Wild bootstrap standard errors with S&A estimator

[AnneSimon233]

Hello,

Thank you very much for developing this estimator and package. I'm using your estimator in one of my PhD projects, which works really well. One thing that I still haven't figured out yet, though, is whether it is possible to use wild bootstrap standard errors with the aggregate estimator (i.e., I don't need this for the event studies, but for the overall ATT that I estimate). I need to run this robustness check to use a more conservative approach with a small number of clusters (11 in total). Is that possible in Stata (or R), and, if so, how can I do that?

Best,
Anne

[lsun20]

Thank you for your interest in the package! I am not sure what bootstrap command you are using, but from Stata's built-in bootstrap command, you can define a program that returns the overall ATT estimate as you are currently doing, and then call this program for bootstrapping (see Example 4 in their documentation ). Perhaps this is what you are looking for?

[AnneSimon233]

Dear Liyuing,
my apologies for never getting back to this and thanks a lot for your thoughts on how to employ bootstrap standard errors with the S&A estimator. I've never done something like that in Stata, so for now I decided to use wild bootstrap standard errors in combination with TWFE (because the results are very similar to those from the S&A estimator), but I might try your approach as well.
Kind regards,
Anne