I am trying to estimate this following triple differences regression in order to test for parallel trends in the difference in differences:

enter image description here

My treatment is binary (it is a law that was passed in one the treated country (c) and not in the control country). I essentially need to see that the coefficient on the three-way-interaction is not significantly different from zero, but I am unsure about how to cluster my standard errors.

I have tried to find an answer online but none of them explicitly talk about this problem in the context of only 2 clusters (a treated and a control country) and a triple difference estimate - I have about 30 years of data, so multiple time periods.

  • $\begingroup$ See section VI in Cameron and Miller's JHR paper. Two clusters are almost certainly insufficient for the asymptotics to kick in, so you will over-reject. The DDD issue is not particularly relevant for this problem. It might also help if you explain what g indexes here and the sample size there. $\endgroup$ – Dimitriy V. Masterov Mar 11 at 22:17

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.