st: problems with Hausman Test

[Χρυσούλα Γιαννικοπούλου <chrygiann@gmail.com>]

Dear Statalisters,

I am facing a problem with Hausman test.

I examine if the coditional (absolute) and uncoditional convergence
hypothesis hold for NUTS IIΙ REGIONS within a country.

The Fixed Effects and Random Effects Methods of Estimation are used.
My regressions are:
Absolute:
xtreg dlngdp_pc lnlag_gdp_pc, fe cl(regions)
xtreg dlngdp_pc lnlag_gdp_pc, re cl(regions)

conditional:

xtreg  dlngdp_pc lnlag_gdp_pc lnbed_places, fe cl(regions)
xtreg dlngdp_pc lnlag_gdp_pc lnbed_places,re cl(regions)

my results are listed below:
absolute convergence: Δ ln⁡〖y_(i,t) 〗=γ+b ln⁡〖y_(i,t-1) 〗+u_(i,t)

                F.E.          R.E.
ln yi,t-1     -0.2735       -0.0576
                (-7.32)        (-3.79)
constant    2.5574        0.5476
                (7.35)          (3.89)


Conditional convergence : Δ ln⁡〖y_(i,t) 〗=γ_i+bln⁡〖y_(i,t-1)
〗+c(ln⁡〖〖TOUR〗_(i,t) 〗 )+u_(i,t)



                F.E.          R.E.
ln yi,t-1     -0.2867       -0.0576
                (-7.54)        (-3.65)
lnTOURi,t    2.5574        0.5476
                (7.35)          (3.31)

constant (γ)  2,1820      0,6389
                  (5,49)         (3,58)

SO, I would say whether I trust more the RE or the FE results.
I guess the RE results are more realistic that the very optimistic FE
results. However,
a Hausman test will probably suggest the FE model.

My main problem is that I cant find the right command about Hausman test.

After the above estimations I run the commands:

*****

estimates store dta_NUTSIII
hausman dta_NUTSIII

***********

But, I get the error message: "the two models need to be different"

Unfortunately, I can't find where is my mistake.

Could you suggest me any commands about Hausman Test in this case?

Thank you for consideration!

Chrysa

*
*   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/