Durbin-Wu-Hausman Test

[Saakshi]

Hi,

I need to perform Durbin-Wu-Hausman Test for an OLS estimate. The null hypothesis is "OLS estimators are consistent". Can you please help me getting some code on this or any idea.

thanks

[KrilleJ]

If you generally use the following formula for the DWH test:

TestStat=d' * inv(Est Asy Var [d]) * d

where d is the difference between the IV and LS estimators, you can use the following command to extract the components needed:

matrix MyOLSBeta=OLS_eq.@coef
matrix MyOLSCov=OLS_eq.@coefcov

matrix MyIVBeta=IV_eq.@coef
matrix MyIVCov=IV_eq.@coefcov

The distribution of the test statistic under the null then depends on an appropriate number of DoF.

/K

[Saakshi]

Thanks.

I need one more explanation. I used TSLS for an estimation and then followed regressor endogeneity test that is Durbin-Wu-Hausman. My regressor turns out to be endogenous as the null hypothesis is accepted. So far this means that the regressor is endogenous, so in this case can I go back and perform my OLS or I have to stick to my TSLS estimates only.

I am bit confused in this.

Please help.

[Krille]

If I recall correctly (you have to check) and read the test in the right way, a small observed value of the test statistic is an indication of OLS being "OK". The procedure should be described in detail in standard econometric books.