How did the largest expansion of unemployment benefits in U.S. history affect household behavior? Using anonymized bank account data covering millions of households, we provide new empirical evidence on the spending and job search responses to benefit changes during the pandemic and compare those responses to the predictions of benchmark structural models. We find that spending responds more than predicted, while job search responds an order of magnitude less than predicted.
In sharp contrast to normal times when spending falls after job loss, we show that when expanded benefits are available, spending of the unemployed actually rises after job loss. Using quasi-experimental research designs, we estimate a large marginal propensity to consume out of benefits. Notably, spending responses are large even for households who have built up substantial liquidity through prior receipt of expanded benefits. These large responses contrast with a theoretical prediction that spending responses should shrink with liquidity.
Simple job search models predict a sharp decline in search in the wake of a substantial benefit expansion, followed by a sustained rebound when benefits expire. We instead find that the job-finding rate is quite stable. Moreover, we document that recall plays an important role in driving job-finding dynamics throughout the pandemic. A model extended to fit these key features of the data implies small job search distortions from expanded unemployment benefits.
Jointly, these spending and job finding facts suggest that benefit expansions during the pandemic were a more effective policy than predicted by standard structural models. Abstracting from general equilibrium effects, we find that overall spending was 2.0-2.6 percent higher and employment only 0.2-0.4 percent lower as a result of the benefit expansions.
Spending and Job Search Impacts of Expanded Unemployment Benefits: Evidence from Administrative Micro Data
This note updates the job-finding analysis in Ganong et al. (2021), estimating the disincentive effect of supplemental unemployment benefits between April 2020 and April 2021. We estimate the causal effect of the supplements using both a difference-in-difference research design and an interrupted time-series research design paired with administrative data. These empirical strategies can be used respectively to identify micro disincentive effects (the effect of increasing benefits for one worker) and macro disincentive effects (the effect of increasing benefits for all workers). Both designs imply a precisely estimated, non-zero disincentive effect. However, the disincentive effect of expanded benefits is quantitatively small: implied duration elasticities are substantially lower than pre-pandemic estimates and suggest that eliminating the supplements would have restored only a small fraction of overall employment losses. Extending the difference-in-difference design through April 2021 suggests that the disincentive effect of the supplements remains modest even after vaccines are broadly available. We conclude that unemployment supplements are not the key driver of the job-finding rate through April 2021 and that U.S. policy was therefore successful in insuring income losses from unemployment with minimal impacts on employment.