Did_imputation and event study analysis

[John lenon]

I want to perform an event study analysis. I have a staggered did setting. I observe the data at a bi-annual level from 1999-to 2019 at the state level. I have never performed event study analysis and I want some help with the code and some basic question.

1) Can I do an event study when there are a couple of states treated since the start of the data?
2) Can I perform an event study analysis when the outcome variable is binary? What if there are many zero, can I still perform event study?
3) Do I use the covariates for event study regression that I have used for DID analysis?
4) Finally, how do I perform event study analysis. I am so confused. I cant figure out with the online codes or materials.
5) I have read about new event study methods such as DID_Imputation. Does anyone know any code to use the command.

My regression command for TWFE

y(ist) = beta1* Treatment(ist) + beta2 * X(ist) + beta3* Statecontrols(ist) + state + time

Treatment = 1 if the state is treated at time t.
Can anyone give me an idea of how would I perform event study analysis? Thank you

[John lenon]

Hello everyone!

My reg model suggests increasing in y variable significantly. However, when I do event study analysis, the graph suggests that y variable is decreasing with time significantly. Is this possible or I am might be doing something wrong. Let me know your response. Thank you

[Jared Greathouse]

I know what you're looking for, but I don't advise you use it unless you have sophisticated understanding of the methods they're positing. New DD methods are not for the faint of heart.



There's no reason to think there's anything peculiar with your second question. Before you worry about signs or directions of effects, worry about it you've set up your design correctly from an econometric standpoint