PSM - Propensity score matching on control group never treated

[Elton Ferreira]

Hi everyone,

Help about PSM - Propensity Score Matching: Is it possible to apply PSM to a sample whose control group has never been treated?
Context: I have a sample of companies that are divided in 2 groups:

1) Treated: companies that are impacted by a new regulation - specific of your country and the implementation occured in different years (e.g. companies of country "A" in 2016, country "B" in 2019...);

2) Control: companies that has never been treated because their countries do not have regulation.

Objective: use PSM as a supplementary test to DID - Difference in Differences to assess the impact of this new regulation.

Thanks!

[Jeff Wooldridge]

PSM shouldn’t be used as a supplement to DiD because PSM assumes unconfounded assignment in levels, whereas DiD only assumes it in changes.

You can apply PSM after differencing the data across time, assuming you have a panel.

[Elton Ferreira]

Thanks Wooldridge.

Perfetct! I have a panel data and I will aplly PSM after differencing the data across time to campare the results.

I would appreciate if anyone has any suggestions on how to do this in stata.