# st: Panel with low within variance and -xtfmb-

 From "Ivan Etzo" To Subject st: Panel with low within variance and -xtfmb- Date Thu, 4 Oct 2007 13:06:07 +0200 (ora legale Europa occidentale)

Hi,

I'd like to test a basic form of my model (panel T=7 N=380) with one regressor that is time invariant and two regressors that have low within variance with the following percentage

Y: 15%
X1: 0.6%
X2:0.6%
X3:0

hausman test rejects the RE model (using -xtreg-) so I cannot estimate the model with the time invariant regressor that for me it's very important.

I tried the -xtfmb- which is an implementation of the Fama and MacBeth (1973) two step procedure, I report the very clear description of the command from the help

-xtfmb- "is an implementation of the Fama and MacBeth (1973) two step procedure. The procedure is
as follows: In the first step, for each single time period a cross-sectional regression is
performed. Then, in the second step, the final coefficient estimates are obtained as the
average of the first step coefficient estimates."

The coefficient I obtain from this procedure are practically the same with the RE model.. I attached the results  below. I wonder whether the -xtfmb- produces inconsistent estimates as well.. any help would be very appreciated.

Ivan

Random-effects GLS regression Number of obs      = 2660
Group variable (i): region Number of groups   = 380

R-sq:  within = 0.0097 Obs per group: min = 7
between = 0.8153 avg = 7.0
overall = 0.7953 max = 7

Random effects u_i ~ Gaussian Wald chi2(3)       = 1665.29
corr(u_i, X) = 0 (assumed) Prob > chi2        = 0.0000

lnmig | Coef.   Std. Err. z    P>|z|     [95% Conf. Interval]

lndpop | .9560363   .0338689 28.23   0.000     .8896544 1.022418
lnopop | 1.023108   .0338689 30.21   0.000     .9567262 1.08949
lndist | -.3231335   .0587596 -5.50   0.000    -.4383002 -.2079668
_cons | -21.17337    .799259 -26.49   0.000    -22.73989 -19.60685

sigma_u | .68680964
sigma_e | .27499318
rho | .86183578   (fraction of variance due to u_i)

Fama-MacBeth (1973) Two-Step procedure Number of obs     = 2660
Num. time periods = 7
F(  3,     6)     = 11189.96
Prob > F          = 0.0000
avg. R-squared    = 0.7962

|            Fama-MacBeth
lnmig |      Coef.   Std. Err.      t P>|t|     [95% Conf. Interval]

lndpop |   .9552167   .0101939    93.70 0.000     .9302731 .9801604
lnopop |   1.017995   .0061794   164.74 0.000     1.002874 1.033115
lndist |  -.3232731   .0117134   -27.60 0.000    -.3519349 -.2946114
_cons |  -21.08686   .1716352  -122.86 0.000    -21.50683 -20.66688