st: Validity of endogeneity test with weak instruments

["Lim Boon Leong" <lboonl@streamyx.com>]

This is a multipart message in MIME format.

Dear Statalists,

 

I have five suspected endogenous regressors in a regression equation and I
have also identified a set of excluded instruments for these endogenous
regressors.  I use xtivreg2 to run the regression analysis and the test
results are as shown below: 

 

 

Underidentification test (Kleibergen-Paap rk LM statistic):
3.641

                                                   Chi-sq(7) P-val =
0.8200

 

----------------------------------------------------------------------------
--

Weak identification test (Cragg-Donald Wald F statistic):
0.146

 

                         (Kleibergen-Paap rk Wald F statistic):
0.319

 

Stock-Yogo weak ID test critical values:                       <not
available>

 

----------------------------------------------------------------------------
--

Hansen J statistic (overidentification test of all instruments):
6.281

                                                   Chi-sq(6) P-val =
0.3925

 

The test results show that there is the problem of under identification and
the instruments are weak.  Nevertheless, the Hansen J statistic shows that
these instruments are valid/do not correlate with the error terms.

 

Next I carry out the endogeneity test to examine whether these five
suspected regressors are truly endogenous.  I use the endog(varlist_en)
option of xtivreg2 to carry out the test.  The test results are as shown
below:

 

-endog- option:

 

Endogeneity test of endogenous regressors:
5.755

                                                   Chi-sq(5) P-val =
0.3308

Regressors tested:    ownexr ownex2r ownex3r L.ldiv L.debt

----------------------------------------------------------------------------
--

Instrumented:         ownexr ownex2r ownex3r L.ldiv L.debt

Included instruments: ceodual ined remcom1 extblock1 L.lmktbk L.roa L.ret

                      L.ltotasset yr04 yr05 yr06

Excluded instruments: lrisk lsale lsale2 ks ks2 ys ik L2.ldiv L.roe
L.ltangible

                      L2.debt

Dropped collinear:    yr02 yr03 yr07

----------------------------------------------------------------------------
--

  

The test results show that these suspected regressors can actually be
treated as exogenous.  

 

My question is with the presence of weak instruments, are the above
endogeneity test results still valid and reliable?

 

Thank you in advance.

 

Best regards,

Boon Leong 

*
*   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/