Re: st: RE: panel data fixed vs random

[Christopher Baum <kit.baum@bc.edu>]

<>
On Feb 4, 2013, at 2:33 AM, Pietro said:

> If i run the hausman test in the reverse order:
> 
> hausman random fix
> 
> i get the following output:
> 
> Note: the rank of the differenced variance matrix (2) does not equal the number of coefficients
>         being tested (4); be sure this is what you expect, or there may be problems computing
>         the test.  Examine the output of your estimators for anything unexpected and possibly
>         consider scaling your variables so that the coefficients are on a similar scale.
> 
>                  ---- Coefficients ----
>              |      (b)          (B)            (b-B)     sqrt(diag(V_b-V_B))
>              |     random        fix         Difference          S.E.
> - -------------+----------------------------------------------------------------
> Penetratio~2 |   -29177.79    -83456.95        54279.16        56513.77
> GastosSaud~a |   -4.087394     .6067533       -4.694147        2.920822
> MortHomicp~b |    65.12774      4.54167        60.58607        31.10263
> interinsfi~a |    .3677846      .238441        .1293436        .0192787
> - ------------------------------------------------------------------------------
>                            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

Garbage in, garbage out. You cannot arbitrarily reverse the order of arguments to -hausman-. As the footer says, you
have specified that the first set of estimates is always consistent, i.e. FE, not RE. You have to use fixed as the first argument.

The huge difference in point estimates suggests to me that RE is not acceptable. Furthermore, when you run the 
test properly, you get a negative Chi2 value, and the command tells you that you are not meeting the underlying assumptions.
In my experience this often signals that you are comparing two inconsistent estimators, i.e., that your FE model is
seriously misspecified, and is itself junk. I would work with the FE model, analyze its residuals, and consider whether
there are obvious problems of omitted variables, endogeneity, etc.

Kit

Kit Baum   |   Boston College Economics & DIW Berlin   |   http://ideas.repec.org/e/pba1.html
                             An Introduction to Stata Programming  |   http://www.stata-press.com/books/isp.html
  An Introduction to Modern Econometrics Using Stata  |   http://www.stata-press.com/books/imeus.html


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/