Can fixed effect absorb the anticipation effect?

[Kris James]

I'm working on a research regarding natural disasters across counties in the US. I'm using these events as exogenous shocks for my Difference-in-Difference analysis. One assumption for the DiD is the absence of anticipation. However, some disasters are not random and can happen within a particular period of the year (E.g., Wildfires in California often occur around June-July). I think this might cause some anticipation for these states/counties. My question is that how should I account for this potential anticipation? If I include county/state fixed effects will the anticipation effect be absorbed?

Thank you for your answers!

[George Ford]

This is an interesting problem. I'm subscribed and await answers.

I would guess that a time fe would be required to catch the anticipation (so you get June/July without a fire, and then with a fire by the DID dummy, but the anticipation effects might be in May), but that depends on how the data is organized (monthly data?).

What I feel confident about is that you're going to be asked for more specifics to get a good answer.