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

Re: st: dynamic short panel


From   Giovanni Bruno <giovanni.bruno@uni-bocconi.it>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: dynamic short panel
Date   Thu, 29 Sep 2005 11:17:58 +0200

sistoand80 <sistoand80@libero.it>:

> Dear Dr. Bruno,
> I'm Andrea Sisto, a PHD student in Economics at the Univeristy of Turin. As
> I'm working with Prof. Zanola, I thank you also on behalf of Roberto. Our
> problem is that we have to account for heterogeneity in panel. This
> heterogeneity should be captured introducing cross-section dummy variable.
> But standard GMM DP estimators drop individual effect with first-difference.
> For ivreg2, the question was whether a FE2SLS, with a model in level with
> Cross section dummy variables, was a correct way to deal with dynamic  panel
> (our strategy was to simply instrument the dependent variable with some
> exogenous instruments).
> Thank you for your suggestion
> AS

Careful here. Time invariant individual heterogeneity is 
typically accommodated either  1) by adding individual dummies,
which boils down to transforming variables in deviations from the 
group means; or 2) by first-differencing. So at the end of the day, 
*both* methods actually do the same job of purging the regression 
from unobserved individual heterogeneity. 

First-differencing is held as a more convenient transformation 
in dynamic panel data models since under conventional assumptions 
on the var-cov matrix of disturbances in levels first-differenced 
disturbances are not correlated to past realizations of the 
dependent variable, which can be used as instruments.  A reason not 
to do that is a non-spherical var-cov matrix of disturbances in 
levels, e.g. due to within-group serial correlation. In this case, 
lagged values of y may not be used as instruments, but lags (and 
leads) of strictly exogenous explanatory variables can always be 
exploited to identify the relationship of interest (in levels or 
first-differences) and -ivreg2- can certainly take care of all 
estimation issues. Nevertheless, such IV estimators might have poor
finite sample performances.
 
Useful readings are the first part of the Arellano-Honore (2001) 
chapter or the Arellano (2003) manual.

Arellano, M. 2003. Panel Data Econometrics. Oxford: Oxford 
University Press

Arellano, M. and B. Honorè, 2001. Panel Data Models: 
Some Recent Developments. In Handbook of Econometrics vol. 5 
ed. J.J. Heckman and E. Leamer. Amsterdam: Elsevier

Giovanni 


> 
> 
> 
> 
> 
> > Roberto Zanola <zanola@sp.unipmn.it>:
> > 
> > > Dear all,
> > > we need to estimate a dynamic short panel (T=6 and N=20). Two
> possibilities:
> > > (1) lsdvc
> > 
> > This is implemented in Stata by the user written code -xtlsdvc-. 
> > It behaves relatively better than IV-GMM estimators in small panel
> > data-sets, in terms of both bias and root mean squared error,
> > but needs strictly exogeneity of regressors and neither 
> > heteroskedasticity or serial correlation of disturbances.
> >  
> > > (2) ivreg2 with dummies
> > 
> > I'm not clear what estimator Roberto has in mind in this
> > case. -ivreg2- is a flexible routine that can implement many 
> > IV estimators and tests, and clearly not all of them are appropriate
> > methods for dynamic models. A simple N-consistent estimator for
> > dynamic panel data models that can be supported by -ivreg2- is that 
> > developed by Anderson and Hsiao (1982). It is carried out by taking 
> > variables in first-differences and using the dependent variable 
> > lagged two times, y(t-2), as an instrument for Dy(t-1). One can 
> > also deal with endogenous x's in this case, provided valid instruments 
> > are available. However, Monte Carlo evidence demonstrates that the AH 
> > estimator, although virtually unbiased, is rather imprecise in 
> > small samples (very large root mean squared error).
> > 
> > References
> > Anderson, T. W. and C. Hsiao. 1982. Formulation and Estimation 
> > of Dynamic Models Using Panel Data. Journal of Econometrics 
> > 18: 570–606
> > 
> > Roberto may also find it useful my paper on small panel data-sets, 
> > downloadable from
> > 
> > http://ideas.repec.org/p/cri/cespri/wp165.html
> > 
> > 
> > Giovanni
> > (author of -xtlsdvc-) 
> > 
> > > 
> > > *
> > > *   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/
> > > 
> > 
> > 
> > -- 
> > Giovanni S.F. Bruno
> > http://ideas.repec.org/e/pbr136.html
> > Istituto di Economia Politica, Università Bocconi
> > Via U. Gobbi, 5, 20136 Milano
> > Italy
> > tel. + 02 5836 5411
> > fax. + 02 5836 5438
> > *
> > *   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/
> 


-- 
Giovanni S.F. Bruno
http://ideas.repec.org/e/pbr136.html
Istituto di Economia Politica, Università Bocconi
Via U. Gobbi, 5, 20136 Milano
Italy
tel. + 02 5836 5411
fax. + 02 5836 5438
*
*   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