Unmet Social Needs And Worse Mental Health After Expiration Of COVID-19 Federal Pandemic Unemployment Compensation.

Abstract

Federal Pandemic Unemployment Compensation (FPUC) provided unemployment insurance beneficiaries an extra $600/week during the unprecedented COVID-19 related economic downturn, but initially expired in July 2020. We applied difference-in-difference models to nationally-representative data from the US Census Bureau’s Household Pulse Survey to examine changes in unmet health-related social needs and mental health among unemployment insurance beneficiaries before and after initial expiration of FPUC. Initial expiration was associated with significantly increased risk of missed housing payments (risk difference [RD] 10.8 percentage points, 95 confidence interval [95%CI] 8.0 to 13.6), food insufficiency (RD 3.8 percentage points, 95%CI 1.9 to 5.9), as well as depressive (RD 6.0 percentage points, 95%CI 3.1 to 9.0) and anxiety symptoms (RD 5.8 percentage points, 95%CI 2.9 to 8.8) among unemployment insurance beneficiaries. As further unemployment insurance reform is debated, policymakers should recognize the potential health impact of unemployment insurance.

The COVID-19 pandemic has had a massive economic impact in the US, leading to the largest 1-quarter economic contraction since record keeping began in 1945.¹ The unemployment rate peaked at 14.7% in April 2020 and remains elevated through December 2020.² Unemployment insurance (UI) has been a key part of the COVID-19 response. Over 60 million individuals have applied for UI benefits during the pandemic.³ A recent modeling study found that UI would likely play a key role in aiding the recovery of consumer spending and averting poverty.⁴ Prior work has also suggested that UI may offer important health benefits.^5–8 First, UI may help individuals meet health-related social needs, such as food⁹ and housing. Unmet food and housing needs are associated with worse health in a number of studies.^10–14 Further, by helping individuals meet basic needs, UI benefits may affect mental health¹⁵, such as reducing depressive and anxiety symptoms.^5,16

UI is administered as a federal-state partnership, with eligibility, benefit levels, and duration of benefit set by states, with broad oversight from the U.S. Department of Labor.¹⁷ States provide benefit funds (typically through employer/employee contributions as a form of social insurance), and the federal government provides funds for administration costs.^17–19 Pre-pandemic, commissions on unemployment insurance reform had raised concerns about relatively narrow eligibility and declining benefit levels.^18–21 The March 2020 Coronavirus Aid, Relief, and Economic Security (CARES) Act expanded eligibility for UI and the generosity of its benefits in three key ways.²² First, the Pandemic Emergency Unemployment Compensation extended the maximum benefit duration for those receiving state unemployment insurance by up to 13 weeks. Next, Pandemic Unemployment Assistance provided UI benefits to workers not eligible for state UI programs, for example ‘gig economy’ workers, the self-employed, and low wage workers. Finally, Federal Pandemic Unemployment Compensation (FPUC) added $600 per week in benefits on top of benefits received through state unemployment insurance or Pandemic Unemployment Assistance (adding to state unemployment benefits averaging around $350 per week).²³ The additional amount made benefits received during the period when FPUC was in effect much more generous than historical averages—and for lower wage workers typically represented more income than they had earned while working.²⁴

Pandemic Emergency Unemployment Compensation and Pandemic Unemployment Assistance have been continuously active since March 2020 (except for a brief lapse at the end of 2020). However, FPUC initially expired on July 31, 2020. After FPUC expired, the Lost Wages Assistance program (LWA) provided 6 weeks of $300 supplemental payments, and then it in turn expired in September 2020.^25,26 After FPUC expiration, UI beneficiaries received substantially lower weekly payments. For example, a recipient may have gone from a mean state benefit of $350 + federal benefit of $600 in July 2020 to a mean state benefit of $350 + federal benefit of $300 (a 32% reduction) in September 2020, followed by a further decrease only to $350 in state provided funds after the LWA program expired (a 63% reduction from July 2020 levels). Thus, owing to FPUC expiration, UI beneficiaries received substantially smaller benefits beginning in August 2020, compared with the period when FPUC was in effect. FPUC was re-activated in January 2021 through March 2021 as part of H.R.133 - Consolidated Appropriations Act, 2021²⁷, but with a lower supplement of $300/week (the same supplement as the LWA program).

