Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Two-Way FE Estimation


From   Maarten buis <maartenbuis@yahoo.co.uk>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Two-Way FE Estimation
Date   Fri, 15 Aug 2008 19:04:04 +0100 (BST)

--- Asgar Khademvatani <akhademv@ucalgary.ca> wrote:
> Thank you for your answer and time. But, I got the following problem
> in running your suggestion for implementing two-way FE:
> 
> As You agreed on that the following pseudo-code is a two-way FE 
> implementation:
> 
> xtreg  Y  d1 d2 d3 X,  fe
> where Y is dependent variable(s) and X is independent variable(s),
> and (d1 d2 d3) are industry dummies.
> 
> I run this model and got the following error:
> 
> independent variables are collinear with the panel variable ind
> r(198);
> 
> This shows that to prevent collinearity,  we do not need include 
> industry dummies (d2 d2 d3) , since -xtreg- itself takes care of time
> periods and industries defined as panel id. This means, we need to
> run the above pseudo-code without industry dummies. It will be
> implicitly accounts for a two-way FE. By the way, we can not add time
> dummy as well, since T is enough large and will lose degree of 
> freedoms.
> 
> I am wondering whether you are agree with my above discussion and 
> suggestion in implementing FE model?

I assumed that because you wanted to include year dummies your panel id
was industry, and that you knew enough about fixed effects to change
your panel id to year when you added industry dummies. You should have
told us at the beginning that your panel id included industry *and*
year. (Basically, this was the suggestion I gave you the second time)

A convenient way of thinking about fixed effect model (not of
estimating them) is that this model adds dummies for each level of
panel id, if the levels are industry, than you can add dummies for
year, so you are controlling for industry and year but not their
interaction. If the levels are industry and year, than that implies
dummies for year, industry, and their interactions. In the latter case
you obviously can not manually enter dummies for industry again.

You seem to focus a lot on loosing degrees of freedom when adding
dummies. You will loose that information regardless of whether you add
dummies or use -xtreg-. This is the very nature of fixed effects: you
only use variation within a panel id and ignore any variation between
panel id's. This necesarrily means that you loose a lot of information
when doing a fixed effects regression, regardless of whether you do it
by adding dummies or whether you do it using -xtreg-. The advantage of
fixed effects is that it allows you to relax some (but not all!) of the
assumptions concerning unobserved variables, but the price you pay is
loss of power. As any economist will tell you: There is no such thing
as free lunch.

-- Maarten

-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434

+31 20 5986715

http://home.fsw.vu.nl/m.buis/
-----------------------------------------

Send instant messages to your online friends http://uk.messenger.yahoo.com 
*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index