Using CMP for selection bias with binary explanatory variable

[Philippe Vaz]

Hello everyone,

I need help to conclude my work.
I've been working with logit regressions and I have my final results. To conclude this I need to test for selection bias. Usually, a 2 step Heckman model would solve the problem. The thing is my main explanatory variable is already a binary one, so I can't have this variable in the main equation and in the selection equation (which is a probit) at the same time, due to endogeneity.

I was told I could solve this using the cmp model, but no one was able to explain me the process. Is it possible? If so, which command should I use?

Thank you

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

There has been a lot of discussion of similar issues on Statalist. I suggest you look at it. I don't really understand your problem. Why exactly do you think having a binary rhs variable (that is not the selection variable I assume) invalidates heckman?

[Philippe Vaz]

Thank you for your answer Phil!
I'll try to read some topics here on the forum

The problem isn't having a binary variable. It's that in the heckman, the dependent variable in the selection equation can't be the same as the main explanatory variable in the main equation. If I run the model with the same variables I'll have an endogeneity issue.
That's the reason why I'm trying to use CMP, so I can use the same variable in both equations..

Regards,
Philippe