AW: st: AW: heteroscedasticity in panel data (Huber-White robust standard errors)

[P K <statistics_2009@yahoo.de>]

Hi Martin,

yes, the std errors for the two alternatives are the same. However, the differ from the version without robust std errors (which I didn't include in my questions below).

ie.

1) fe option only => significant results for some variables of interest
2) fe i(number) robust => insignificant results
3) fe i(number) robust cluster(number) => insignificant results


So I was wondering about what to report in my paper. I guess the robust std errors are the better option?
I tried bpagan for my data as well (as estat hettest is more suited for the cross-sectional case), but it didn't work.
So I thought using robust std. errors is a good alternative, but now I am not sure what to conclude from the output. 
I appreciate your help on this.

Thanks
Pat




----- Ursprüngliche Mail ----
Von: Martin Weiss <martin.weiss1@gmx.de>
An: statalist@hsphsun2.harvard.edu
Gesendet: Mittwoch, den 17. Juni 2009, 17:41:40 Uhr
Betreff: st: AW: heteroscedasticity in panel data  (Huber-White robust standard errors)


<> 

Don`t you use -robust- in both cases? So I would bet that the standard errors are the same across these two alternatives?

*************
//     Setup
webuse nlswork, clear

gen age2 = age*age
gen ttl_exp2 = ttl_exp*ttl_exp
gen tenure2 = tenure*tenure
gen byte black = race==2

xtset idcode

xtreg ln_w grade age* ttl_exp* /* 
*/  tenure* black not_smsa south, fe
est store fe

xtreg ln_w grade age* ttl_exp* /* 
*/  tenure* black not_smsa south, /* 
*/ fe robust
  est store ferob

  xtreg ln_w grade age* ttl_exp* /* 
*/  tenure* black not_smsa south, /* 
*/ fe robust cluster(idcode)
  est store ferobclu
  
  estimates table fe /* 
*/ ferob ferobclu, /* 
*/ se style(oneline)
*************

In either case, simply visually comparing the errors to conclude that there is heteroscedasticity is not a scientific method. Note, for instance, the  command -estat hettest- for the cross-sectional case...


HTH
Martin

-----Ursprüngliche Nachricht-----
Von: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] Im Auftrag von P K
Gesendet: Mittwoch, 17. Juni 2009 23:29
An: statalist@hsphsun2.harvard.edu
Betreff: st: heteroscedasticity in panel data (Huber-White robust standard errors)

Hi,

I tested my fixed effects panel model for heteroscedasticity using the Huber-White robust standard errors. 
The commands used are:

xtreg AverChangeROEadj2 strategy_01 duration_stability_t1 ceo_change ceo_int_t1 tmt_turnover_t1 sizelnempl age prior_slack_avail2 ROEadj_1 external_change_dummy lgcount_strategy performance_crisis_ROEadj strategy_01_stability_t1 strategy_01_ceo_changet1 stab_ceo_t1 stability_ceoint_strategy, fe i(number) robust

xtreg AverChangeROEadj2 strategy_01 duration_stability_t1 ceo_change ceo_int_t1 tmt_turnover_t1 sizelnempl age prior_slack_avail2 ROEadj_1 external_change_dummy lgcount_strategy performance_crisis_ROEadj strategy_01_stability_t1 strategy_01_ceo_changet1 stab_ceo_t1 stability_ceoint_strategy, fe i(number) robust cluster(number)


Results with robust std errors are different from the results without robust std errors, implying that there is heteroscedasticity.

Any hints on how to deal with this, i.e. shall I report the results with robust std errors?

Thanks,
Pat



      

*
*   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/



      

*
*   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/