Unemployment Insurance Generosity and Aggregate Employment

Arindrajit Dube, Chris Boone, Lucas Goodman, and Ethan Kaplan:

Unemployment Insurance Generosity and Aggregate Employment: Abstract We estimate the impact of unemployment insurance (UI) extensions on aggregate employment during the Great Recession. Using a border discontinuity design, we compare employment dynamics in border counties of states with longer maximum UI benefit duration to contiguous counties in states with shorter durations between 2007 and 2014. ... Using either the baseline or the “refined” border design, we find no statistically significant impact of increasing unemployment insurance generosity on aggregate employment. Similar results obtain from instrumental variables estimates that only use variation in UI benefit duration induced by national-level policy changes—namely the 2008 benefit extension and the subsequent 2014 expiration of the Emergency Unemployment Compensation program. Our point estimates vary in sign, but are uniformly small in magnitude and most are estimated with sufficient precision to rule out substantial impacts of the policy. Based on the confidence intervals of all of our preferred specifications, we can reject negative impacts on the employment-to-population ratio exceeding 1.2 percentage points from an increase in maximum UI benefit duration from 26 to 99 weeks. Our more precise estimates rule out decreases in excess of 0.5 percentage points from the policy expansion. Overall, our macro employment estimates from the UI benefit extention can be rationalized with a small, negative labor supply effect from the UI benefit expansion—as suggested by a number of recent studies—along with a moderately sized fiscal multiplier. 

Also, our results differ substantially from Hagedorn, Karahan, Manovskii and Mitman (2015) and Hagedorn, Manovskii and Mitman (2016)  despite employing apparently similar strategies. In an online Appendix, we compare our results to those papers and discuss in detail what accounts for the substantial differences in our respective estimates, and decompose the differences due to choice of dataset and specifications.

Here is the link to the paper: http://bit.ly/2dq1IYA.

Here is a link to a media advisory with a synopsis of the study: http://bit.ly/2dLTrkX.