Recently, there is an emerging line of the study said that the traditional two-way fixed effect(TWFE) is failed in a lot of case because of the heterogeneous effects of laws over time, follow some paper Goodman-Bacon (2019), Chaisemartin, 2020.

Especially, Goodman, 2019 did a great job to decompose the single post-treatment dummy. In his note, he answered the question "Is DD wrong":

Not in general. The DD research design—comparing outcomes for groups whose treatment status changes to groups whose treatment status does not change—still can be a good idea. The DD specification—estimating the coefficient a single post-treatment dummy—is a bad idea when your treatment effects vary over time (get bigger with time since treatment). In this case, just summarize your findings in a different way—event-study or a linear trend-break, for instance.

My question here is, if I expect and argue that the treatment effects get smaller with time since treatment, so whether DD specification now is a bad idea?


1 Answer 1


It depends on a few things. First, if you expect treatment effects change over time, then you want to estimate an event-study style DD specification.

If you have a single treatment timing (all treatment starts at the same time), then an event-study will unbiasedly estimate the treatment path. If you have variation in treatment timing, you want to use a modern method, e.g. Callaway and Sant'Anna (2020), Gardner (2021), or Sun and Abraham (2020)

  • $\begingroup$ Is there any reference or simple explanation for "* event-study style DD specification*" ? $\endgroup$ Jul 6, 2021 at 19:41
  • 2
    $\begingroup$ This is a pretty user-friendly intro guide: mixtape.scunning.com/… $\endgroup$
    – Kyle Butts
    Jul 6, 2021 at 23:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.