60.0	38.0	43.8
Living with another adult, %	58.5	55.1	74.8	50.8	60.4
Education, %
Less than high school diploma	8.8	9.2	7.0	6.6	16.8
High school diploma	63.1	65.0	54.3	69.4	57.7
More than high school diploma	28.1	25.9	38.8	24.0	25.5
Employment
Mean no. of jobs (SE)	1.2 (0.015)	1.1 (0.016)	1.2 (0.039)	1.1 (0.019)	1.2 (0.034)
Mean no. of months at primary job (SE)	38.5 (1.803)	35.0 (1.878)	50.5 (4.769)	34.9 (2.463)	30.0 (3.059)
Mean hourly wage at primary job, $ (SE)	12.44 (0.206)	12.33 (0.206)	12.81 (0.645)	12.48 (0.228)	12.08 (0.452)

source Authors’ analysis of baseline data from authors’ survey of service workers in the city of Philadelphia, collected at the time of study enrollment.

notes BIPOC is Black, Indigenous, and people of color. In our data, it refers to all respondents who identified as other than White non-Hispanic. We use it here for brevity of presentation.

These patterns held true across race and ethnicity, meaning that within our sample, respondents of different racial and ethnic identities were broadly similar on observable characteristics. In addition, respondents’ wage and job histories suggest that they were broadly eligible for the expanded unemployment insurance available during the pandemic, meaning that differences in access were likely due to systemic issues, including the widespread administrative failures and computer errors documented during the pandemic, rather than to respondent characteristics that might function as confounders in our analyses. Nonetheless, relationships shown are associational, and we interpret them with caution.

LAYOFFS AND RECEIPT OF UNEMPLOYMENT INSURANCE

Our respondents were employed in retail, food service, or hospitality jobs when initially recruited; thus, it is not surprising that nearly a quarter lost their jobs during the pandemic. Of greater note is that Black (26.6 percent) and Hispanic (25.1 percent) respondents were significantly more likely to be laid off than were White (19.8 percent) respondents (p < 0:01), despite having worked in the same set of industries before the pandemic (exhibit 2; also see appendix 2).²¹

EXHIBIT 2.

Pandemic era unemployment insurance (UI) application and receipt among low-wage service workers with young children in Philadelphia, Pennsylvania, by race and ethnicity, May 2020-November 2021

Respondent experiences	Black	Hispanic	White
Laid off	26.6%**	25.1%***	19.8%
Laid off, applied for UI	92.7	79.8***	93.3
Applied for and received UI	64.5**	65.0*	78.3
Received UI while supplements were available	73.3	79.1	81.5
Received UI during supplement window, received supplement	77.4	79.2	88.7
Laid off, received UI and supplement	33.9***	32.6***	52.8

source Authors’ analysis of panel data from authors’ surveys of service workers in the city of Philadelphia, collected in four waves.

notes “Supplement” refers to the extra Pandemic Emergency Unemployment Compensation UI benefits that were available (either $300 or $600 per week, depending on the period). Significance was calculated in comparison with White respondents.

^*p < 0.10

^**p < 0. 05

^***p < 0.01

Moreover, among our sample, not all workers who were laid off received unemployment insurance. Here we report findings for our entire sample, which includes the small number of respondents who reported a race or ethnicity other than Black, White, or Hispanic and thus are not included in exhibit 2. Across all respondents and all waves, nearly all (90 percent) of laid-off respondents filed for unemployment insurance (data not shown). However, of those who filed, only two-thirds reported having received unemployment insurance at the time of the survey. Of those who did receive unemployment insurance, 77 percent received it when supplemental unemployment insurance benefits were available; 80 percent of those who received benefits while supplements were available also received their supplemental payments. In total, only 60 percent of the sample of laid-off respondents received unemployment insurance, and only 37 percent received supplemental payments.

