Matching for Whole Baseline period in Propensity Score Matching Difference in Differences

[Pranshu Tripathi]

Dear Members
I am working on unbalanced panel data.
I have 2017 as a treatment year and groupdummy as a treatment group dummy.
I have 2011-2024 as the period of study, where 2011-2016 is the baseline period and 2017-2024 is the endline period.
I want to run the psmatch2 command to match the groups in the baseline period.


But the issue is that most of the studies in panel data psmdid match at the end year of baseline.

For example: psmatch2 groupdummy x1 x2 x3 i.industryid, logit common noreplacement, if Year==2016

It is causing issues as when I do a placebo test using 2013 and 2014 as placebo years, then the effect of these placebos is significant.

What should I do to match the whole baseline period?

[George Ford]

You could match on the pre-treatment means.

egen x1_mean = mean(X1) if !post //and so forth

If you do placebos, then you can construct the means and match before 2013.

The fact the placebo's are significant is not a problem with the approach. It may just be true.

[Pranshu Tripathi]

Thanks, George
I will do it.

[George Ford]

you may want to use this:

egen x1_mean = mean(cond(!post,x1,.), by(whatever)

It will populate all rows unlike the simpler version.

and make sure you "by" using whatever the cross section is in either case.

[Pranshu Tripathi]

Thanks, George