In my model I have one endogenous variable of interest x and 20 control variables, out of which one may be endogenous as well. I am basically not interested in the coefficient of my control variables. Do I have to instrument for both/all variables to have an unbiased estimate for x my variable of interest ? It is sufficient to just instrument for x, as I my instrument for my control variable is very weak. Secondly, I am confused about the ivreg2 command. Does the order matter how I set it up to instrument for multiple endogenous variables, e.g. ivreg2 y (x c1=i1 i2) c2 c3... versus vreg2 y (x c1=i2 i1) c2 c3... ?
-
Login or Register
- Log in with
