Statalist


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

st: RE: xtivreg2 and year dummies


From   "Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: xtivreg2 and year dummies
Date   Thu, 29 Nov 2007 15:22:55 -0000

Robert,

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Robert
> Sent: Wednesday, November 28, 2007 6:12 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: st: xtivreg2 and year dummies
> 
> Dear all,
> 
> I need some help (again...) to solve an applied econometrics problem. 
> In short: Starting from an unbalanced panel database (10 
> years, many variables), I made my xtivreg2, as the following:
> 
> * xtivreg2 h x (d=z), fe i(id) cl(id)
> 
> I was later suggested to add year dummies in my IV 
> regression. I tried this way:
> 
> - befofe making the xtivreg2..., I added the following strings:
> 
> * tabulate year, generate(dyear)
> * global xextra dyear1 dyear2 dyear3 dyear3 dyear4 dyear5 
> dyear6 dyear7
> dyear8 dyear9 dyear10
> 
> and then I re-estimated this regression:
> 
> * xtivreg2 h x $xextra (d=z), fe i(id) cl(id)
> 
> The results are not that good, in particular almost all 
> "dyear" dummies in the h (second stage) equation are visibly 
> not significant, while they are in the first stage. Besides, 
> the Sargan-Hansen is lower now!
> 
> My questions:
> 
> 1) Is it empirically correct the way I added the dummies in 
> the xtvreg2?

I'm not sure what you mean by "empirically correct", but I think this
way to add year dummies is fine.

> 2) Regardless of my specific interests, is it useful to add 
> year-dummies as exogenous in both equations ($xextra belong 
> to both first and second stage equation)?

That depends on you and your model.  If they're in the first stage but
not the second, you're saying that the year dummies are valid
instruments for whatever you're treating as an endogenous regressor.

> 3) Since some year dummies seem to be significant in the d (first
> stage) equation (while not in h regression), can I include 
> them in this, as if they were z-like instruments?

Same as above - it depends on the model.  If you think they're valid
instruments, then fine.  But you should have a reason for thinking that
year dummies are appropriate for the endogenous regressor d and not the
dependent variable h.

> 4) Is there a test of joint significance for these dummies (I 
> tried "test  dyear2 dyear1 dyear3 dyear4 dyear5 dyear6 dyear7 
> dyear8 dyear9 dyear10", but I were wrong...) to justify their 
> inclusion?

This syntax works for me when I use it on the abdata.dta dataset
provided with Stata.

Cheers,
Mark

> 
> 
> I really hope your kind help.
> 
> Thank you in advance for all.
> 
> Robert
> ______________________________________________________________
> __________
> More new features than ever.  Check out the new AOL Mail ! -
> http://o.aolcdn.com/cdn.webmail.aol.com/mailtour/aol/en-us/tex
t.htm?ncid=aolcmp00050000000003
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
> 

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   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