biprobit with two endogenous variable

[Ketan Reddy]

Dear Statalist users,

I am trying to estimate a model where both dependent (y) and two endogenous variables (x and z) are binary in nature. Given the binary nature of both dependent and explanatory variable I proceed with biprobit estimation.

Now biprobit (X=Y) (Y=R) will be the standard procedure when only Y was endogenous. However, with two endogenous variables, how can I estimate the following

biprobit (X=Y) (Y Z= Instrument for Y, Instrument for Z)

[Fei Wang]

As you have three binary endogenous variables (including the DV), -biprobit- won't work. You may use -mvprobit- or -cmp-, user-written commands, for the multiple equation analysis. If you only care about the average partial effects of x and z on y regardless the binary nature of y, x, and z, then it's harmless to simply use -ivregress 2sls- with two endogenous regressors.

[Ketan Reddy]

Fei Wang Thank you!