Logit / probit regression with binary endogenous explanatory variable

[Christian Hackober]

Dear all,
I am currently facing an issue with regard to binary endogenous explanatory variables in probit or logit models and try to figure out what type of model to use and how to implement it.

I try to fit the following model:

y = α + β x1 + γ x2 + δx3 + e

y = binary DV
x1 = whether an individual company received an Venture Capital investment (1/0) (which is biased by a treatment effect and hence I want to account for the implied endogeneity)
x2 = continous variable
x3 = continuous variable

I found already an IV for x1 and thought about using ivprobit. However, given the binary nature of my endegenous variable, I assume that ivprobit should not work.
Do you have any suggestions which model could be used instead?

Thanks for your support in advance.

Bests
Christian

[Phil Bromiley]

If this were a continuous dv, then the binary endogenous variable would not be a problem. Look in the last couple of weeks on Statalist - I think Wooldridge commented on exactly your problem. It is also possible to do this directly in GSEM.