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

Re: st: Heckman on panels


From   Stas Kolenikov <skolenik@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Heckman on panels
Date   Wed, 20 Oct 2004 08:14:58 -0400

It is (ii), no good answer. Heckman's model is complicated enough per
se, and putting another layer of complexity with panel is not an easy
task. The simplest thing to do is to make -cluster()- correction, if
-heckman- syntax allows that.

if you are desperate, try -gllamm- with three levels: level 1
observations are (a) your probit decision and (b) continuous response;
level 2 units are those pairs, so that the unobserved random effect
allows for the correlation of the two, and is specific to individual
and time; level 3 units are individuals.

Hope this helps, and think over Nick's comments. A good email starts
with "I read the [R] and [U] manuals, I searched Stata FAQ and
Statalist archive, but cannot seem to find an answer to the following
question: ...". (Soon we would have to add, "... and -gllamm- manual
... " :))

Stas

On Wed, 20 Oct 2004 09:52:32 +0200, hwiig <henrik.wiig@econ.uio.no> wrote:
> Hello again everyone! No reply yet on my question I posted yesterday, I
> just need to know whether its because (i) There is a technical error
> somewhere, and the question never reaches you? (ii) Nobody knows the answer
> or (ii) the question is so "stupid" that nobody bothers to answer. Anyway,
> I try again:
> 
> >Hi!
> >
> >I got a panel data set of some small scale farmers in developing
> >countries. Some sell volume X at individual price P, some don't sell at
> >all. The control variables vector Z for all farmers is available.
> >
> >Heckmans method controls for self selection by (i) estimating a probit for
> >sell / no sell (ii) introducing the resulting Mills ratio in the sales
> >equation before running a normal OLS (on the households that do sell their
> >product).
> >
> >I got information from two years, so I wonder whether there exist some
> >kind of Stata procedure to correct for estimated individual effects
> >(either fixed or random).
> >
> >The Stata command "xtprobit" does exactly this for a isolated first step, so
> >
> >(1) How do I introduce xtprobit in the first stage of the Heckman in order
> >to correct for individual (unobserved) differences in choosing to sell or not?
> >
> >(2) How do I introduce individual differences the sales function (second
> >step) of the Heckman model (i.e. also individual unobserved differences in
> >sales volume for the farmers who do sell)?
> >
> >Anyone out there who knows how to solve this? I would most appreciate your
> >help!
> 
> Henrik
> 
> Henrik Wiig
> Department of Economics
> University of Oslo
> PO Box 1095 Blindern, N-0317 Oslo, Norway
> Eilert Sundts hus, 12th floor, Moltke Moes vei 31
> Telephone: 22 85 51 35
> Fax: 22 85 50 35
> E-mail: henrik.wiig@econ.uio.no
> Private telephone: 22 55 24 84
> Mobile telephone: 47 75 75 09
> http://folk.uio.no/hwiig/
> 


-- 
Stas Kolenikov
http://stas.kolenikov.name
*
*   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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index