Do temporary extensions to unemployment insurance benefits matter? The effects of the US standby extended benefit program

Abstract

During the 2007–2010 economic downturn, the US temporarily increased the duration of Unemployment Insurance (UI) by 73 weeks, higher than any prior extension, raising concerns about UI's disincentive effects on job search. This article examines the effect of temporary benefit extensions using a Regression Discontinuity (RD) approach that addresses the endogeneity of benefit extensions and labour market conditions. Using data from the 1991 recession, the results indicate that the Stand-by Extended Benefit (SEB) program has a significant, although somewhat limited, impact on county unemployment rates and the duration of unemployment. The results suggest that the temporary nature of SEB benefit extensions may mitigate their effect on search behaviour.

Keywords:

• unemployment insurance
• unemployment
• regression discontinuity

JEL Classification::

• J65
• J68

Acknowledgements

I wish to thank Donald Parsons, Tara Sinclair, Roberto Samaniego, Stephanie Cellini and Anthony Yezer, as well as the other participants of the 10th Annual SOLE/IZA Transatlantic Meeting of Labor Economists for their helpful comments and discussions regarding this article.

Notes

¹ States trigger on 2 weeks after they first reach the trigger threshold (Wenger and Walters, 2006).

² In 1992, a Total Unemployment Rate (TUR) trigger greater than 6.5%, and exceeding the same rate in the previous 2 years by 10.0%, was added (Vroman and Woodbury, 2004).

³ In 2009, several workforce agencies said that they notified individuals on their last benefit check, when filing their last claim and when claimants respond to reporting requirements over the phone.

⁴ This test is conducted with five bins on either side of the border. There are only nine treated observations with a distance greater than 150 miles making estimates for these bins highly imprecise.

⁵ The AIC is given by where N is the sample size, σ is the mean squared error and p is the number of parameters. The AIC reflects the trade off between a more precise model and a more parsimonious one.

⁶ I consider only symmetric samples where the sample is always restricted to the lesser of the maximum distance of a county in the treated or the untreated group.

⁷ Epanechnikov kernel is as follows K = 0.75(1 − (D _i /h)²).

⁸ The log unemployment rate corrects for some skewness in the distribution of county unemployment rates.

⁹ The AIC also, selects a linear specification for both of these specifications.