Durbin-Wu Hausman test

[Jo Smith]

Hi
Any help please about the Stata code to run Durbin-Wu Hausman for panel data?

Thanks

[Carlo Lazzaro]

Jo:
type -search Durbin-Wu Hausman-

[Jo Smith]

Carlo, thanks
Can I use ivregress 2sls command?

[Marcos Almeida]

Hello, Jo,

As Carlo precisely stated, the commands "help" and "search" always give you excellent information.

With - ivregress - you are supposed to use another command: - estat endogenous -. so as to perform the Durbin-Wu Hausman test.

Please, also type: - help ivregress postestimation - . There you´ll find further specifications,

Best,

Marcos

[Carlo Lazzaro]

Jo:
questions like the one you posted are difficult to reply (yes or no are really unhelpful in this instance) unless you provide the list with more details.

[Jo Smith]

Thanks for that.

[Jo Smith]

I tried what you advised,
However, after estat endogenous , I find some unusual values (dots) - see below-

ivregress 2sls naf lnauditfees

Instrumental variables (2SLS) regression Number of obs = 690
Wald chi2(1) = 0.80
Prob > chi2 = 0.3704
R-squared = 0.0012
Root MSE = .30365

------------------------------------------------------------------------------
naf | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lnauditfees | -.0147123 .0164263 -0.90 0.370 -.0469073 .0174826
_cons | .2389657 .1523581 1.57 0.117 -.0596507 .5375821
------------------------------------------------------------------------------
(no endogenous regressors)

. estat endogenous

Tests of endogeneity
Ho: variables are exogenous

Durbin (score) chi2(0) = 0 (p = .)
Wu-Hausman F(0,688) = . (p = .)


Is something wrong here?

[Carlo Lazzaro]

Jo:
are you sure that you have instrumented the endogenous regressors properly?
At its facevalue, the slip of your Stata session shows that you have not chosen any instrument variable for your endogenous regressors. It is likely that you have simply re-run your model by replacing -reg- with -ivreg-. It this were the case, your result is not puzzling at all: Stata simply reports the same results calculated with -reg-,

[Eric de Souza]

is something wrong here?

Yes. To elaborate on Carlo's reply, look carefully at your output. It says (no endogenous regressors). So how can you test for endogeneity? By the way, the null hypothesis is that the regressor is exogenous.Read carefully the help pages for ivregress and ivregress post-estimation; and if necessary an appropriate textbook. Instrumental variable issues are very easy to mes up.