@HUDgov published 20-year RCT results for Jobs Plus, providing public housing residents with employment services along with financial incentives & supports for work. My quick take: No impact on earnings & employment for the full 6-site sample, & the reported positive impacts at 3 sites aren't reliable.


  • HUD's Jobs Plus demonstration, over 1998-2003, provided public housing residents with employment-related services at onsite job centers, rent-based work incentives allowing residents to keep more of their earnings, & activities to promote neighbor-to-neighbor support for work.

Study Design & Earlier Findings:

  • The study randomized 15 housing developments at 6 US sites to treatment vs control.

  • An earlier, 6-year follow-up found small earnings impacts for the full sample & more sizable impacts at the 3 best-implementing sites. (The latter is a post-hoc subgroup analysis, so is only suggestive.)

New Findings:

  • In years 20-21 after program launch, no impact on adults' employment & earnings for the full 6-site sample. At the 3 best-implementing sites, the study reports statistically significant (or near significant) impacts on earnings (11%) and employment (4% points).

  • But, as the report notes, these findings of statistical significance are based on an incorrect analysis of the data as if the study had *individually* randomized 4,105 adults at the 3 sites (whereas it was a cluster RCT that randomized 7 housing developments at these sites).

  • When analyzed correctly as a cluster RCT in a study appendix, these effects weren't close to statistical significance (2-sided p=0.50 for earnings & p=0.34 for employment) - so are only suggestive (not reliable) as they could well be due to chance.

  • The study report notes this limitation and describes the findings as "somewhat suggestive."

We use cookies to improve your experience and to help us understand how you use our site. Please refer to our cookie notice and privacy statement for more information regarding cookies and other third-party tracking that may be enabled.

Intuit Mailchimp logo