I have a pre-/post-treatment-control design (diff-in-diff) with one treatment event. The data is a firm panel data set (5 pre and 5 post firm-years, ~200 firms). The problem - my treatment and control sample show significant differences on multiple covariates (also in the pre-period).
Therefore, I want to balance the two samples via entropy matching. Now my question: What is the right way to match the treatment and the control sample here? My suggestion would be to compute entropy weights using the average value (by firm) for each covariate in the pre-period, compute the weights by firm (in the pre-period) and then use the weights also for the post-period.
Is that the correct way?