Monday, April 25, 2016

What Do We Know about Subsidized Employment Programs?

A subsidized employment program is when the government offers a subsidy to an employer (could be private sector or public sector) to hire those from a certain eligible group. The eligible group can be defined in many ways: for example, those who live in a certain set of neighborhoods, or those who have been unemployed for a certain time, or single mothers, or the disabled, or older workers, or those just emerging from jail or prison, or those who are already in some income-support program and trying to make a transition to work, or other groups. The basic idea of subsidized employment is that it's better to pay people to work than it is to support them while they aren't working--and further, there is a hope that subsidized work experience can be a transition to being hired by an employer who doesn't need a subsidy.

 Indivar Dutta-Gupta, Kali Grant, Matthew Eckel, and Peter Edelman provide a useful overview of the existing research in "Lessons Learned from 40 Years of Subsidized Employment Programs," published by the Georgetown Center on Poverty and Inequality in Spring 2016. The report calls subsidized employment a "promising strategy," which seems a fair judgement if one mentally adds "but not yet proven." There are two major federal government studies now underway involving subsidized employment. The "Subsidized and Transitional Employment Demonstration (STED), 2010-2017" being run by the US Department of Health and Human Services in seven cities and the Enhanced Transitional Jobs Demonstration study is being run in seven cities by the US Department of Labor. Results from these studies should become available over the next couple of years. 

It's easy to hypothesize about why subsidized employment programs might work, or not, but as we wait for these big new studies to be completed, what does the preexisting evidence say? As one might suspect, the reason that the government is doing the additional studies is that the existing evidence isn't as clear as one might like. 

For example, one of the biggest subsidized employment programs tried in the US was the Comprehensive Employment Training Act which ran from 1973-1982. It offered a combination of public service jobs, classroom training, subsidized on the job training and work experience. As of 1980, for example, there were more than 700,000 people participating in CETA. But as the report writes, CETA "was not rigorously evaluated." Modern studies are often set up with a big group of eligible participants who are then randomly assigned either to get the subsidized employment or not. Based on this study design, it's relatively straightforward to look at the difference in outcomes between two very similar groups--some of who are randomly in the program and some who are not. 

But CETA wasn't based on randomization. People decided whether to enroll, and presumably those who had more initiative or responsibility or better skills or fewer life problems were more likely to enroll. Economists tried to use statistical tools to sort out these factors, but while some factors are measureable (say, years of schooling or prior work experience), lots of factors like initiative or sense of responsibility aren't collected in data. So figuring out how much difference CETA made, as opposed to these other characteristics, was very hard. As the report says: "Little can be said with certainty about CETA ...  but non-experimental studies suggest sometimes contradictory findings, with one analysis suggesting positive effects only for women in classroom training, OJT [on-the-job training], and public service employment (not work experience), and another analysis of the impacts of training on men found large positive effects from classroom training and smaller, positive effects from OJT." It's also fair to point out that the labor market for lower-skilled labor has evolved considerably during the four decades or since CETA, so even if the evidence was a lot more clear-cut than it is, it seems hazardous to draw lessons for getting people jobs in 2016 based on evidence from the 1970s and early 1980s. 

Here's a table from the report summarizing results from "rigorously evaluated models," by which they mean models that involved in some way a randomized element in who was assigned to receive the subsidized employment. (You can click on the table to enlarge it.  In the final column, the results that appear in bold type are statistically significant.)

How you evaluate this kind of table may depend on what sort of mood you're in. Some of the studies are relatively old, like the 1970s and 1980s. Some show no effect, or no statistically significant effect. But some data is more recent, and some results are more positive. As the report summarizes: 
  • Subsidized employment programs have successfully raised earnings and employment. This effect is not universal across programs or target populations, but numerous rigorously evaluated interventions offer clear evidence that subsidized employment programs can achieve positivelabor market outcomes. Some of these effects derive from the compensation and employment provided by the subsidized job itself, but there also is evidence that well-designed programs can improve outcomes in the competitive labor market after a subsidized job has ended.
  • Subsidized employment programs have benefits beyond the labor market. Fundamentally, subsidized jobs and paid work experience programs provide a source of both income and work experience. A number of experimentally-evaluated subsidized employment programs have in turn reduced family public benefit receipt, raised school outcomes among the children of workers, boosted workers’ school completion, lowered criminal justice system involvement among both workers and their children, improved psychological well-being, and reduced longer-term poverty; there may be additional effects for some populations, such as increases in child support payment and improved health, which are being explored through ongoing experiments.
  • Subsidized employment programs can be socially cost-effective. Of the 15 rigorously evaluated (through experimental or quasi-experimental methods) models described in this report, seven have been subject to published cost-benefit analyses. Keeping in mind that more promising and effective models are more likely to lead to such analyses, all seven showed net benefits to society for some intervention sites (for models implemented at multiple sites) and some target populations. Four of these seven models were definitively or likely socially cost-effective overall.
Academic researchers are congenitally fond of ending their studies by saying "more research is needed." Isn't it always? But in this case, it seems a fair conclusion--and the good news is that the additional research has been underway for some years already and is fairly close to completion. Until then, this overview by Dutta-Gupta, Grant, Eckel, and Edelman is a solid summary and overview of the existing evidence.