[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

# st: RE: estimating a heckman sample selection model with panel data

 From "Alejandro Lopez-Feldman" To statalist@hsphsun2.harvard.edu Subject st: RE: estimating a heckman sample selection model with panel data Date Wed, 15 Jun 2005 06:25:14 -0700 (PDT)

```Colin,

It seems to me that your procedure might work well, altough it will only
work for random effects and not for fixed effects since xtprobit is only
valid for random effects. Also, as you do with the regular heckman, I think
that you will have to obtain the errors in a way that accounts for the fact
that you are including a predicted value (instead of an observed one) in
the second stage.

There are two alternatives that I know of to get fixed effects (or
something close to it) this are the references:

Kyriazidou, E. (1997). Estimation of a Panel Data Sample Selection Model.
Econometrica. 65 (6), pp. 1335-1364.

Wooldridge, J. (1995). Selection Corrections for Panel Data Models under
Conditional Mean Independence Assumptions. Journal of Econometrics. 68, pp.
115-132.

Alejandro
>
> ----- Original Message -----
> From: <Colin.Vance@dlr.de>
> To: <statalist@hsphsun2.harvard.edu>
> Sent: Wednesday, June 15, 2005 4:08 AM
> Subject: st: estimating a heckman sample selection model with panel data
>
>
>> Hello Everyone,
>>
>> I have a panel of data and would like to estimate a Heckman sample
>> selection model. This can apparently be done with gllamm, though I'm
>> still struggling a bit with the code and interpretation (and may have
>> questions on that later).
>>
>> My current question is whether the following simple alternative could
> be
>> availed: 1. estimate the random effects probit part of the model using
>> xtprobit 2. calculate the inverse Mills ratio from the results, which
>> equals normden(linear_pred)/norm(linear_pred) 3. Include the Mills as
> an
>> explanatory variable in the second stage regression to control for
>> selectivity bias. In the second stage regression, one would have to
>> decide between a fixed effects and random effects specification. And I
>> believe one would also want to use the robust option.
>>
>> This approach seems simple enough, but I'm hesitant because I've never
>> seen it done in the literature. Instead, most articles on the panle
>> Heckman are fairly equation dense, and few of the findings have made
>> their way into statistical software (with a few exceptions like
> gllamm).
>>
>> If anyone has any insights as to whether the suggested alternative is
>> defensible, please pass them along.
>>
>> Many thanks,
>>
>> Colin
>>
>> Colin Vance, Ph.D.
>> German Aerospace Center
>> Institute of Transport Research
>> Rutherfordstrasse 2
>> 12489 Berlin
>> Germany
>>
>> tel: +49 30 67055147
>> fax: +49 30 67055202
>> email: colin.vance@dlr.de
>>
>>
>> *
>> *   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/
>

-----------------------------------------
Alejandro Lopez-Feldman
Ph.D. Candidate
Agricultural and Resource Economics
University of California-Davis
*
*   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/
```

 © Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index