search

chaisemartinPackages / did_multiplegt_dyn

|| Stata | R || Estimation of event-study Difference-in-Difference (DID) estimators in designs with multiple groups and periods, and with a potentially non-binary treatment that may increase or decrease multiple times.
38 stars 8 forks source link

Number of placebos for parallel trends assumption #76

Closed mblancoschroeder closed 2 months ago

mblancoschroeder commented 3 months ago

Hello Diego,

My dataset allows me to estimate 2 pre-treatment (placebos) and 4 post treatment periods. Is there any suggestion on how many placebos should I estimate in order to be sure that parallel trends assumption holds? I think 2 periods is not a lot, but as I don't have never treated I don't see how can I extend it more.

Thank you very much for your help,

Magdalena

chaisemartinPackages commented 3 months ago

Dear Magdalena,

In general, you would always want to estimate as many placebos as you estimate post-treatment effects to justify your claim that parallel trends hold for as many periods as you estimate your treatment effects. Looking at this issue might also be interesting as someone already posed a quite similar question some time ago. In your case, when your design is such that you can not estimate as many placebos as effects, we suggest estimating as many placebos as possible.

Best regards, Felix

mblancoschroeder commented 3 months ago

Hello Felix,

Thank you very much for your answer.

Best regards,

Magdalena