Propensity Score Matching with Longitudinal Data

[Thilo Keppler]

Dear Statalists,

I try to use the user-written -psmatch2- for a PSM analysis in a longitudinal data setting.

I have five years of data from different companies resulting in a strongly balanced panel data.
Each company is assigned an unique ID and is either assigned to the treated group (dummy variable which equals 1 if a company is located in a certain country) or to the untreated group (=0).

I'd like to use the PSM to achieve a better fit between the treated and untreated group.
Is it possible to cluster the company ID in the PSM so that it regards the matching to the IDs and within the IDs the years available?
Otherwise it generates a matching between different years of different companies.

Thank you very much in advance,
Thilo

[Marcos Almeida]

Hello Thilo,

Welcome to the Stata Forum,

I'm not sure about the user-written psmatch2, but this is what we get from the Stata Manual on - teffects - command, related to PS matching:

Caveats and assumptions To use the estimators implemented in teffects, we must make several assumptions about the process that generated our data. Different estimators and statistics may require slightly more or slightly less restrictive assumptions and may exhibit varying degrees of robustness to departures from these assumptions, but in general, all the estimators require some form of the following three assumptions. The independent and identically distributed (i.i.d.) sampling assumption ensures that the outcome and treatment status of each individual are unrelated to the outcome and treatment status of all the other individuals in the population. Correlated data arising from hierarchical or longitudinal study designs do not meet this assumption.