Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
st: Fixed Effects Model for comparison between groups
From
Christian Schroetel <[email protected]>
To
"[email protected]" <[email protected]>
Subject
st: Fixed Effects Model for comparison between groups
Date
Sun, 15 Sep 2013 15:21:13 +0200
Hey,
I feel like you might be annoyed by those beginner-in-statistics but
I'm still trying it as I got the feeling you were a really trustable
source. So, here goes my issue (some might already know my model):
I'm trying to get to know drivers of firm growth. Therefore, I got
data for 3105 firms from 16 years, unbalanced (between 40k and 10k
observations per variable). I'm currently using 12 potential
explanatory variables. I clustered my data by firm and year using
tsset. Furthermore, I evaluated whether Random- or Fixed-Effects-Model
was the right one running a Hausman test. It said, Fixed-Effects model
was the proper one. I also wanted to try the Chamberlain test for the
validity of the restrictions of the reduced form model as suggested by
Baltagi (2005) but couldn't find a way to do that in Stata. So I just
did the FE-Model.
What I'm wondering about is the following: I get R-squared (within) of
around 0.25 to 0.3, R-squared (between) of around 0.02 and R-sq
(overall) of around 0.03. Does that mean the model is only appropriate
for explaining the variance of firm growth within each firm and not
between the firms? I mean, actually I wanted to conclude things like
"firm A grows at higher rates than firm B because it has higher
research & development intensity" but it seems like inferences like
that aren't really possible with that model. Is that right? I also
found the OLS, Random Effects and Between estimations all delivered
better results considering the R-squared (and what I expected the
resulst to be ;-)). Only the FE model has completely different results
what I suspected to be due to not considering the variation between
the firms and due to some explanatory variables not vary enough within
one firm. If so, what could I do about it? Is it possible clustering
by firms is too narrow? Or is the FE model just not suitable for such
inferences?
Thanks for any help.
Best regards, Christian
Baltagi (2005): Econometric Analysis of Panel Data
*
* 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/