Following up on the points rightly mentioned by Rafa,
The traditional exclusion restrictions in the literature (the number
of children in the household, the presence of an active head of the
household and the expected variance of household income) are
unconvincing if you consider for instance that all these factors
impact the hours decision specially for women, and if you beleive in
the possiblity of division of labor in the household between husbands
and wifes.
Unfortunately, other selectivity correction methods won't get you out
of this dilemma... actually as far as I remember the Heckman method is
the only one that allows you to have the same variables in both
equations, precisely because of the nonlinearities implied by the
Normality assumption. Any semiparametric correction method will
require exclusion restrictions. Finally, the identification based on
the nonlinearity of the IMR is very weak, as it has been shown that
for a good range of values of the XB index the IMR is virtually
linear, so if you don't have a good spread of values in the index,
your identification might fail.
Robert
On 11/10/05, R.E. De Hoyos <[email protected]> wrote:
> Wasim/John,
>
> Two points here. (1) You have to think if observed hours of work are really
> the outcome of a utility maximizing process or if they are primarily driven
> by institutional rigidity. If institutional rigidities matter then almost
> all instruments will capture the effect upon participation (a
> utility-maximizing decision) without affecting hours of work (see Heckman
> 1990, AER). In many cases the density of hours worked is concentrated around
> two points, i.e. full time (40 hrs per week) and part time (20 hrs per
> week). The relative importance of the self-employed (and/or informal sectors
> in less developed countries) will make this distribution less concentrated
> around these two points. So you really need to think about the decision
> process behind the data. Depending on your data, it could be preferable to
> estimate a discrete choice model (multinomial logit or probit) to
> approximate participation and---indirectly---hours worked. (2) Some of the
> most common instruments in the literature of female labour participation
> are: the number of children in the household, the presence of an active head
> of the household (typically the husband or father) and the expected variance
> of household income (specially under borrowing constraints).
>
> I hope this helps,
>
> Rafa
> ________________________
> R.E. De Hoyos
> Faculty of Economics
> University of Cambridge
> CB3 9DE, UK
> www.econ.cam.ac.uk/phd/red29/
>
> ----- Original Message -----
> From: "Wasim Akram" <[email protected]>
> To: <[email protected]>
> Sent: Thursday, November 10, 2005 5:18 AM
> Subject: st: Heckman Method
>
>
> > Dear Statalist Members,
> > I am trying to estimate labor force participation and
> > work hour model using heckman's two step estimator. I
> > do not have wage data and consequently I am relying on
> > reduced from approach.
> > So far I have not used any exclusion restriction in
> > the first stage probit probit model and consequently
> > the entire identification strategy is based on the
> > normality assumption of the error terms. Many
> > economists argue to avoid such situation.
> > So I need a variable that effect labor force
> > participation but not work hours. I find it difficult
> > to identify such variable given the fact that labor
> > force participation and work hours are derived from
> > the same utility maximizing model.
> > Can any one suggest me of any such variable? Or should
> > I drop Heckman method and look for other method for
> > correcting selection bias? Which method?
> >
> > Thanks,
> >
> > Wasim
> >
> >
> >
> >
> >
> >
> > __________________________________________________________
> > Find your next car at http://autos.yahoo.ca
> > *
> > * 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/
> >
> *
> * 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/
>
*
* 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/
