An op-ed by Angela Rachidi and Robert Doar in Real Clear Policy, reprinted on the American Enterprise Institute’s blog, challenges some points I made about the effects of work requirements in my commentary on the topic last week. Rachidi and Doar’s critique focuses on the interpretation of data from the National Evaluation of Welfare to Work Strategies (NEWWS), which analyzed the results of 11 controlled experiments conducted around the country in the 1990s. Each experiment compared the experience of a group of welfare recipients who were subject to work requirements with a control group who did not face work requirements.

Rachidi and Doar point out that some of the programs were “jobs-first” programs that placed a primary emphasis on job placement, while others were “education-first” programs that emphasized human capital development before job placement. They claim that only the jobs-first variants are relevant to today’s debate over work requirements for noncash welfare such as SNAP, Medicaid, and housing assistance. Linking specifically to my commentary, they maintain that critics:

incorrectly cite the results from the education-focused programs, which were found to be largely ineffective compared to jobs-first programs, to suggest that work requirements would not be effective today in SNAP or Medicaid. The education-first programs are largely irrelevant to today’s discussion of extending work requirements to other safety-net programs — no one proposes that recipients be required to go to college in order to receive SNAP or Medicaid.

Fair enough. Let’s take a closer look at the results of the jobs-first experiments to see how effective they really were in the three cities (not four) where such experiments were conducted side-by-side with the education-first variant. (And, by the way, the education-first variants did not focus on “going to college,” but rather on basic education, high-school equivalency, and vocational training.)

Rachidi and Doar claim that:

The jobs-first programs in NEWWS most relevant for today’s discussions showed impressive gains in employment and earnings. In those programs, increases in the average number of quarters worked in a five-year follow-up period ranged from 8 percent to 21.1 percent, with similarly high increases for average total earnings.

That sounds impressive until you look at the actual numbers. Let’s start with the effects on employment. The NEWWS report measures success in finding work by the number of calendar quarters, out of the five-year experimental period, when participants did any paid work. Since that could mean just a few days of work over a three-month period, it is a considerably lower bar for success than stipulated by today’s work-to-welfare reformers. For example, the recent report on work requirements from the Council of Economic Advisers envisions a requirement of 20 to 30 hours per week, depending on the number of children in a participant’s family. Here are the results, taken from Table 4.1 of the NEWWS report and stated in quarters worked over the 20-quarter experimental period:

  • Atlanta: Control group, 7.8; Jobs-first group, 8.5; education-first 8.3
  • Grand Rapids: Control group, 9.1; Jobs-first group, 9.8; education-first 9.5
  • Riverside: Control group, 4.7; Jobs-first group, 6.0; education-first 5.5

Two things stand out here. First, the absolute gains in employment were distinctly underwhelming in all cases. In Atlanta and Grand Rapids, the jobs-first approach produced less than one added quarter in which any work at all was done out of the 20 quarters of the experiment. In Riverside, the gain was less than two quarters out of the 20. Second, the differences between the jobs-first and education-first approaches were trivial. The jobs-first approach resulted in just 0.2 more quarters worked Atlanta, 0.3 in Grand Rapids, and 0.5 in Riverside.

The results for total earnings were similarly unimpressive. Here they are, stated as the average annual increase in earnings for participants in the two experimental groups compared with the control group:

  • Atlanta: Jobs-first group, $492; education-first $403
  • Grand Rapids: Jobs-first group, $310; education-first $169
  • Riverside: Jobs-first group, $456; education-first $272

Again, the absolute gains in income are small, and the difference in gains from jobs-first to education-first approaches is downright trivial. In Atlanta, the jobs-first group earned $1.71 more a week on average than the education-first group. In Grand Rapids, the difference was $2.71 a week, and in Riverside, $3.54 a week. Even allowing for the change in the cost of living since the 1990s, two or three extra dollars a week is not much progress toward self-sufficiency.

The greater success of the jobs-first approach in Riverside than in the other cities comes with an interesting twist. The Riverside experiment was different from the other two, in that it accepted only applicants without a high-school diploma or GED. Table 1 of the NEWWS report breaks out the earnings results for all three cities according to education. In those cities, too, the work-first approach showed relatively strong results for participants with less than high school, but not for those with more education. In Atlanta, the work-first approach had no significant advantage for participants with a high-school education or more, and in Grand Rapids, the education-first approach actually produced a greater gain in earnings than the work-first approach. Since any real-world work-to-welfare policy would cover people at any level of education, the overall results from Atlanta and Grand Rapids offer a more realistic idea of what such a policy could accomplish. Including results from Riverside is misleading.

The bottom line: Rachidi and Doar are technically correct to point out that the jobs-first variants of the NEWWS experiments were somewhat more successful and more similar to today’s work-requirement proposals than the education-first variants. However, the absolute differences in the effects of the two approaches were quite small, whether stated in terms of employment or of earnings. Most importantly, neither the jobs-first nor the education-first variants of work-requirements came close to producing the dramatic gains in self-sufficiency that today’s work-requirement advocates hope to achieve.