Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: heckman like model with panel data (Not heckman, cluster)

From   Stas Kolenikov <[email protected]>
To   [email protected]
Subject   Re: st: heckman like model with panel data (Not heckman, cluster)
Date   Mon, 8 Nov 2004 16:54:20 -0500

I don't think that what you want is feasible. To my knowledge (based
on stuff like Wooldridge's black book), fixed effect probit cannot be
estimated, as there is no way to condition on the panel level error
term; and probit is in general a part of -heckman-. So I would be
surprised if there was any way to do fixed effect Heckman model.

-gllamm- should be able estimate a random effects model instead, but
you would need to put some effort into the data management, and
convincing -gllamm- that a missing RHS variable is a valid outcome,
since the missing values are, by default, dropped by -ml-. -heckman-
was programmed specifically to handle that as an exception, but I am
not sure there is an easy way to let -gllamm- know how to handle the
missing values.


On Mon, 08 Nov 2004 13:08:10 -0800, [email protected]
<[email protected]> wrote:
> I was wondering if any of you know if there is a code
> available in STATA to estimate a fixed effects selection
> model with panel data? Ideally I will like to estimate a
> model similar to a Heckman two steps estimator but using
> panel data. I know that Kyriazidou (1997) developed a model
> like this in Econometrica and she has a code in Gauss and I
> have heard that there is something similar available in
> Limdep 8 but I will like to know if there is something
> similar for STATA.
> I have seen in the statalist that some people have suggested
> for a similar problem to use heckman, cluster
> but I am not sure if that solution is really consistent. In
> particular I think it will not solve the problem of
> unobserved heterogeneity at the individual level that is
> solved using fixed effects, it will change the standard
> errors but the coefficients might still be inconsistent.  I
> have also seen some suggestions to use GLLAMM for this, I
> have to admit that I am not familiar with GLLAMM but again I
> am not sure if such procedure is consistent with econometric
> theory and if it is really suitable to what I want to do.

Stas Kolenikov
*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index