Wednesday, April 22, 2020

REA experimental evaluation released

The final report for the recent experimental evaluation of the U.S. REA program has been released. I was involved on the margins of this as a consultant to Abt Associates, who conducted the evaluation under contract to the U.S. Department of Labor. Jacob Klerman, the lead author, overlapped with me at Chicago, and was my TA for Gary Becker's first year micro class.

The report is notable on a few dimensions: First, relative to other government reports on experimental evaluations I have been involved with, it does an exceptional job of embedding the analysis and results into the broader literature, and an exceptional job of incorporating the underlying economics of the problem. Second: the REA program is the programmatic sibling of the WPRS (= Worker Profiling and Reemployment Services) program considered in Black, Smith, Berger and Noel (2003). In line with our findings there, requiring people to show up for a meeting with a staff person seems to drive a big part of the impact in REA. Also in line with BSBN (2003), but more compellingly due to the much larger sample, the impacts do not clearly vary with the claimant's estimated probability of exhausting their benefits.

Here is Abt's summary:
The Challenge
The  goals of the Department of Labor’s Reemployment and Eligibility Assessment (REA) program were to address the reemployment needs of Unemployment Insurance (UI) claimants and prevent and detect improper payments. Abt Associates conducted a study of 250,000 UI claimants to determine the impact of the low-intensity program, which involved a few hours of group engagement and a few hours of one-on-one counseling. The cost generally was less than $100 per claimant.
The Approach
This REA evaluation is among the largest random assignment studies of a social program ever conducted in the United States. Nearly every REA-eligible UI claimant in the four participating states during the study intake period was randomized. Despite this highly inclusive design, the resulting sample sizes are only borderline sufficient to address some of the research questions. This appears to be because—consistent with the low intensity of the intervention—impacts are small. Abt previously conducted an implementation evaluation.
The Results
The program cut the time people spend drawing UI by an average of 1.3 weeks and increased earnings by $465 in the first year, or two percent of wages in the control group. About half of the decline in claimants represented an increase in work. The other half did not receive UI, but were not working. State responses to failure to attend required REA meetings appear to be an important factor in the drop in UI weeks. The REA program has been replaced by the Reemployment Services and Eligibility Assessment Program (RESEA). Our study will inform the redesign of RESEA program and the evalulation design.
You can find the implementation report and the final report on Abt's web page here.

No comments: