To an economist (at least this one), the choice has nothing to do with panel
length. All random effects models are inconsistent if the individual
specific error component is correlated with any of the independent
variables. In the type of questions that economists typically study (for
example, the relationship between earnings and schooling) it is hard to tell
a story where this isn't the case (for example, the econometrician wishes to
use panel data to remove the confounding influence of unobserved ability on
earnings and unobserved ability is almost certainly correlated with measured
education). Along with these 'theoretical' arguments against random effects
models, econometrically a hausman test invariably also rejects the
consistency of random effects (at least in every panel model I have ever
estimated).
Basically, the within and first difference estimator are very powerful tools
for controlling for the confounding influence of invariant unobserved
heterogeneity. Random effects estimators can also be extended to allow for
restricted correlation with the independent variables (often called
correlated random effects models). These models are very flexible and in
some dimensions require fewer assumptions than the simple within estimator.
However, no canned routine is available in Stata (or most software packages)
for estimating these models, which seems to keep most people away from using
them.
Anyway, this is my two cents worth.
Steve
> -----Original Message-----
> From: Joseph Coveney [SMTP:[email protected]]
> Sent: Tuesday, July 29, 2003 1:36 AM
> To: Statalist
> Subject: Re: st: xttobit with fixed effect???
>
> David Reinstein asked whether there is a fixed-effect extension to
> -xttobit-.
> In responding, Steven Stillman mentioned that fixed-effects nonlinear
> models
> provide inconsistent parameter estimates.
>
> If the list will tolerate a couple of na�ve questions: not being an
> econometrician, I'm curious as to why a preference for fixed-effects
> models is
> often expressed for longitudinal data? Is it because the number of panels
> in
> econometrics studies is sometimes too few to justify handling panels as a
> random effect? If so, does it matter that a fixed-effect estimator is not
>
> consistent, since with a fixed-effect model, prediction and hypothesis
> testing
> are confined to the panels in the data in-hand?
>
> Joseph Coveney
>
>
>
>
> *
> * 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/
The information contained in this document is intended only for the
addressee and is not necessarily the views nor the official
communication of the Department of Labour. All final/official papers
which are sent from the Department will be sent by non-electronic
means, on appropriate letterhead, signed by authorised personnel.
*
* 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/
