Hi, I have a more general econometric question.

I am trying to estimate the causal effect of retirement on health using IV 2SLS. Retirement is a binary outcome, and health is categorical. I am using OLS in both stages, even though this do not provide the most efficient results. But, I am wondering how to interpret the coefficient when I use OLS on a categorical outcome.

E.g., I have a variable taking on 12 categories: 1, 2, ... 12., counting number of specific diseases. When I use stata command "reg" to estimate the effect of retirement on this health-variable, in the second stage (using OLS), what is the interpretation then?

If the variable was continuous instead, it would have been straight foreward. Or if binary, it would also be straight foreward using the linear probability regressions. But I am in doubt when my variables are categorical.

(Extra question: to get more efficient estimates, I could have been told I could exploit a control function method instead of IV. This method is newer and more complicated, but is there a built in function in STATA for this method?)

Thank you.