I am trying to estimate a difference in differences model. For each customerid I have either two observations (one before the treatment and one after) or one observation (only before the treatment). I would like to ask whether in this case it is possible to estimate a random effects model? Does it make sense to use also clustered errors (clustered by customerid)?

Secondly, I’m looking into matching methods to make my treatment group and control group more similar. If I match with replacement, I have duplicate rows in my dataset (the same customerid and period). Does this pose a problem in case I use a random effects model?

The regression I am trying to estimate: Y= B0 +B1 Treatment+ B2 Period+ B3 Treatment*Period

Any help is greatly appreciated! Thanks FG



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.