I am working on the following two models.

Logit: Binary outcome $y$ ($y=1$ if paid more than the item value to win the item, $y=0$ otherwise) $$ y=b_0+x_1b_1+x_2b_2+x_3b_3. $$

OLS: Continuous dependent variable $z$ (amount of overpayment) $$ z=b_0+x_1b_1+x_2b_2+x_3b_3+e. $$

My questions:

  1. Do I need to estimate these two models jointly (instead of separately) since the dependent variables seem to be related?
  2. If joint estimation is a must, which method (or package in Stata) should I use? Seemingly unrelated regression suest?
  3. Any other thoughts in terms of methodology?

1 Answer 1


I think you are looking for the Heckman model (sample selection model/Heckit model). It is a two stage estimator. In Stata, you use command heckman for that.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.