In 2002, the Bush administration decided that the federal government needed to get in the business of promoting (heterosexual) marriage and proposed a new federal program—the Healthy Marriage Initiative—dedicated to that purpose. In 2005, the Republican-controlled Congress authorized the program and appropriated $500 million for it over five years. Although some conservative libertarians opposed the program, other conservative think tanks, including the Heritage Foundation, lobbied strongly for it as part of a conservative approach to welfare reform.

The first rigorous evaluation of a program funded with these dollars has just come out, and shows—surprise, surprise—that it's a failure. Conducted by Mathematica Policy Research, the evaluation assigned 5,000 unmarried couples—ones who were romantically involved and either expecting a baby together or already had a baby younger than three months—to program or control groups in eight U.S. cities. The program group received three services: 1) group sessions on relationship skills; 2) individual support from family coordinators; and 3) assessment and referral to support services. The group sessions were the core of the program—in essence, program participants received roughly 20 hours of "relationship skills education" that control group participants did not. The random-assignment structure of the evaluation allowed the researchers to isolate the effects these services had outcomes including marriage, living together, domestic violence, and relationship quality.

The researchers found that the program had no effect on the likelihood of couples staying together or getting married, and no effect on relationship quality (measured by variables including subjective relationship happiness) or other measured variables. There was some difference in effects by location—one of the eight sites had generally positive effects, while another had generally negative ones (including an increase in domestic violence). The researchers note that the site with positive effects (Oklahoma City) differed from the other ones in that it included already married couples, while the site with the negative effects (Baltimore) served a higher share of couples in less committed and more tenuous relationships than other sites.

There were some reported positive effects for couples in which both partners are African American, but, on closer examination, these turn out to be underwhelming. The program did not increase the share of African American couples who were romantically involved, living together, or married. Of the six statistically significant positive impacts for African Americans, most were related to "relationship quality," only one was significant at the .01 level, and the effects themselves don't look particularly large. For example, the one outcome significant at the .01 level was an increase in the "use of constructive conflict behaviors," which, as measured by a 4-point scale, was 3.22 for the program group compared to 3.14 for the control group. Moreover, the Baltimore site, which had negative results, had the highest share (92%) of African Americans. In sum, I'm pretty skeptical that the overall apparent positive results for African American couples are real rather than random.

The Healthy Marriage Initiative was funded through 2010 with money originally diverted from other more productive uses, including child care assistance and social insurance for low-wage workers. That was a mistake then, and, given these findings, one that shouldn't be repeated going forward.