Aggregate employment effects of unemployment benefits during deep downturns: Evidence from the expiration of the federal pandemic unemployment compensation
BibTeX
@techreport{dube2021aggregate,
title={Aggregate employment effects of unemployment benefits during deep downturns: Evidence from the expiration of the federal pandemic unemployment compensation},
author={Dube, Arindrajit},
year={2021},
institution={National Bureau of Economic Research}
}
Abstract
The expiration of the temporary $600 boost to weekly UI benefits under the Federal Pandemic Unemployment Compensation (FPUC) led to a sharp, unprecedented, 98 percentage point reduction (on average) in the replacement rate during a time when employment was recovering during the Covid recession. Leveraging the considerable variation in this drop across states, I use a difference-in-differences event study design to estimate the macro employment effects. I find little impact of job gains from the benefit reduction, especially when I focus on groups (non-college graduates, and those from non-high-income households) that comprise of most UI recipients. The estimates rule out job gains implied by much of the micro UI duration elasticities from the existing literature.
Notes and Excerpts
I use an event-study design using the new, high-frequency, Census Household Pulse Survey (HPS), and consider how non-college, adult worker employment evolved differentially by the state level variation in the median earnings replacement rates following the expiration of the FPUC.
Replacement rates varied across states. Used to predict employment effect of 600 dollar FPUC ending. Point estimate is negative (opposite to what you might expect) but statistically indistinguishable from zero.
In both panels, the DID estimates are not statistically distinguishable from zero. In my preferred specification with all controls, the magnitude of the DID estimate on employment probability (-0.011, s.e. = 0.015), along with the July sample employment rate of 0.60, implies a 1.8% reduction in jobs from the expiration of FPUC. Recall that labor supply consideration suggests employment should rise from the benefit cut, so the point estimate are of the “wrong” sign.
And the confidence interval rules out more than a few percentage points of employment increase.
He compares this to micro estimates of duration elasticity. That is, how does job finding react to the duration of benefits? Range of 0.1-0.9 is likely. But these elasticities would suggest a ΔE of 0.9M-8.2M
When we focus on those without a college degree—who compose of around 79% of the UI recipients, and 67% of labor force participants—the 95% confidence intervals rule out out elasticities larger than 0.19.
Going forward, future research would benefit from better understanding the source of this very modest employment response to UI benefit generosity, and whether it was driven by unusually low micro-level response of unemployment duration, or macro-level wedges affecting aggregate employment.7