Heckman two stage procedure

[Theresia Verena]

Hello,
I want to use the Heckman Two Stage Procedure, but I don't understand why the selection equation in heckit uses probit instead of logit. Can someone explain it to me? Thank you in advance.

[Øyvind Snilsberg]

It might help to think about what are the components of the inverse Mills ratio and how would one go about estimating them.

[Fei Wang]

Heckman (1979) assumes normally distributed errors in the first-step latent-variable regressions, and that's why probit is used. I guess one can derive another set of equations by using logit in the first step, but Heckman won't be happy to see it called "Heckman" two-step procedure. For censored data, logit can be used in the first step of a two-part model, it's statistically similar to Heckman's procedure but applies to different situations.

[Theresia Verena]

Thank you very much for Fei Wang and Øyvind Snilsberg