st: [Fwd: Two-step Probit after merging]

[mmolina@uniroma3.it]

Dear Statalist,

Thanks Marteen. Appart I agree that it depends on the nature of the data
but I was wondering that as the 2SIV and Newey¡¯s estimator do not
estimate ¥ä and ¥ë but rather

¥ä¥ñ = [1/(1-¥ñ©÷)¨ö]¥ä     and    ¥ë¥ñ = [1/(1-¥ñ©÷)¨ö] ¥ë

where ¥ñ is the correlation coefficient between ui and vi

and we assume that

¥íi = vi©¬ + ui. Because ui and vi are jointly normal, ¥íi is also normal

and I have a merged database of only two years (two observations for each
id) my concern is that I'm not controlling for the possible lack of
independence between the two observation.

Then my question is are these estimates necessarily biased?

Thanks a lot