Re: st: Problems with Fixed Effects Regression

[Maarten Buis <maartenlbuis@gmail.com>]

On Mon, May 30, 2011 at 11:45 AM, Abubakr Saeed wrote:
> I'm performing a panel data analysis (400 firms, 12 years, and 6
> industries). I got two questions regarding my analysis:
> My LM test supports for random individual effects. After estimating
> the Hausman test for deciding between fixed or random effect
> regression, result was in favor of the fixed effect regression. I am
> using Industry dummies which are constant over time. While running FE
> my all industry dummies were dropped because of collinearity. What I
> do? Should I just neglect industry dummies or choose random effect?

When you estimate a fixed effects model you are controlling for
everything that is constant within a firm. This is why Stata is
dropping every variable that is constant within firms, these cannot
explain anything after controlling for the fixed effects. If industry
is just a control variable and you are not substantively interested in
it, than you can just use a fixed effects model and leave industry
out. You are not ignoring anything, as the fixed effects are already
controlling for all firm-constant variables, including industry. The
same is true for all other firm-constant variables, like your other
variable whether or not the firm was listed.

Things will obviously get more complicated when you are substantively
interested in these effects.

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany


http://www.maartenbuis.nl
--------------------------
*
*   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/