Hi everyone!

Because I am lacking parallel trends, I have decided to study the effect of X on Y using propensity score matching or weighting in order to try and get a double robust estimator. I have panel data for most European countries (44 of them) between 2003 and 2016 and I want to study the effect of the EU enlargements in 2004 and 2007 on Y in the EU countries that were members before the enlargements (called the EU15 in 2004 and EU25 in 2007); the control group is thus European countries that are not EU members. This means that the 12 countries that joined in 2004 and 2007 are not of interest when looking at the 2004 enlargement and the 2 countries that joined in 2007 are not of interest when looking at the 2007 enlargement.

My questions are:
- Is matching or weighting preferred?
- How does one implement propensity score matching/weighting on panel data in STATA? And, specifically, how does one match countries over all years on pre-treatment covariates only? So far I haven't found any good tutorials.
- Is there any limit to the number of covariates I should match on? Background variables are readily available from the World Bank Development Indicators.
- Last but not least, in a difference-in-difference approach, my variable of interest would be the interaction between being an EU member and being in the post-treatment period. However, am I right in this case to match on EU-membership and to later estimate the double robust estimator for post-treatment or will the variable of interest still be the interaction term between the two?

Very grateful for any help I can get, thanks in advance!