Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Structural Break Type Test with Fixed Effects IV and GMM


From   Dominic Soon <dominic.soon.etc@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Structural Break Type Test with Fixed Effects IV and GMM
Date   Sun, 11 Apr 2010 22:12:26 +0800

Yes, my mistake!

I'm intending to use three approaches, from -xtreg- , -xtivreg- and
-xtabond- (i.e. dynamic panel)  Unfortunately, I don't have any other
instrumental variables, let alone instrumental variables which impact
S "differentially" in the early and the late period.  In these
circumstances, would the following make sense?  SxL refers to S*L.

xtreg Y L S SxL, fe
xtivreg Y L (S SxL = L.S L.SxL)  /* Assume that lagged S does not
affect shocks in current Y */
xtabond Y L , lags(1) endogenous(L SxL)

(Other covariates are omitted for ease of discussion)

My problem is that clearly the idea that lagged SxL is probably going
to be a very poor instrument for SxL close to the beginning of the
"late" period. For instance, if "late" refers to 2001, lagged SxL will
be zero.


On Sun, Apr 11, 2010 at 3:07 AM, Austin Nichols <austinnichols@gmail.com> wrote:
> Dominic Soon <dominic.soon.etc@gmail.com>
> -fe- is not an option for -xtabond-.  Are you thinking you have a
> dynamic panel model, or S is endogenous (i.e. you mean -xtivreg-)?  If
> the latter, what instruments are you using?  You probably want L and
> L*S included as regressors, and you then have 2 endog vars: S and L*S.
>  Then you want instruments that differentially affect S in the early
> and late periods.
>
> On Fri, Apr 9, 2010 at 11:34 PM, Dominic Soon
> <dominic.soon.etc@gmail.com> wrote:
>> Dear Statalisters,
>>
>> I am trying to estimate a regression of two variables output (Y) on
>> R&D capital stock (S), as well as some other variables (e.g. labour,
>> capital, so on and so forth).  For simplicity, let's say the model is:
>>
>> Y_it = a_i + beta * S_it + error term
>>
>> I am trying to see whether the coefficient beta is different between
>> an (assumed) "early" and "late" period.  I'm also attempting to run
>> the regression using both fixed effects and GMM.
>>
>> The question is - are there any issues with this methodology,
>> particularly when running a GMM estimation.  Suppose I ran something
>> like:
>>
>> xtabond Y S S_late, fe
>>
>> where S_late is equal to L times S, with L being a dummy variable that
>> is equal to 1 if t is later than my (assumed) breakpoint, can the
>> t-statistics be interpreted sensibly?
>>
>
> *
> *   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/
>

*
*   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   |   Site index