st: Re: Hadrilm test

[Kit Baum <baum@bc.edu>]

The first panel, in which you consider the level of lngdp, provides overwhelming evidence against the stated null hypothesis of stationarity.

The second panel, in which you consider differences in lngdp (roughly, GDP growth rates), provides mixed evidence against the stated null. If you assume that the errors are homoskedastic (Homo), you do not reject. If you assume they are heteroskedastic or exhibit serial dependence, you do not reject using the mu test, but reject using the tau test. The mu test assumes, under H0, that the series are stationary around a mean; the tau test assumes they are stationary around a linear trend, analogous to the assumptions made in a Dickey-Fuller test.

So there is clear evidence against stationarity for at least some of the series in levels, and mixed evidence against stationarity of _all_ series for the series in first differences. Keep in mind that like any panel unit root test, the assumption that _all_ series are stationary is strong, and can be violated if even one of the series in differences exhibit nonstationary behaviour.

Kit 

On Mar 28, 2010, at 2:46 PM, asma esseghir wrote:

> Good morning Mr Baum,
> 
> I have a question about the Hadrilm test of statinarity in Panel data.
> 
> With Stata I had the result bellow, but I can't conclude if there's stationnarity or no. and how to correct it.
> Should I use dependent variable only?
> And what is the diffrence betwin "hadrilm lngdp" and "hadrilm D.lngdp"
> 
> 
> hadrilm lngdp
> Hadri (2000) panel unit root test for lngdp
> with 26 observations on 7 cross-sectional units
> -----------------------------------------------------------
>  eps       Z(mu)    P-value      Z(tau)    P-value
> -----------------------------------------------------------
> Homo      38.173     0.0000      19.703     0.0000
> Hetero    35.993     0.0000      12.917     0.0000
> SerDep     5.641     0.0000       3.454     0.0003
> -----------------------------------------------------------
> H0: all 7 timeseries in the panel are stationary processes
> Homo: homoskedastic disturbances across units
> Hetero: heteroskedastic disturbances across units
> SerDep: controlling for serial dependence in errors (lag trunc = 6)
> .
>  
>  hadrilm D.lngdp
> 
> Hadri (2000) panel unit root test for D.lngdp
> with 25 observations on 7 cross-sectional units
> -----------------------------------------------------------
>  eps       Z(mu)    P-value      Z(tau)    P-value
> -----------------------------------------------------------
> Homo       0.256     0.3991       1.053     0.1461
> 
> Hetero     0.285     0.3878       1.957     0.0252
> 
> SerDep     0.415     0.3390       3.148     0.0008
> -----------------------------------------------------------
> H0: all 7 timeseries in the panel are stationary processes
> Homo: homoskedastic disturbances across units
> Hetero: heteroskedastic disturbances across units
> SerDep: controlling for serial dependence in errors (lag trunc = 6)
> .
> Yours sincerely
> ESSEGHIR ASMA
> 
> Get your new Email address! 
> Grab the Email name you've always wanted before someone else does!


Kit Baum   |   Boston College Economics & DIW Berlin   |   http://ideas.repec.org/e/pba1.html
                              An Introduction to Stata Programming  |   http://www.stata-press.com/books/isp.html
   An Introduction to Modern Econometrics Using Stata  |   http://www.stata-press.com/books/imeus.html


*
*   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/