How to estimate a binary panel model with endogeneity in Stata?

[Pablo Avendano]

Hi,

I’m thinking about estimating a binary panel data model (logit or probit) with an endogenous variable. I’d like to know if Stata offers any option to correct for endogeneity in this type of model.

So far, I’ve considered estimating models like ivprobit, xteprobit, simple logit, spatial logit, and comparing them with and without clustered errors to see how the results differ. However, I’m not sure if any of these are suitable for my case.

Could anyone suggest how to approach this estimation, or what alternatives exist?

Thanks!

[Jeff Wooldridge]

Is the endogenous variable (roughly) continuous, or is it also binary?

[Pablo Avendano]

Hi professor Wooldridge, the variable is continuous.

[Jeff Wooldridge]

Assuming you have a time-varying instrument, you can combine the correlated random effects and control function approaches. We explain it in this paper:

https://www.sciencedirect.com/science/article/pii/B9780128143674000022

I can send the Stata do files for the fractional probit case; the modification for a binary outcome is essentially trivial.

[Pablo Avendano]

Thank you for the answer. I would greatly appreciate it if you could send me the Stata do files to my email: [email protected]. Thank you in advance for your help!

[Pablo Avendano]

Assuming you have a time-varying instrument, you can combine the correlated random effects and control function approaches. We explain it in this paper:

https://www.sciencedirect.com/science/article/pii/B9780128143674000022

I can send the Stata do files for the fractional probit case; the modification for a binary outcome is essentially trivial.

Hi professor, I'd really appreciate it if you could kindly send the Stata do files to my email: [email protected]. Thanks in advance for your support!