For this study, we used the end of FPUC in July 2020 as a ‘natural experiment’ that demarcated those who received more generous benefits from those who received less generous benefits. Using difference-in-difference analyses, we tested the hypothesis that, for individuals with ongoing COVID-related income disruption, the expiration of FPUC UI benefits would be associated with more unmet health-related social needs and worse mental health.

Methods

Data source, study setting, and participants

In this repeated cross-section study, we used data from the US Census Bureau’s Household Pulse Survey Public Use Files (https://www.census.gov/householdpulsedata).²⁸ It is a brief, internet-based survey, offered in English and Spanish, designed to enable population estimates of the household experience during COVID-19 across the US.²⁸ So far, it has been fielded in two phases. Phase 1 was fielded weekly over 12 weeks (April 23, 2020 to July 21, 2020). During the final 6 weeks (June 11 to July 21, 2020), a question about unemployment insurance benefits was added. After a pause, Phase 2 began August 19, 2020, fielded in two-week blocks. Though Phase 2 included additional questions, many of the same questions asked during Phase 1 were retained, to permit analysis of trends. A respondent could complete the survey up to three times within a phase, but no respondents participated in both Phase 1 and Phase 2 during the period we analyzed. Only one respondent per household completed the survey.

We used data from the final six weeks of Phase 1 (June 11 to July 21, 2020, during which FPUC was active) and the first eight weeks of Phase 2 (August 19 to October 12, 2020, during which FPUC had expired). We included working age adults (those born from 1955 to 2002) with ongoing pandemic-related income disruption, defined as people who reported a loss of employment income in their household on or after March 13, 2020, and, during the 7 days preceding the survey, not having the kind of earned income they had pre-pandemic to meet household spending needs. We used these criteria to identify those with ongoing household income disruption, as those who had lost jobs initially but then returned to work (and thus re-established a source of earned income) would not need UI. More information about the Household Pulse Survey, including both the Phase 1 and Phase 2 survey instruments, is publically available (https://www.census.gov/programs-surveys/household-pulse-survey/technical-documentation.html). This study was considered non-human subjects research, and thus exempted from the need for IRB approval, by the UNC IRB.

Unemployment insurance benefits

We categorized as receiving UI those who reported using UI benefits to meet household spending needs in the last 7 days, while those who did not report using UI were categorized as not receiving UI. The method of classification was the same across the FPUC and post-FPUC periods.

Outcomes

We considered several outcomes relevant to the pathways between UI and short-term health impacts.

We considered two health-related social needs outcomes: housing instability (whether the respondent had made the prior month’s housing payment on time), and food insufficiency (sometimes or often not having enough to eat).²⁹ The food insufficiency question was derived from NHANES III, and was scored according to standard practice.^29,30 We also examined two mental health outcomes: depressive and anxiety symptoms. Respondents were asked the Patient Health Questionnaire (PHQ) 2 for depressive symptoms and Generalized Anxiety Disorder (GAD) 2 questions for anxiety symptoms.^31,32 Scores ranged from 0 to 6 (more depressive or anxiety symptoms), and, in keeping with scoring recommendations, we used a cutpoint of ≥3 on both the PHQ2 and GAD2 to indicate potentially clinically meaningful symptoms.^31,32

Covariates

We considered several covariates that may confound the association between receipt of UI benefits and health outcomes. These were: age, gender (male or female), self-reported race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, non-Hispanic Asian, and non-Hispanic other or multiracial), education (< high school diploma, high school diploma, > high school diploma), 2019 (i.e., pre-pandemic) annual household income category (less than $25,000, $25,000 – $34,999, $35,000 – $49,999, $50,000 – $74,999, $75,000 – $99,999, $100,000 – $149,999, $150,000 – $199,999, and $200,000 and above), marital status (married versus not), pre-pandemic food insufficiency, work in the last seven days, and household size. Because the impact of the pandemic was heterogeneous across states and time, we included variables for state of residence (all 50 states plus the District of Columbia were included, which we refer to as states for convenience), state level COVID cases per capita at the beginning of the survey week³³, and the calendar date of survey administration.

Statistical analysis

The Household Pulse Survey contains person weights to produce nationally representative estimates, which we used for all analyses. Our primary research question was whether there was a change in study outcomes in the post-FPUC period for UI beneficiaries. To examine this question, we used difference-in-difference analyses that compared those who did and did not receive UI benefits in the FPUC and post-FPUC period. To conduct these analyses, we fit regression models with indicators for receipt of UI benefits (yes or no), time period (post-FPUC or during FPUC), and their product term. To avoid interpretability issues with nonlinear models, we fit linear probability models.^34,35 The unit of analysis was the survey response for a given week, and participants could complete the survey in more than one week. For descriptive statistics and unadjusted analyses, we used a respondent’s first survey response. For regression analyses, we included all survey responses, and used robust variance estimation with standard errors clustered by respondent to account for repeated measures within respondents. Regression models included all of the above listed covariates for adjustment.

