Running Two-way Fixed Effects with Multiple Treatments in did_multiplegt

[Roberto Santos]

Hi, I am trying to run a two-way fixed-effects (TWFE) model with multiple (actually just two) treatments in did_multiplegt. According to this paper ("Two-way fixed effects and differences-in-differences estimators with several treatments"), that should be possible. However, I have not been able to find any documentation or examples of how to do so. If anyone can show me an example of the syntax I should use in order to have two treatments in my regression that would be great. Thanks!

[George Ford]

help did_multiplegt.

Can I estimate the effect of a treatment, controlling for other treatments that may also change over the panel?

The command can compute the DIDM estimator, controlling for other treatments that may change over the panel, as proposed by de Chaisemartin and D'Haultfoeuille (2020c). In that case, let othertreat be a variable containing the other treatment, and
let fd_othertreat be the first difference of that other treatment. To compute the estimator proposed by de Chaisemartin and D'Haultfoeuille (2020c), one should write if_first_diff(fd_othertreat==0) trends_nonparam(othertreat) always_trends_nonparam
in the command's options. With two rather than one other treatments, one should write if_first_diff(fd_othertreat1==0&fd_othertreat2==0) trends_nonparam(othertreats) always_trends_nonparam in the command's options, where othertreats is generated as
follows: egen othertreats=group(othertreat1 othertreat2). Etc. for the cases with three, four ... other treatments. See de Chaisemartin and D'Haultfoeuille (2020c) for further details.