did2s: pre-trend tests checking for parallel trend?

[Lorenzo Bombino]

Dear all,

I am using the command did2s to estimate a two-stage difference-in-difference model.

I am looking at the effect on candidates' vote share in 300 municipalities, following the treatment. The treatment is staggered. The dataset comprises three elections.

This is the sample code for my estimation:

did2s pref_share if female==1, ///
first_stage(i.mun i.yr age population immigration unemployment_ep) ///
second_stage(post_treat) treatment(post_treat) cluster(mun)

I would like to know what is the best procedure to test for pre-trend, ie. to check whether the parallel trend assumption holds.
I know the csdid command allows for instance for estat pretrend and I am looking for something similar.

Thanks!

[Maxence Morlet]

ssc install eventdd for example. Eyeballing a graph and testing for the joint significance of pre-trends coefficients for instance. But the most convincing test is argumentation: convince your reader that your treatment (timing) was random. Give context depending on your field and the specific policy.