# st: Heckman vs Tobit

 From Omrane Guedhami <[email protected]> To [email protected] Subject st: Heckman vs Tobit Date Wed, 24 Nov 2004 16:14:12 -0500

```Hi All,

I appreciate your feedback on the following questions.
I would like to compare the results of Heckman and Tobit. As I understand,  the
estimates from both models (Heckman and Tobit) are not directly comparable and
I have to compute the marginal effects on the probability of selection (y>0)
The steps are as follows:
1. Run Heckman:
. heckman pforeign Z1 Z2 Z3, twostep select (dforeign = X1 X2 X3 X4 Z1 Z2 Z3)
dforeign (selection) is a dummy variable whether foreign are present or not and
pforeign (regression)  is the % of foreign.
After heckman, I run mfx:
. mfx compute, eqlist(dforeign)  (predict, psel)
If I would like to compute the marginal effect of the independent variables on
the probability of being observed. Is this appropriate?
2.  I run a Tobit. The ultimate goal is to compare the results of both
approaches.
. tobit pforeign X1 X2 X3 X4 Z1 Z2 Z3, ll ul
I run mfx:
mfx compute, predict(pr(0,.))

However, there are two other variants of mfx (tobit case):
mfx compute, predict(e(0,.))
mfx compute, predict(ystar(0,.))

Which one computes the marginal effects of  the independent variables on the
probability of being observed (positive foreign or pforeign>0)? Is the computed
Tobit marginal effects directly comparable to those from Heckman?

A final question, if I would like to compare Heckman to Tobit, does this mean I
have to compare first step estimates (marginal effects) of Heckman (Probit) to
the Tobit or both equations of Heckman (Probit and OLS) to the Tobit?

Omrane

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```