Abstract

This article investigates the socio-demographic differences in household responses to the COVID-19 pandemic regarding employment and consumption. We find that the significant racial disparities in employment observed during the pandemic can be explained, in part, by differences in household income, composition, education, and occupational sorting. Nonetheless, we document pervasive racial, income, and educational gradients when focusing on household food insecurity and individuals' reliance on social insurance programs and other government assistance during the pandemic. Overall, our results highlight that the disparities observed for household income and education tend to be the most significant and most pervasive following the onset of the COVID-19 crisis.

George-Levi Gayle is a professor of economics at Washington University in St. Louis and a research fellow at the Federal Reserve Bank of St. Louis. Andrea Flores is an assistant professor of economics at EPGE Brazilian School of Economics and Finance.

INTRODUCTION

Since the onset of the COVID-19 pandemic, households have experienced significant disruptions that could potentially have long-lasting economic and social implications. In particular, it is well documented that the pandemic has had an unequal impact on employment with respect to income, race, and education and has contributed to a widening of pre-existing disparities. In this article, we adopt a more comprehensive approach by investigating how these employment gaps relate to socio-demographic gradients in the impact of the pandemic on household consumption, which, to the best of our knowledge, remains an open question. Specifically, we find significant differences across racial, income, and education groups regarding the incidence of food insecurity and the reliance on social insurance programs and other forms of government assistance during the pandemic.

To implement our approach, we rely on data from the first 36 weeks of the Census Bureau's Household Pulse Survey, which contains information on employment, housing, food sufficiency, spending patterns, and educational changes. We supplement this survey data with information on state-level COVID-19 cases and death counts and average mobility changes relative to February 2020 for each corresponding week captured in the Pulse retrieved from the COVID Tracking Project and Google's Mobility Reports, respectively. Furthermore, we use Current Population Survey (CPS) data to control for pre-pandemic trends when focusing on outcomes for which there is information available before the onset of the COVID-19 pandemic. Combining these data sources, we identify the differential impact of the pandemic across socio-demographic groups by implementing a regression-based methodology that allows us to control for differences in individual and household characteristics and state-specific differences related to containment of the virus. 

First, we focus on trends in employment and earnings losses experienced during the pandemic. We find significant disparities in earnings losses with respect to household income and respondents' education and race that persist after controlling for other household and respondents' characteristics. This finding is consistent with the socio-demographic differences in the decline in employment rates observed during the pandemic. Furthermore, we find that racial and education disparities in employment and earnings losses go away once we control for household income. The elimination of the racial gap in earning losses during the pandemic once we control for household income can be attributable to the strong relationship between household income and workers' ability to work from home documented in Mongey, Pilossoph and Weinberg (2020) and Blau, Koebe, and Meyerhofer (2020). To substantiate this argument, we corroborate that those individuals living in households at the lower quintiles of the income distribution and without a bachelor's degree were significantly less likely to experience a switch to telework during the pandemic, which persists after controlling for occupation. This result is in line with the decomposition results in Montenovo et al. (2020), indicating that a significant part of the socio-demographic gradients in employment losses is because of differences in pre-pandemic occupational sorting, which is strongly related to workers' ability to work from home. 

Second, we analyze the different types of non-employment related to the pandemic. We find that non-employment due to business responses to the pandemic and symptoms associated with COVID-19 was more prevalent among Blacks and Hispanics, individuals in households in the bottom quintiles of the income distribution, and individuals without a bachelor's degree. After controlling for other socio-demographic and household characteristics and respondents' occupations, these disparities persist. The patterns observed regarding non-employment due to business responses to the pandemic are consistent with the socio-demographic gradients reported in Adams-­Prassl et al. (2020) and Montenovo et al. (2020) related to job losses and layoffs experienced during the pandemic. Similarly, the patterns observed in non-employment due to symptoms associated with COVID-19 are consistent with the racial disparities documented by Angelucci et al. (2020) and Papageorge et al. (2020) regarding individual adoption of self-protective behavior and reflective of the uneven spread of the virus across socio-demographic groups. 

We also find similar income and education gradients in the impact of the pandemic on non-­employment as it relates to the childcare needs of the household. However, racial differences in household composition explain the disparity across households observed initially from childcare needs inducing non-employment. Furthermore, the pandemic has widened a pre-existing gender gap in non-employment because of the need to provide childcare for children in the household. Women were significantly more likely than men to report childcare needs as the reason for non-­employment during the pandemic, even after controlling for pre-pandemic trends. This finding corroborates that, besides differences in occupational sorting, the increased demand for home childcare has contributed to the adverse impact of the pandemic on women's employment rates documented in Alon et al. (2020) and Montenovo et al. (2020).

Finally, we analyze two measures of consumption disparity. The first pertains to the incidence of food insufficiency experienced by households during the pandemic. The second involves household sources of spending income to mitigate income losses during the pandemic. In terms of food insufficiency, the pandemic had a more adverse impact on Black and Hispanic households, respondents without a bachelor's degree, and households in the lower quintiles of the income distribution. We follow Moffitt and Ziliak (2020) and Ziliak (2020) in using information from the December Food Security Supplement of the CPS to corroborate that the adverse impact documented among these socio-demographic groups during the pandemic significantly worsened pre-pandemic disparities in food insecurity, mainly across income quintiles and education groups. 

We then investigate the different buffers used by households to face the income shock generated by the pandemic. We find that the socio-demographic groups with a higher share of households experiencing employment income losses during the pandemic relied significantly more on borrowing from family and friends and the economic impact payment (EIP) than the socio-demographic groups with a lower share of households experiencing employment income losses. When investigating how the EIP was spent, we find that households in socio-demographic groups more likely to receive these transfers used the money to cover food expenditures and utilities. We find similar gaps in the percentage of households saving this additional income. This share increases monotonically with household income and respondents' education and is higher for White respondents' households than for their non-White counterparts. 

Given that the reliance on unemployment insurance (UI) as a spending income source does not reflect the disparities documented in employment income losses, we analyze differences in the demand for and receipt of these benefits. We find that while the education, income, and race gradients of the respondents applying for UI benefits are consistent with the disparities documented for employment income losses, these disparities are reversed when we focus on the percentage of respondents who reported receiving UI benefits conditional on applying for them. This finding aligns with the criticism raised in Bitler, Hoynes, and Schanzenbach (2020) that social insurance programs in the United States have not effectively responded to the unmet needs of relatively more disadvantaged households during the pandemic. The authors argue that these programs' shortcomings are attributable primarily to delays, coverage gaps, and the magnitude of benefits—particularly related to UI. 

Read the full article.