When doing propensity matching score (PSM), for the frther analysis in Difference-in-Differences, there is a question about whether we should match characteristics of firms before event date or we should match firm-to-firm due to characteristics for the whole sample still trigger me

From a discussion in Statalist, I saw researchers multiple propensity score getting from logit model to 1000*year. I understand that it is for matching observation by observation with the same year. I am owndering if it is the correct thought. If it is a correct thought, I am wondering fi there is any reference for that.

Secondly, is there any case that we used psmatch2 for the whole sample? I mean calculating propensity score for all observations then matching using psmatch2. From doing that I deem that I can get the coefficient of TREAT and POST in Difference-in-Differences afterwards. However, there would be a case that I may match observation in different years to each other. But it seems that the matching observations in different year is not a problem if we only conduct propensity score based on characteristics of firms before event, it is thought correct?

And based on the characteristic of psmatch2, in a panel data, it seems that we are not able to match firm with firm but observation-with-observation. I am wondering which approach is most suitable (calculating propensity score for pre-event or whole-sample characteristics. should we multiple propensity score by1000industry)



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Browse other questions tagged or ask your own question.