Re: st: testing for cross sectional dependence in a dynamic model

[joales salbdralor <joalessakafiora@googlemail.com>]

Hi all,

I have a panel data set consisting of 11 countries and 7 years.
I want to test for  cross sectional dependence using Pesaran's test in
the conext of a dynamic linear fixed effects model.

In the paper "'Testing for Cross-sectional Dependence in Panel Data
Models'. published by stata journal it is mentioned in the concluding
remarks section that

"In dynamic panels,
Pesaranʼs test remains valid under FE/RE estimation (even if the
estimated parameters
are biased) and therefore it may be the preferred choice, since the
properties of the
remaining tests in dynamic panels are not yet known. On the other
hand, if common
time effects have been included in the dynamic panel (and the panel is
short), the test
by Sarafidis, Yamagata, and Robertson (2006) may be used."

 My model specification is the following
 Υ_it=a+bY_i(t-1)+cX_it++a_i+e_it_

where a_i  is a FE. the model does not contain time effects.
I can not understand 2 things
1) should  I apply Pesaran's test to the above dynamic model after
having estimated it with LSDV estimator. That is,

xtreg depvar laggeddepvar other regressors, fe cluster (id)
xtcsd, pesaran abs

However, when I perfom this test stata says that can not find some
regressors. Mysteriously some variables drop from the model. It this
because I use wrong approach in my code?. Does this have to do with
the fact that some regressors are correlated (corr coeff=0.4 or 0.52).
 Does this has to do with the fact that I have 11 countries and 7
years only? Also I use stata 10.
2) If I find cross sectional dependence how do I proceed in terms of
the estimation of the dynamic panel FE model. That is , can  I still
apply Blundel-Bover estimator ? Should I correct the standard errors
with another formula?


Thank you for your time


On 6/12/12, joales salbdralor <joalessakafiora@googlemail.com> wrote:
> Dear all,
>
>
> I have a panel data set consisting of 11 countries and 7 years.
> I want to test for  cross sectional dependence using Pesaran's test in
> the conext of a dynamic linear fixed effects model.
>
> In the paper "'Testing for Cross-sectional Dependence in Panel Data
> Models'. published by stata journal it is mentioned in the concluding
> remarks section that
>
> "In dynamic panels,
> Pesaranʼs test remains valid under FE/RE estimation (even if the
> estimated parameters
> are biased) and therefore it may be the preferred choice, since the
> properties of the
> remaining tests in dynamic panels are not yet known. On the other
> hand, if common
> time effects have been included in the dynamic panel (and the panel is
> short), the test
> by Sarafidis, Yamagata, and Robertson (2006) may be used."
>
>  My model specification is the following
>  Υ_it=a+bY_i(t-1)+cX_it++a_i+e_it_
>
> where a_i  is a FE. the model does not contain time effects.
> I can not understand 2 things
> 1) should  I apply Pesaran's test to the above dynamic model after
> having estimated it with LSDV estimator. That is,
>
> xtreg depvar laggeddepvar other regressors, fe cluster (id)
> xtcsd, pesaran abs
>
> However, when I perfom this test stata says that can not find some
> regressors. Mysteriously some variables drop from the model. It this
> because I use wrong approach in my code?. Does this have to do with
> the fact that some regressors are correlated (corr coeff=0.4 or 0.52).
>  Does this has to do with the fact that I have 11 countries and 7
> years only? Also I use stata 10.
> 2) If I find cross sectional dependence how do I proceed in terms of
> the estimation of the dynamic panel FE model. That is , can  I still
> apply Blundel-Bover estimator ? Should I correct the standard errors
> with another formula?
>
>
> Thank you for your time
>
> *
> *   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/