To provide difference-in-difference estimates of study outcomes comparing the FPUC period in our dataset (June to July 2020) to the LWA period (August to September 2020, which was when the federal supplement was the same as 2021 FPUC benefit levels), we conducted similar analyses using data from the FPUC period and a subset of the post-FPUC period (August 19 to September 14, 2020). We do however note that owing to state variability in distribution of LWA benefits, survey respondents may not have received LWA benefits they were eligible for at the time they responded to the survey.

We conducted two sets of sensitivity analyses. The first used an alternative variance estimation strategy, Balanced Repeated Replication (BRR) weights, in order to make sure that our results were not sensitive to the choice of variance estimation strategy (Technical Appendix).³⁶ Therefore we fit the same models as for our primary analyses, but using BRR variance estimation. Second, though missingness for variables was generally low (< 5%), the missingness for the income variable was 13.2%. Therefore, as a sensitivity analysis, we used multiple imputation by chained equations,³⁷ generating 10 imputed data sets to check that our results were not sensitive to item non-response.

A key assumption of difference-in-difference analysis is that of ‘parallel trends’, which means that the difference in outcomes, if any, between those who did and did not receive UI should remain stable during the FPUC period (when there were no major changes in federal UI policy). To test this assumption, we used 3 approaches (see Technical Appendix³⁶ for details): plots of unadjusted means for each of the four study outcomes, an ‘event-study’ type analysis that fits regression models with a week-by-UI indicator product term, and ‘placebo tests’ for each of the four study outcomes. For the placebo tests, we took all of the Phase 1 data (which corresponds to the FPUC period) and artificially created a divide between the data from the first three weeks of our Phase 1 data and the second three weeks. We then fit the same difference-in-difference models as used for the main analyses, using the artificial divide to demarcate the two periods. In other words, we tested whether we would detect a change during a period when we should not. Using this approach, if the parallel trends assumption holds, the UI-by-period product term should have a coefficient close to 0.

Analyses were conducted in SAS version 9.4, Stata/MP version 16.1, and R version 3.5.3. Given multiple outcomes in this study, we used the false discovery rate approach to control for type 1 error.³⁸ Therefore, we present regression results with both a nominal p-value and a ‘q-value’, which can be interpreted as indicating the proportion of results with that q-value or lower that would be expected to be a false positive accounting for all the analyses conducted.³⁹ Thus a q-value < 0.05 indicates that, accounting for multiple analyses, a given result is expected to be a false positive less than 5% of the time. We interpreted a q-value < 0.05 to indicate statistical significance.

Limitations

As with any survey study, there is the possibility of selection bias owing to non-response (i.e., those who completed the survey are not representative of the underlying population). To permit rapid fielding with minimal staffing, the Household Pulse Survey used a recruitment strategy that invited a large number of participants with minimal follow-up, and fielded each survey for only a brief period of time. This resulted in a response rate much lower than typical for US Census surveys (approximately 3%).⁴⁰ This low response rate was anticipated in the survey design, and representativeness weights were provided to help address this issue. Nevertheless, we recognize that selection bias owing to non-response is an important consideration. Next, the data are self-reported, and limited to what was asked in both phases of the Household Pulse survey: we did not have detailed data on current income, pre-pandemic jobs, or benefits received. Therefore, though we used a strong study design and adjusted for a robust set of potential confounders, the possibility of unmeasured confounding, particularly time-varying confounding, is an important concern. Next, though we conducted several robustness checks, the assumption inherent to difference-in-differences analyses, that the intervention and control group would continue to have parallel trends in outcomes into the future (were it not for the intervention), is fundamentally untestable. Finally, state reporting of unemployment numbers has been unreliable during the pandemic⁴¹, and with available data, we are unable to determine if those who did not report receiving UI benefits did not receive them due to ineligibility, misreporting, or issues with accessing benefits (see Technical Appendix for further discussion³⁶). Combined with only having self-report data regarding UI benefit receipt, these factors could produce misclassification that would tend to bias results to null.

