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

Re: st: Autocorrelation and heteroskedasticity in panel models


From   "Clive Nicholas" <[email protected]>
To   [email protected]
Subject   Re: st: Autocorrelation and heteroskedasticity in panel models
Date   Mon, 26 Jan 2004 20:38:12 -0000 (GMT)

David Greenberg wrote:

> Beck and Katz urge the use of panel-corrected standard errors to accompany
> OLS estimation for the special case when T is large and N is small. Enrico
> Pellizzoni's case is one when T is small and N is large. I don't believe
> that such methods as generalized least squares perform badly in that
> situation.

Thanks for your comments. I've flicked through the three Beck/Katz papers
on their critique of FGLS, and I reckon I'm reasonably sure of my ground.

In their 1995 paper, their Monte Carlo simulations show that FGLS performs
_worse_ in terms of overconfidence (i.e., the degree to which it
underestimates estimator variability) when T is small (pp 639-40).

Many savvy researchers recognised this and simply used FGLS to correct for
panel heteroscedasticity, hence Kmenta's CHTA (or PWLS, as B/K call it).
In their 1996 paper, they ran another MC simulation and concluded that
PWLS should only be used when T=>20 and where there is actually panel
heteroscedasticity in the errors that needs correcting (pp 20-3)!

By the 2001 paper, B/K summarise all this by making their earlier material
(slightly) easier to digest (mainly for thickards like me!), say that
OLS-PCSE performs better against both rivals in these circumstances and
conclude that "... the Parks-Kmenta [FGLS] estimator simply should not be
used" (Beck, 2001: 276).

I don't know about you, but reading this paper after their two others led
me to believe that one should take this injuction as _universal_ and not
partial. Of course, there may be many (F)GLS lovers out there who would
violently disagree with B/K on this, but there you are.

By the way, what -xt- would you suggest where T=6 and N=3456 (he asks
cheekily)? Neal Beck himself simply warned me against the use of PCSEs in
such a context. Anyone else is welcome to make a suggestion as well.

C.

REFERENCES:

Beck N and Katz JN (1995) "What to Do (and Not to Do) With Time-Series
Cross-Sectional Data", AMER POL SCI REV 89(3): 634-47.

Beck N and Katz JN (1996) "Nuisance vs. Substance: Specifying and
Estimating Time-Series Cross-Section Models", POL ANALYSIS 6(1): 1-34.

Beck N (2001) "Time-Series Cross-Section Data: What Have We Learned in the
Past Few Years", ANNU REV POL SCI 4: 271-93



CLIVE NICHOLAS        |t: 0(44)191 222 5969
Politics Building     |e: [email protected]
School of Geography,  |f: 0(44)870 126 2421
 Politics & Sociology |
University of         |
 Newcastle-upon-Tyne  |
Newcastle-upon-Tyne   |
NE1 7RU		      |
United Kingdom	      |http://www.ncl.ac.uk/geps
*
*   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 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index