[Anthony ODowd]

Dear Joao Santos Silva ,
Thanks for your help.
When I run a Poisson with FE my observations drop from around 10,000 to around 1,000 because "of all zero outcomes". However when i run an OLS with fixed effects my observations stay around the 10,000 mark. How is this the case?
Thanks,
Anthony

[Jeff Wooldridge]

Anthony: That the linear FE estimator appears to use all the data is a mirage. If you look at the formula for the estimator you’ll see that units with all zeros contribute nothing to the estimator. For both estimators, such units are are uninformative for estimating beta.

[Anthony ODowd]

Thanks for the help, Jeff