DID parallel trend test with 1 year pre period and 1 year post peiod

[Yun Cheng]

Hello!

I have conducted a staggered DID where different firms' events happen in different years. For each firm, due to data limitations, I only obtain the event year and the year before. My DID includes TREAT, POST dummies and the interaction, TREAT*POST. The ols model includes industry and year FEs and standard errors are clustered at the industry level.

I have read some papers on how to conduct parallel trend test, however, they would typically have several pre and post years, so I can create pre3, pre2, pre1, post1, post 2 etc (relative to the event year, year 0), put those dummies into the regression, and expect the pre dummies to be indistinguishable from 0, and the post dummies to be significantly different from 0. For my DID, the situation is a bit different, I don't know how to test for that only 1 pre year. In talking about trend, I suppose there should be at least 2 years of pre. Any suggestion on how to test the parallel trend assumption? Thanks!

[Øyvind Snilsberg]

as you rightly point out, you need at least two pre-periods to assess the degree to which pre-treatment trends in the outcome are parallel. try instead to argue why you believe the not-yet-treated firms is a good counterfactual for the already treated ones.

[Jared Greathouse]

Not possible.

[Maxence Morlet]

The most important (if you're an economist) is argumentation. You need to argue the policy you're evaluating is plausibly exogenous and there was no non-random selection into treatment.

[Maxence Morlet]

Just like Oyvind Snilsberg pointed out.