Caution on rejecting the null at Hausman test?

[Minnie Rock]

Hey everyone,

I performed the following Hausman test to decide on the model I need to use.
The results show that I should use the FE model.

However, I also know that I should be cautious on rejecting the null (not using RE).

When can this be the case? What do you think I should keep in mind?

hausman fe re

---- Coefficients ----
| (b) (B) (b-B) sqrt(diag(V_b-V_B))
| fe re Difference S.E.
-------------+----------------------------------------------------------------
logMobilec~s | .478315 .6718178 -.1935028 .0086275
logExchRate | .3271507 .2607558 .0663949 1.018707
dUnemploym~l | -.0472854 -.0269096 -.0203758 .
dlogGDP | 1.150435 1.604068 -.4536325 .
------------------------------------------------------------------------------
b = consistent under Ho and Ha; obtained from xtreg
B = inconsistent under Ha, efficient under Ho; obtained from xtreg

Test: Ho: difference in coefficients not systematic

chi2(4) = (b-B)'[(V_b-V_B)^(-1)](b-B)
= 262.36
Prob>chi2 = 0.0000
(V_b-V_B is not positive definite)

[Carlo Lazzaro]

Minnie:
you can repeat -hausman- with the -sigmamore- option. You may also want to consider checking the reliability of -hausman- via the user-written command -xtoverid- (type -search xtoverid- from within Stata to install it).
If your sample if large enough, -hausman- can reject the null even for trivial difference between -fe- and -re- coefficients.