Furthermore, receipt of unemployment insurance was unequal across racial and ethnic groups (exhibit 2). The overwhelming majority of both Black (92.7 percent) and White (93.3 percent) laid-off respondents applied for unemployment insurance, but Hispanic laid-off respondents (79.8 percent) were significantly less likely to do so (p < 0:01). We note that these rates of unemployment insurance application were much higher than national rates for unemployed workers, likely because our sample consisted of mostly high-need, eligible workers. (In the overall population, in contrast, many of the eligible are low-need and many of the high-need are ineligible, both of which lead to lower application rates.)^17,18

Among those who applied, both Black (64.5 percent) and Hispanic (65.0 percent) respondents were significantly less likely to have received unemployment insurance benefits than were White (78.3 percent) respondents (p < 0:05 for Black versus White and p < 0:10 for Hispanic versus White). Black (73.3 percent) and Hispanic (79.1 percent) respondents were also less likely to have received unemployment insurance during the period when supplements were available than were White (81.5 percent) respondents, although the difference did not reach statistical significance. Similarly, conditional on availability, Black (77.4 percent) and Hispanic (79.2 percent) respondents were less likely than White (88.7 percent) respondents to actually receive their supplemental unemployment insurance payments, although again the difference did not reach statistical significance. Patterns for respondents of color overall were similar to those for Black and Hispanic respondents.

The attrition at each point, as illustrated in exhibit 2, led to stark racial and ethnic disparities in unemployment insurance receipt. Three-fifths of Black laid-off respondents and just over half of Hispanic laid-off respondents received unemployment insurance, compared with 73 percent of White laid-off respondents (p < 0:01 for White versus Hispanic and p < 0:05 for White versus Black; results not shown). Only one-third of laid-off Black and Hispanic respondents received both unemployment insurance and a supplement, compared with more than half of laid-off White respondents (p < 0:01 for both disparities) (exhibit 2).

As shown in appendix 3, regression model results demonstrate that differences in education, industry, work attachment, and wages do not explain these racial and ethnic differences in unemployment insurance receipt.²¹

ASSOCIATION WITH MATERIAL HARDSHIP AND MENTAL HEALTH

Exhibit 3 displays the relationship of layoffs to material hardship and mental health for our sample overall and by race and ethnicity. Patterns of results generally held across racial and ethnic groups, although smaller sample sizes for subgroups meant less precision. Across all respondents, layoffs were associated with negative outcomes, including income loss and severe income loss, being unable to pay rent or mortgage, anxiety, and depression. Running out of food was unaffected by layoffs, perhaps because of nutrition programs such as the Supplemental Nutrition Assistance Program and the Special Supplemental Nutrition Program for Women, Infants and Children. Harms from layoffs were generally larger for non-White than for White respondents, although differences were not statistically significant.

EXHIBIT 3.

Effects of layoffs on material hardship and mental health among low-wage service workers with young children in Philadelphia, Pennsylvania, overall and by race and ethnicity, May 2020-November 2021

Race and ethnicity	Income loss	Severe income loss	Unable to pay rent or mortgage	Ran out of food	Probable anxiety diagnosis	Probable depression diagnosis
All (N = 2,174)	0.209****	0.216****	0.087****	0.018	0.048**	0.047*
BIPOC (n = 1,709)	0.219****	0.233****	0.088***	−0.002	0.053*	0.058**
White (n = 389)	0.174***	0.137**	0.07*	0.052	0.048	−0.006
Black (n = 1,058)	0.232****	0.234****	0.046	0.009	0.025	0.027
Hispanic (n = 441)	0.159***	0.191****	0.179****	−0.065	0.071	0.116**

source Authors’ analysis of panel data from authors’ surveys of service workers in the city of Philadelphia, collected in four waves.

notes Each cell shows the regression coefficient from a separate regression. Complete regression results are in appendix 5 (see note 21 in text). Numbers reflect person-wave observations. BIPOC is Black, Indigenous, and people of color. In our data, it refers to all respondents who identified as other than White non-Hispanic. We use it here for brevity of presentation.

^*p < 0. 10

^**p < 0.05

^***p < 0. 01

^****p < 0.001

Exhibit 4 displays the relationships of unemployment insurance and supplement receipt with these outcomes among respondents who were laid off and applied for unemployment insurance. Among applicants, receipt was related to lower rates of material hardship and mental health difficulties. The associations between receipt and severe income loss, anxiety, and depression were statistically significant. Compared with those who did not receive unemployment insurance, those who received it were 14.5 percentage points less likely to experience severe income loss, 16.5 percentage points less likely to have probable clinically significant anxiety, and 9.9 percentage points less likely to have probable clinically significant depression. Importantly, these differences did not meaningfully change when we controlled for baseline anxiety or depression (see appendix 4).²¹ However, unemployment insurance receipt did not by itself significantly prevent income loss (it is designed to replace only a fraction of lost wages), inability to pay rent or mortgage, or running out of food in the sample overall.

