st: RE: xtivreg2

["Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>]

Ikuho,

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Ikuho Kochi
> Sent: 02 July 2012 21:58
> To: statalist@hsphsun2.harvard.edu
> Subject: st: xtivreg2
> 
> Dear fellow econometricians/statisticians
>  
> I have a question regarding the "xtivreg2" command and would 
> appreciate very much if somebody can help me to clarify some issues.
> 
> Since I have to estimate HAC fixed models, I am using the 
> xtivreg2 command.  
> However, the xtivreg2 command does not estimate a constant 
> term in the fixed effect model (as described in the command 
> description), and I am wondering why it is programmed in this 
> way.  My understanding is that the fixed effect model usually 
> includes a constant term

Actually, that's not the case.  Strictly speaking, a fixed effects model
cannot provide an estimate of the constant term.  The reason is that a
"pure" fixed effects model uses *only* within variation to obtain
estimates of the coefficients.  It can't provide estimates of
coefficients on anything that does not vary within panels, and that
includes the constant term.

> (when I run the same model with 
> xtivreg command, the constant term appears, so it is not the 
> data problem),

This is because official -xtivreg-, and for that matter official
-xtreg-, use cross-sectional variation to obtain estimates of the
constant term.  Rather than go into the gory details, I'll just say
"it's in the manual".  Which it is.

Personally, I always found it odd that these commands will report an
estimated constant term that relies on cross-sectional variation even
when the -fe- option is specified.  This is non-standard in econometrics
(my home territory), but maybe it's standard in other areas - I don't
know.  But I'm the author of -xtivreg2-, and I programmed it to behave
more in accord with the conventions in my own field of work.

HTH,
Mark

> or am I missing something?  I wonder if the 
> user of this command is expected to include an extra variable 
> with constant values in the fixed effect model so that the 
> constant term will be estimated in fixed model estimation? 
>  
> I appreciate for your time and input...
> 
> Ikuho Kochi
> *
> *   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/
> 


-- 
Heriot-Watt University is the Sunday Times
Scottish University of the Year 2011-2012

Heriot-Watt University is a Scottish charity
registered under charity number SC000278.


*
*   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/