One of the biggest challenges in designing social programs is to make sure you're not just rewarding people for doing what they were already planning to do.
A recent study (via Lee Crawfurd) suggests that this specific worry may be a bigger deal than previously thought. MIT superstar econometrician Esther Duflo, along with a group of researchers at France's Center for Research in Economics and Statistics and the Paris School of Economics, conducted a randomized trial of a program meant to help unemployed people in France find jobs. The Ministry of Labor was experimenting with having private contractors run job-placement programs previously administered by the state, and Duflo and her co-authors were brought on to evaluate one privatized program aimed at helping recent college graduates.
The program was extremely hands-on, with each jobless person assigned a caseworker to meet with him or her at least once a week. It lasted for a full year, the first half of which was devoted to finding the jobs, and the second half to making sure the new employees were adjusting to their work and finding them new positions if not. The study looked at about 50,000 people who were randomly sorted into a treatment group getting placement help or a control group not receiving aid. Almost 12,000 of those participants were jobless and thus in particular need of job placement services.
It worked — sort of. Only 43 percent of unemployed workers invited to participate signed up (the number was, understandably lower for workers who held jobs or at least claimed to). But unemployed workers who did participate were 11 percent more likely to get jobs with a fixed contract lasting six months or longer, and 4 percent more likely to get any kind of long-term job. On its own, that suggests that the program worked well at directing participants into work, or at least better than they would have done if the program hadn't existed.
But what about the control group? Aye, there's the rub. The authors also find that the program had significant negative effects on employment rates among non-participants, effects that were stronger in weak labor markets and, for whatever reason, among men. Altogether, their analysis suggests that out of every 1,000 participants, 36 found jobs through the program, but 48 non-participants were displaced. The net effect on employment is negative.
Things didn't always stay rosy for those participating in the program. Even ignoring displacement, the positive effect on employment for participating workers goes away within a year. In other words, they don't use the jobs gained through the program as a steppingstone to more permanent work.
So if job search programs don't work, what would? A better alternative, the authors argue, would be to beef up unemployment insurance (UI). They argue that their results imply that the "replacement rate" of UI — or the ratio of UI benefits to the recipients' previous wages — should be increased by 10 percentage points, and 15 percentage points during economic downturns. That keeps those who are looking for work still spending and gives them resources to search for work without benefiting one group of unemployed over the rest.