Results

Descriptive Statistics

For this repeated cross-sectional study, there were 122,133 unique individuals who met inclusion criteria, representing 38 million Americans, and they provided 132,254 survey responses (range: 1 – 3 per participant). Exhibit 1 presents characteristics of participants based on their first recorded survey response (see Appendix Exhibit 1 for more detail). There were 49,700 respondents, representing 14 million individuals, who reported household use of UI benefits in the past week. There were 72,433 respondents, representing 24 million individuals, who did not report household use of UI, despite pandemic related income disruption. Overall, those who did not receive UI were more likely to be Hispanic, have lower education, and have lower pre-pandemic income than those who did receive UI. Study outcomes were patterned by race/ethnicity, education, and pre-pandemic income (Appendix Exhibit 2).

Exhibit 1.

Participant Characteristics Household Pulse Survey (June to October, 2020)

	Overall (N = 122,133)	Did Not Receive Unemployment Insurance Benefits (N = 72,433)	Received Unemployment Insurance Benefits (N = 49,700)
	Weighted % or mean (SD)	Weighted % or mean (SD)	Weighted % or mean (SD)
Age, years** [mean (SD)]	40.7 (13.3)	40.5 (13.4)	41.0 (13.0)
Women****	51.6	50.5	53.5
Race/ethnicity****
NH White	45.3	43.5	48.2
NH Black	17.2	17.0	17.5
Hispanic	27.3	29.8	22.9
NH Asian	5.6	4.8	7.0
NH Other	4.7	4.9	4.4
Education****
< HS Diploma	14.6	17.4	9.9
HS Diploma	35.6	35.7	35.4
> HS Diploma	49.8	46.9	54.7
Pre-pandemic annual household income****
Less than $25,000	28.1	31.9	22.0
$25,000 – $34,999	16.3	16.5	16.0
$35,000 – $49,999	15.0	14.3	16.0
$50,000 – $74,999	17.3	15.7	19.7
$75,000 – $99,999	10.0	9.0	11.6
$100,000 – $149,999	8.3	7.5	9.6
$150,000 – $199,999	3.0	2.8	3.2
$200,000 and above	2.1	2.3	1.8
Married	43.7	43.8	43.4
Pre-Pandemic Food Insufficiency****	21.1	24.4	15.5
Worked in Past 7 Days****	35.7	43.0	23.5
State COVID cases per capita [mean (SD)]	0.01 (0.01)	0.01 (0.01)	0.01 (0.01)

Notes:

Weighted N = 37,717,054 overall, 23,763,053 who did not receive unemployment insurance, and 13,954,001 who did receive unemployment insurance

P values from weighted t-tests (age, state COVID cases per capita) or chi-squared tests (all other variables)

^*p < 0.1,

^**p < 0.05,

^***p < 0.01,

^****p < 0.001

NH = non-Hispanic; HS = high school; FPUC = Federal Pandemic Unemployment Compensation

Unadjusted Analyses

In unadjusted analyses, 26.5% of respondents reported current food insufficiency, compared with 21.1% reporting pre-pandemic food insufficiency—a significant increase (McNemar’s test p-value < 0.001). Further, 42.0% had a PHQ2 score ≥3 (Exhibit 2; Appendix Exhibit 3). Health related social needs and mental health outcomes were worse for those who did not receive UI benefits—for example, 30.3% of those who did not receive UI benefits reported missing a housing payment, compared with 23.0% of those who received UI (p<0.001).

Exhibit 2.

Unadjusted Relationship between Receipt of Unemployment Insurance Benefits and Health-Related Social Needs and Mental Health Outcomes

	Overall (N = 122,133)	Did Not Receive Unemployment Insurance Benefits (N = 72,433)	Received Unemployment Insurance Benefits (N = 49,700)
	Weighted %	Weighted %	Weighted %
Missed Housing Payment****	27.4	30.3	23.0
Food Insufficiency****	26.5	29.6	21.5
PHQ2 Depression Score ≥ 3**	42.0	42.6	40.9
GAD2 Anxiety Score ≥ 3	50.2	50.4	49.8

Notes:

Weighted N = 37,717,054 overall, 23,763,053 who did not receive unemployment insurance, and 13,954,001 who did receive unemployment insurance

P-value from weighted chi-square tests

^*p < 0.1,

^**p < 0.05,

^***p < 0.01,

^****p < 0.001

COVID = Coronavirus Disease; PHQ = Patient Health Questionnaire; GAD = Generalized Anxiety Disorder

Testing the Parallel Trends Assumption

Demographics and outcomes in the FPUC period are shown in Appendix Exhibits 4–5. Trends in study outcome by week did not lead us to reject the parallel trends assumption (Appendix Exhibits 6–9). Event study findings were consistent with the parallel trends assumption holding during the FPUC period, with coefficients typically close to 0 and not statistically significant (Appendix Exhibit 10). Similarly, for all outcomes, placebo tests found coefficients near 0 and were not statistically significant (Appendix Exhibit 11).

Adjusted Analyses

In difference-in-difference analyses adjusted for age, gender, race/ethnicity, education, income, household size, marital status, prior food insufficiency, work status in the past 7 days, state COVID cases per capita, and state and week of survey fixed effects, we consistently found that receiving UI benefits was associated with lower risk for unmet health-related social needs, and depressive and anxiety symptoms (Exhibit 3; Appendix Exhibits 12–16). For example, the adjusted risk difference (RD) for food insufficiency in those who received UI benefits compared with those who did not was 5.01 percentage points lower (95%CI 6.51 lower to 3.51 lower, p <0.0001, q = <0.0001).

Exhibit 3.

Difference-in-Difference Estimates Comparing Receipt of Post-FPUC Unemployment Insurance to Receipt of Unemployment Insurance when FPUC was Active

	Difference in Difference Estimate in Percentage Points of Outcome Risk	P	Q	Unemployment Insurance Estimate in Percentage Points of Outcome Risk (95% CI)	P	Q
Missed Housing Payment	10.79	<.0001	<.0001	−10.70	<.0001	<.0001
Food Insufficiency	3.88	0.0002	0.0003	−5.01	<.0001	<.0001
Depressive Symptoms	6.04	<.0001	0.0002	−4.15	0.0003	0.0005
Anxiety Symptoms	5.82	<.0001	0.0002	−3.15	0.006	0.007

Notes:

The units for the difference-in-difference estimate and the unemployment insurance estimates are percentage points of risk for the outcome indicated by the row. The unemployment insurance estimate compares those who did to those who did not receive unemployment insurance.

Point estimates, 95% confidence intervals, and p-values are from linear probability regression models with robust standard errors clustered by respondent (to account for repeated survey responses within individuals) and representativeness weights.

The q-value comes from the False Discovery Rate approach to control type I error. The q-value can be interpreted as indicating the proportion of results with that q-value or lower that would be expected to be a false positive accounting for all the analyses conducted. Thus a q-value < 0.05 indicates that, accounting for multiple analyses, a given result is expected to be a false positive less than 5% of the time.

Models were adjusted for age, gender, race/ethnicity, education, income, household size, pre-pandemic food insufficiency, marital status, work in the last 7 days, state COVID cases per capita, state, and week of survey.

FPUC = Federal Pandemic Unemployment Compensation

When comparing post-FPUC to the period when FPUC was active, we observed significantly higher risk of unmet health-related social needs, and depressive and anxiety symptoms among UI recipients during the post-FPUC period. The difference-in-difference estimate of missing a housing payment was 10.79 percentage points greater (95% CI 7.99 to 13.58, p < 0.0001, q < 0.0001) for those receiving UI in the post-FPUC period, relative to those receiving UI during the FPUC period. Similarly, the difference-in-difference estimate of food insufficiency was 3.88 percentage points greater (95% CI 1.87 to 5.89, p = 0.0002, q = 0.0003), of depressive symptoms was 6.04 percentage points greater (95% CI 3.10 to 8.97, p < 0.0001, q = 0.0002), and of anxiety symptoms was 5.82 percentage points greater (95% CI 2.90 to 8.75, p < 0.0001, q = 0.0002).

Results were similar when comparing the period when FPUC was active to data restricted to August 19^th – September 14^th (when the LWA program was active) (Exhibit 4; Appendix Exhibit 17).

Exhibit 4.

Difference-in-Difference Estimates Comparing Receipt of Unemployment Insurance when FPUC was Active to Receipt of Unemployment Insurance from August 19 to September 14, 2020 (when LWA was active)

	Difference in Difference Estimate in Percentage Points of Outcome Risk (95% CI)	P	Q	Unemployment Insurance Estimate in Percentage Points of Outcome Risk (95% CI)	P	Q
Missed Housing Payment	11.09	<.0001	<.0001	−10.44	<.0001	<.0001
Food Insufficiency	4.77	<.0001	0.0002	−4.94	<.0001	<.0001
Depressive Symptoms	5.94	0.0006	0.001	−4.11	0.0004	0.0007
Anxiety Symptoms	7.03	<.0001	0.0001	−2.85	0.01	0.01

Notes:

The units for the difference-in-difference estimate and the unemployment insurance estimates are percentage points of risk for the outcome indicated by the row. The unemployment insurance estimate compares those who did to those who did not receive unemployment insurance.

