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

Re: st: Panel Cointegration

From   Ian Sue Wing <[email protected]>
To   [email protected]
Subject   Re: st: Panel Cointegration
Date   Wed, 11 Aug 2004 07:34:10 -0400

Are there any -ado files that address panel cointegration tests, such as
Pedroni, Levin & Lin, Harris & Tzavalis, Maddala & Kim? Or anything dealing
with DOLS(Dynamic OLS) or FMOLS (Fully Modified OLS) estimation?

Other Statalisters probably have much more detailed knowledge on this topic than I do, but here goes...

For the Pedroni heterogeneous panel tests, use RATS. Everything is set up for you there. The Kao homogeneous panel tests can be easily coded in a do file for the DF-rho and DF-rho-star cases. The other tests require computation of the nuisance parameters in the expressions for the test statistics---but Kao's monte carlo analysis finds that the DF-rho-star test out-performs the others most of the time. See the papers on Kao's website for details.

As far as estimation goes, you can perform DOLS easily in a panel setting with -xtreg, fe-. Just add forward and lagged differences of the covariates in systematic combinations, testing for critical lag and lead lengths after each regression using the Bayes Information Criterion. This test is available in -icomp-. This process may well be tedious, but again you can write a do file that loops over lags and leads, whose stopping criterion is based on the values returned by the -icomp- tests. I haven't found a way to implement FMOLS in Stata.

To check the results of your tests of cointegration in a heterogeneous vs. homogeneous panel, it may prove useful to compare the results of LSDV, DOLS and Swamy random coefficient (see -xtrchh2-) specifications of your estimating equation, evaluating the extent to which the parameters remain constant across the different specifications, especially if you include a time trend. If I remember correctly, a very high t-stat on the trend (10 or more) is often indicative of spurious regression problems in a time-series context.


Ian Sue Wing
Assistant Professor
Center for Energy & Environmental Studies
Department of Geography & Environment
Boston University
675 Commonwealth Ave. Rm. 141, Boston MA 02215
Phone: (617) 353-5741 * Fax: (617) 353-5986

* For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index