Social Insurance with Imperfect Eligibility Screening: Theory and Evidence from Pandemic UI

This paper studies social insurance with imperfect eligibility screening, focusing on unemployment insurance (UI) during its expansion in 2020 and 2021. We study the extent of imperfect screening by identifying anomalous payments using administrative tax data and UI policies, finding $214 billion in potentially-improper payments—concentrated in the Pandemic Unemployment Assistance (PUA) program—with approximately half detectable ex-ante through improved federal-state data sharing. There is substantial geographic variation, and a border design shows this is partly due to policy decisions made by states. To assess the implications for optimal policy, we first conduct simulations that replace PUA with means-tested, lump-sum transfers, finding that these transfers would have better insured against income losses at lower administrative cost. Second, we develop a model of opt-in versus automatic transfers that shows the targeting advantage of opt-in programs can reverse when ineligible recipients pass the benefit screen. Calibrated to 2020 UI, the model implies that shifting toward automatic transfers would have increased social welfare.