In Dehejia and Wahba (2002), after the average treatment effect column, there is also a new column "regression treatment effect". I know the average treatment effect is just the average difference of the treated and the matched controlled outcome. I am not sure what the regression treatment is supposed to be and how can I obtain it from Stata.

  • $\begingroup$ You need to use a purposeful strategy. See for example BBR chapter 17: fharrell.com/p/blog-page.html $\endgroup$ Oct 21, 2017 at 15:07
  • $\begingroup$ Minimal name and date references should (please) be fleshed out as if giving a journal reference in a paper or dissertation (or, if more appropriate, a stable URL). You shouldn't assume that people are familiar with the literature you know. $\endgroup$
    – Nick Cox
    Oct 23, 2017 at 17:51
  • $\begingroup$ Correct link: hbiostat.org/bbr $\endgroup$ Apr 20 at 16:18

1 Answer 1


On page 154 of D&W (2002)$^*$, it says

In addition to using a weighted difference in means to estimate the treatment effect, we also consider a weighted regression using the treatment and matched comparison units, with the comparison units weighted by the number of times that they are matched to a treated unit. A regression can potentially improve the precision of the estimates

This is a regression of the form

reg outcome_var i.treated age education ... if matched==1 [iw = weight_var]

$^*$ Propensity Score-Matching Methods For Nonexperimental Causal Studies by Rajeev Dehejia and Sadek Wahba, The Review of Economics and Statistics, 2002, vol. 84, issue 1, 151-161


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.