ivreg2 can instrumental variables be put out of the parentheses

[Johnny YANG]

I am reading a paper stating about their instrumental variables and controls. It's really weird because they used variable A as instrumental variable and then they used A also as control. I tried their model, the command is as follows,
ivreg2 Y (X=A B) C A B
The fact wether we put A and B as controls (out of the parentheses) has a big impact on both the Sargan statistic, weak instrumental test and the coefficient estimations. My question is can we really use A and B as instruments and controls at the same time?

[Joro Kolev]

There are two separate issues here.

The theoretical issue: In Two Stage Least Squares the instruments and the exogenous regressors are fundamentally the same type of variables. The first stage is fundamentally a projection on all exogenous variables (both excluded instruments and included exogenous controls).

And then there is the programming issue of what exactly the user written -ivreg2- does, and how exactly it interprets the syntax you are typing. For this latter issue you need to read the help file of -ivreg2-.