I have a panel dataset where I am trying to estimate the effect of receiving a grant on publication. I had hoped to use differences-in-differences estimator, but the problem is that many people receive grants multiple times over the course of the dataset, so they aren't simply pre- and post-treatment. I have not been able to find a technique for dealing with treatments that occur to the same person multiple times. I also can't use a matched sample because people who have never received grants are not similar to those who have. Does anyone know of a technique?

  • $\begingroup$ You can recast the DID as a fixed effects regression where you include the post and the interaction of post and treated in the model. The interaction coefficient is the DID effect. Recipients can stay treated as long as the grant is active. If people can have simultaneous grants, you can have a series of post and post-treated interactions, one pair for each grant. Have you considered this approach? $\endgroup$ – Dimitriy V. Masterov Dec 2 '15 at 1:28

Your Answer

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

Browse other questions tagged or ask your own question.