REGWUTEST Procedure

@RegWuTest performs a Wu (Wu-Hausman, Durbin-Wu-Hausman) specification test on a regression just estimated by instrumental variables. This needs to be preceded by a LINREG with the INSTRUMENTS option.

@REGWUTEST (no options or parameters)

Variables Defined

%CDSTAT	Test statistic (for LM version) (REAL)
%SIGNIF	Significance level of %CDSTAT (REAL)

Example

* Wooldridge, Introductory Econometrics, 4th edition

* Example 16.5 from pages 557-558

open data mroz.raw

data(format=free,org=columns) 1 753 inlf hours kidslt6 kidsge6 age educ $

wage repwage hushrs husage huseduc huswage faminc mtr motheduc fatheduc $

unem city exper nwifeinc lwage expersq

* 2SLS is done by using an INSTRUMENTS instruction to list the set of

* instruments, and a LINREG with the INSTRUMENTS option.

instruments constant educ age kidslt6 nwifeinc exper expersq

linreg(inst) hours

# constant lwage educ age kidslt6 nwifeinc

* Endogeneity test - done using the RegWuTest procedure

@RegWuTest

Sample Output

LWAGE is being instrumented out by EXPER and EXPERSQ. The null hypothesis in the Wu test is that the apparently endogenous variable is actually exogeneous and so doesn't need to be instrumented. Here, we strongly reject that. The procedure computes both an F and a chi-squared version of the test; these will usually produce almost identical significance levels.

Wu/Hausman Specification Test

F(1,421)= 43.66478 with Significance Level 0.00000000

Chi-Squared(1)= 40.219372 with Significance Level 0.00000000