Interaction effect in the 2nd stage of 2SLS

[james xu]

Dear all,
In 2SLS, if I were to test the interaction effect between a variable and the instrumented variable from the 1st stage, is it right that I use the following command:

ivreg2 dep_var controls (interaction_var##instrumented_var=instrument)

my goal is to test the interaction effect ONLY in the 2nd stage as part of my robustness checks, but it seems ivreg2 now uses the interaction term altogether as the instrument, which is not exactly what i wanted. I cannot use the fitted value from the 1st stage in my case because my instrument is at state-level while my 2nd stage is at firm-level. Using the fitted value would be inconsistent for my estimation in the 2nd stage.

Does anyone know how to implement the 2nd stage in this case correctly?

Much appreciated for any help and suggestions!

Pls let me know if anything above is unclear.

[Andrew Musau]

Note that interaction_var##instrumented_var = interaction_var instrumented_var interaction_var#instrumented_var. You don't want the interaction variable to be treated as an endogenous variable. Therefore, you want

ivreg2 dep_var controls (instrumented_var c.interaction_var#c.instrumented_var=instrument c.instrument#c.interaction_var)

[james xu]

Thank you, Andrew.