Point estimates, 95% confidence intervals, and p-values are from linear probability regression models with robust standard errors clustered by respondent (to account for repeated survey responses within individuals) and representativeness weights.

The q-value comes from the False Discovery Rate approach to control type I error. The q-value can be interpreted as indicating the proportion of results with that q-value or lower that would be expected to be a false positive accounting for all the analyses conducted. Thus a q-value < 0.05 indicates that, accounting for multiple analyses, a given result is expected to be a false positive less than 5% of the time.

Models were adjusted for age, gender, race/ethnicity, education, income, household size, pre-pandemic food insufficiency, marital status, work in the last 7 days, state COVID cases per capita, state, and week of survey.

FPUC = Federal Pandemic Unemployment Compensation

LWA = Lost Wages Assistance Program

Sensitivity Analyses

BRR analyses had point estimates identical to the main analyses, with smaller confidence intervals, as expected, meaning that post-FPUC UI was more strongly associated with unmet health-related social needs and worse mental health in these analyses (Appendix Exhibit 18). Multiple imputation analyses also had estimates similar in magnitude to the main analyses, and post-FPUC UI was significantly associated with unmet health-related social needs and worse mental health across all outcomes.

Discussion

When examining nationally-representative survey data among those with COVID-19 pandemic related income disruption, we found that being in a household that received unemployment insurance benefits was associated with fewer health-related social needs and better mental health. However, the lower benefit levels received by UI beneficiaries after the expiration of FPUC were associated with greater risk for unmet health-related social needs and worse mental health.

These associations are consistent with UI having its intended effect—providing resources to help mitigate the economic impacts of the pandemic—but also with the concern that UI is less effective without FPUC. It is also important to consider the high prevalence of the study outcomes overall. Among adults with pandemic related income disruption, one in four report missed housing payments and food insufficiency, two in five report clinically meaningful depressive symptoms, and more than half report anxiety symptoms above a clinically meaningful threshold.

This study extends prior literature on potential health benefits of UI. Pre-pandemic work found that UI benefits improved mental health, particularly depression.⁶ Work conducted in the context of the ‘Great Recession’ also found that more generous UI was associated with better mental health, and may have prevented deterioration of self-rated health.^5,42 Other work examining austerity-related cuts to social programs found that such cuts were associated with worsening depression⁴³, and worse access to healthcare.⁴⁴

This study suggests several directions for future research. First, studies should examine how state-level variability in UI benefits is associated with health-related social needs and mental health outcomes. Next, given the barriers that likely prevented some eligible individuals from receiving UI, a study that sought to estimate an intention-to-treat average treatment effect (e.g., the effect of specific UI policies across all who might be eligible for them, as opposed to an effect estimated among those who received UI) would complement the information provided here.

Given the ongoing debate surrounding UI in the US, the study findings have important implications. Pandemic UI programs incorporated several features of UI reform that had been recommended, but not enacted, pre-pandemic.^18,20,21 These include greater income replacement and more inclusive eligibility, particularly for low-income and self-employed workers. An additional proposed reform is to simplify, modernize, and possibly, nationalize (i.e., assume responsibility at the federal level) the unemployment insurance program. The massive spike in UI claims at the beginning of the pandemic overwhelmed legacy systems, leading to large backlogs and frustration for users.⁴⁵ Finally, the comparison between the FPUC period and then period when LWA was active is revealing. The re-activation of FPUC through at least March 2021 as part of H.R. 133 is welcome news. However, the study findings suggest that given the lower supplement level ($300 versus $600), re-activated FPUC may be less beneficial than initial FPUC. Conversely, owing to variation in LWA implementation among states, some people eligible for LWA may not have been receiving LWA benefits when they completed the survey.

In conclusion, we found enormous economic disruption wrought by the pandemic. Unemployment insurance benefits may help mitigate this, but the initial expiration of FPUC was associated with increased risk for unmet health-related social needs and worse mental health among unemployment insurance beneficiaries. In future debates about both short-term and longer-term unemployment insurance reform, it will be important to remember that unemployment insurance is a vital form of social insurance that could provide meaningful health benefits.