st: question about endogeneity test after 2SLS

[Deepankar Basu <dbasu@econs.umass.edu>]

Hi Statalist,

I am using STATA 11.1 on Windows 7. I estimated a model using "ivregress":

. ivregress 2sls D.lemp L.D.lrgva L.L.D.lrgva (D.lrgva = L.D.lrpcg  
L.D.lrfinv L.D.lrgva L.D.rgexp L.L.D.lrgva) if tin(2001q1,2007q4),  
vce(hac nwest)

After the estimation, I do the following specification test:

. estat endogenous

  Tests of endogeneity
  Ho: variables are exogenous

  HAC score chi2(1)               =  .253277  (p = 0.6148)
    (Prewhitening performed with 1 lag)

  HAC regression F(1,12)          =  3.21364  (p = 0.0983)
    (Based on Bartlett kernel with 15 lags)


The HAC score test indicates that the the null cannot be rejected  
(p-value is 0.6148); on the other hand, the regression-based test  
indicates that the null hypothesis can be rejected at the 10 percent  
level (p-value is 0.0983). Thus, the two tests give conflicting  
results regarding the endogeneity of the regressor (D.lrgva). What is  
the best practice in such situations of conflicting result?

Thanks in advance.

Deepankar Basu


*
*   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/