EXHIBIT 4.

Effects of unemployment insurance (UI) and supplemental benefit receipt on material hardship and mental health among low-wage service workers with young children in Philadelphia, Pennsylvania, who were laid off and applied for UI, May 2020-November 2021

Race and ethnicity	Income loss	Severe income loss	Unable to pay rent or mortgage	Ran out of food	Probable anxiety diagnosis	Probable depression diagnosis
All (N = 534)
UI	−0.060	−0.145***	−0.084	−0.044	−0.165***	−0.099*
Supplement	−0.159****	−0.078**	−0.092***	−0.065*	0.039	−0.002
BIPOC (n = 433)
UI	−0.075	−0.156**	−0.100	−0.067	−0.134**	−0.156**
Supplement	−0.164****	−0.087**	−0.112***	−0.074*	0.027	0.040
White (n = 83)
UI	0.058	−0.061	0.081	0.186	−0.195	0.321**
Supplement	−0.190*	−0.014	−0.065	−0.104	0.037	−0.273**
Black (n = 290)
UI	−0.055	−0.112	−0.022	−0.004	−0.039	0.001
Supplement	−0.185***	−0.105**	−0.148***	−0.104*	−0.057	−0.064
Hispanic (n = 103)
UI	−0.175	−0.230*	−0.217*	−0.255**	−0.331**	−0.510****
Supplement	−0.123	−0.096	−0.077	−0.032	0.128	0.238**

source Authors’ analysis of panel from authors’ surveys of service workers in the city of Philadelphia, collected in four waves.

notes Each cell shows the regression coefficient from a separate regression. Complete regression results are in appendix 5 (see note 21 in text). Numbers reflect person-wave observations. “Supplement” refers to the extra Pandemic Emergency Unemployment Compensation UI benefits that were available (either $300 or $600 per week, depending on the period). BIPOC is Black, Indigenous, and people of color. In our data, it refers to all respondents who identified as other than White non-Hispanic. We use it here for brevity of presentation.

^*p < 0.10

^**p < 0.05

^***p < 0. 01

^****p < 0. 001

Unemployment insurance supplements were more effective than basic unemployment insurance at preventing material hardship, with each additional $300 reducing the chance of any income loss by 15.9 percentage points, inability to pay rent or mortgage by 9.2 percentage points, and running out of food by 6.5 percentage points. However, supplements generally did not enhance the effects of basic unemployment insurance in reducing anxiety and depression in our sample (exhibit 4; full regression results are in appendix 5).²¹

SPECIFICATION CHECKS

We conducted several specification checks to test the robustness of our findings (appendix 4).²¹ First, although our sample was employed at the time of recruitment in late 2019 and had, on average, multiple years of job tenure when recruited (exhibit 1), it is possible that some were no longer working in 2020 and were thus ineligible for unemployment insurance (expansions under Pandemic Unemployment Assistance made all parents of young children who had worked since January 1 effectively eligible for benefits). Only thirty-nine respondents were not working when contacted before the pandemic between February 20, 2020, when our reenrollment began, and March 13, 2020, when Pennsylvania enacted initial COVID-19 restrictions. Because job separations in retail, food service, and hotels more than double in January, after the winter holidays, it is likely that nearly all of those respondents were still employed January 1, 2020, making them eligible for Pandemic Unemployment Assistance under “recent labor force attachment.”²⁸ Nevertheless, we ran our regression models excluding participants who reported no longer being employed when surveyed between February 20 and March 13, 2020. The results were substantively similar to those reported in exhibits 3 and 4. We ran our regression models including participant fixed effects, which control for all time-invariant measured and unmeasured characteristics of participants; and for the models predicting anxiety and depression, we added baseline measures of mental health as an additional covariate. In both of these analyses, the results were substantively similar to those shown in exhibits 3 and 4.

