Hello there,
as it was suggested by a tenured member of the forum i should repost my questions another time. So I splitted up my questions (part one refers to the more general topic of control functions, it is seen in my last post).
This post concerns an actual empirical finding which gives me some hard time to interpret.
The main question is: I´m estimating and comparing a model by ols an various iv-estimators (one endogenous regressor and one instrument). For the baseline model OLS and IV-estimators are pretty much the same. But as soon as I integrate interaction terms and use multiplicative constructed variables for the interaction terms in the second stage, everything is messed up. While OLS demonstrates significant coefficients, the IV estimators demonstrates no significance on the interaction terms at all and completely other coefficients.
In order not to spam the forum, for a concrete example i would like to refer to the following links, where i´ve already posted some of the results and the used stata code:
1. Statalist --> http://www.statalist.org/forums/foru...nous-variables
2. Cross Validated --> http://stats.stackexchange.com/quest...eraction-terms
Best regards
Fred
as it was suggested by a tenured member of the forum i should repost my questions another time. So I splitted up my questions (part one refers to the more general topic of control functions, it is seen in my last post).
This post concerns an actual empirical finding which gives me some hard time to interpret.
The main question is: I´m estimating and comparing a model by ols an various iv-estimators (one endogenous regressor and one instrument). For the baseline model OLS and IV-estimators are pretty much the same. But as soon as I integrate interaction terms and use multiplicative constructed variables for the interaction terms in the second stage, everything is messed up. While OLS demonstrates significant coefficients, the IV estimators demonstrates no significance on the interaction terms at all and completely other coefficients.
In order not to spam the forum, for a concrete example i would like to refer to the following links, where i´ve already posted some of the results and the used stata code:
1. Statalist --> http://www.statalist.org/forums/foru...nous-variables
2. Cross Validated --> http://stats.stackexchange.com/quest...eraction-terms
Best regards
Fred