Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: RE: Is xtivreg2 appropriate with a small dynamic panel?


From   DE SOUZA Eric <edesouza@coleurop.be>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: Is xtivreg2 appropriate with a small dynamic panel?
Date   Thu, 6 Aug 2009 18:45:29 +0200

To elaborate on Austin Nichols' reply, using sophisticated techniques
that rely on asymptotic theory is not worth it with your sample. Even
the Windmeiejer correction for small sample bias would not be very
useful in your case.

Eric de Souza
College of Europe
Brugge (Bruges), Belgium
http://www.coleurope.eu
 

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Steve Doogue
Sent: 06 August 2009 15:06
To: statalist@hsphsun2.harvard.edu
Subject: st: Is xtivreg2 appropriate with a small dynamic panel?

Hello all,

Grateful for your help on the following. I'm new to STATA and an
econometric rookie, so please bear with me.

I am working with macroeconomic data for which T = 17 and N = 4 (four
countries, over 1990-2006), and want to know whether the xtivreg2
command is appropriate for this panel size.

The model I am looking at will be dynamic with a lag of the dependent
variable on the right hand side.

I suspect endogeneity of one of my explanatory variables, so want to use
an instrumental variable approach. I don't want to use the Arellano-Bond
(AB) GMM estimator, which I understand is bias when T is
small: Judson and Owen (1996) for example find significant bias using
the AB estimator even when T=30. A natural candidate seems to be the
xtivreg2 (fixed effects) set of commands which (a) allows for
endogeneity to be addressed (via the IV) and (b) account for
country-specific effects. My question, then, is how well suited are the
xtivreg2 commands to a panel data of the dimensions I describe (T = 17,
N = 4)?

If I have interpreted the paper by Alvarez and Arrellano (2003)
correctly, the bias of the fixed effect estimator is of the order 1/T
(however, this is as N approaches infinitiy).

Many thanks for your help,

Steve.
(London)

References:
- Alvarez and Arrelano, 2003, The Time Series and Cross-section
Asymptotics of Dynamic Panel Data Estimators,  Econometrica, July 2003.
- Judson and Owen, 1996, Estimating Dynamic Panel Data Models: A
Practical Guide for Macroeconomists.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   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