Discussion

Our research, consistent with recent work, demonstrates that layoffs are harmful to material well-being and mental health and that unemployment insurance, both in general and through the supplemented benefits available during much of the COVID-19 pandemic, can help remediate these harms to service workers with children. Moreover, we show that neither the harms of layoffs nor the protections of unemployment insurance differ significantly across racial and ethnic groups. What does differ is the likelihood of receiving unemployment insurance and of receiving a generous pandemic supplement, with Black and Hispanic respondents much less likely to receive unemployment insurance than White respondents despite their identical locations and similar work histories. Our findings on racial disparities in unemployment insurance receipt are consistent with prior work during the pandemic.²⁹

There are several possible explanations for differential unemployment insurance receipt. First, we found that Hispanic workers were less likely to apply. Although our respondents were all eligible for unemployment insurance on the basis of their work history, and most likely paid into the unemployment insurance system through paycheck deductions, some may have been barred from access because of their documentation status (to protect respondents, we did not collect data on documentation status). In addition, a large body of work has shown that a perceived lack of access due to restrictions on immigrant eligibility for government support can lower applications, even among those who are eligible, from groups with a high immigrant share.^30,31 Fear that accessing unemployment insurance might harm future green card or citizenship applications—the Trump administration’s expansion of the public charge rule was much in the news just before the onset of the pandemic—may also have lowered applications among immigrants or those with immigrant relatives. In addition, low community levels of receipt may be self-reinforcing because of a lack of awareness of and knowledge about accessing the program.

The exact reasons why laid-off Black and Hispanic workers who applied for unemployment insurance were less likely to have received their unemployment insurance benefits are not known, but the patterns are consistent with prior research during the pandemic and the Great Recession.^6,32 Employers may keep poorer records for such workers or be more likely to contest their layoff status to the state for the employer’s benefit. Neighborhoods where Black and Hispanic workers live may experience worse mail service and mail security. These workers may have lower awareness of unemployment insurance, be less comfortable interacting with a state agency, or lack the resources to make it through a complicated bureaucratic process; they also may face discrimination from bureaucrats when they attempt to advocate for themselves.^33,34 Similar explanations may account for lower unemployment insurance receipt among Black respondents during the initial $600 supplement period if these obstacles led to delays in receipt.

Several policies would improve unemployment insurance for all and could close race and ethnicity gaps in its receipt. First, permanently extending coverage to lower-earning workers would reduce racial disparities in receipt driven by lower wages and less access to full-time work among Black and Hispanic workers. Second, permanently raising regular unemployment insurance benefit levels and duration, instead of requiring Congress to enact piecemeal supplements and extensions, would bring the US in line with peer countries and prevent material hardship, as the supplement did early in the pandemic. Third, reforming unemployment insurance financing so that employers are disincentivized from blocking workers’ claims would eliminate a barrier that has been shown to drive racial disparities in receipt.¹⁹ Finally, providing funding for states to modernize application and delivery processes would reduce the contribution of racial disparities in internet access to disparities in unemployment insurance receipt.

Conclusion

It is well documented that workers who are Black, Hispanic, and members of other non-White groups have experienced larger spikes in unemployment, income loss, material hardship, and mental health difficulty than have White workers during the pandemic.^7,9,29 In this article we found that unemployment itself has been a factor contributing to income loss, material hardship, and mental health problems and that these relationships hold true across racial and ethnic groups. However, the people not identifying as White in our sample were more likely to experience layoff despite their holding the same set of jobs as White sample members. Thus, they were more likely to experience the negative effects of job loss on their income and mental health, meaning that racial disparities in layoffs contributed to racial disparities in material hardship and mental health problems.

Consistent with previous work,^10,11,13,14 we found that unemployment insurance, and particularly its pandemic expansions, alleviated each of these negative sequelae of layoff. However, we found that people not identifying as White were less likely to receive unemployment insurance. Although much of the previous work has shown a similar pattern at the population level, we add to the growing body of evidence the fact that his disparity persists even among workers in the same jurisdiction and working in the same set of jobs, who should therefore be similarly eligible for the program. Precisely because unemployment insurance is beneficial, racial disparities in access to unemployment insurance contribute to income, housing, nutrition, and mental health disparities. But that also means that increasing equity in access to the program could help reduce the disparate effects of job loss on non-White workers and related racial health disparities.