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

st: Re: VAR on panel data


From   Christopher F Baum <baum@bc.edu>
To   statalist@hsphsun2.harvard.edu
Subject   st: Re: VAR on panel data
Date   Wed, 1 Sep 2004 07:09:03 -0400

On Sep 1, 2004, at 2:33 AM, John wrote:


I'm interested in fitting a var model with some exogenous variables on panel
data for N countries.

I figured an easy way to do this would be
- -var y x1 x2, exog(country1-countryN other) noc

where "country" are indicator variables giving a country-specific intercept,
and "other" are other exogenous variables.

Are there any problems with such an approach? Does this approach assume that
the country effects are Fixed Effects? Is this a problem?
A VAR is a set of OLS regressions of several Y variables on a set of lags of all of the Ys plus, optionally, some Xs. You seem to have three dependent variables (Y, x1, x2). By default the var command will regress each of the dep vars on two lags of all of the dep vars. If the data are stacked in "long" format, so that you have used a panel tsset to define that these are panel data, var will not run at all. If you don't use tsset, var will treat the first observation on the second country as the T+1st observation of the first country, and so on when it takes lags. This will make no sense at all.

So no, I don't see that VAR is what you're looking for. Why not just fit a dynamic panel data model (xtabond or xtabond2)? That would appear to be closer to the specificatiion you are considering.

Kit

*
* 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