Dear all,

I'd need some advice on how to deal with fixed effects. I am currently working on a cross-sectional dataset and would like to include in my models county fixed effects. Note that this is relative to all 3,000 US counties. I have encoded my "county" variable into a numeric one. Would it be okay to just include the county variable as it is or should I include i.county? The former works just fine, but I doubt that it is the right way to go. The problem with the latter is that, since there are 3,000 counties, Stata slows down by a lot and sometimes crashes. When it does not crash, the usual regression output cannot be seen as all the space is taken up by the coefficients for each of the 3,000 county dummies. Is there a way to avoid this issue? Note that I would also need to run IV regressions, so, for instance, using xtreg, fe would not be a solution for me.

Thank you very much in advance.