Re: st: Re: Using time invariant variable as instrument in fixed effects?

[Lisa Rickles <lrickles@sas.upenn.edu>]

Conor,

I appreciate your insight, but I am not sure I fully understand.  I
was interested in using parent education because it should be
correlated with achievement at every point in time (and therefore with
lagged achievement).

Using fixed effects, I am already controlling for unobserved
time-invariant factors for each child.  If I understand correctly, the
exogeneity problem arises only if parent education level is correlated
with the idiosyncratic error term.  I don't think parent education
level would be correlated with the time-varying error term.

Lisa

On Sat, Apr 9, 2011 at 9:02 PM, Conor Hughes <cbhughes@uchicago.edu> wrote:
> Hi Lisa,
> Maybe I'm missing something, but if you're trying to estimate math
> achievement, would the parents' education level be a valid instrument?
>  It seems pretty reasonable on first look to imagine that the parents'
> education level would be correlated with their child's math
> achievement, however you're defining that, so the instrument wouldn't
> be exogenous.
>
> I know this doesn't really answer your question, but hopefully it's
> helpful in one way or another.
>
> - Conor
>
> On Sat, Apr 9, 2011 at 1:27 PM, Lisa Rickles <lrickles@sas.upenn.edu> wrote:
>>
>> > Thank you so much for the suggestion.
>> >
>> > I looked at xthtaylor, but I am not sure if it would work because I
>> > have an explanatory variable that is correlated with both the
>> > individual time-constant unobserved effects and the idiosyncratic
>> > error.
>> >
>> > In particular, I am trying to estimate math achievement, and I include
>> > a lagged dependent variable (lagged math score) in the regression.  I
>> > am not sure exactly how xthtaylor works, but it seems that the lagged
>> > score would be correlated with the idiosyncratic error, as well as the
>> > unobserved ability.  (I would have used a second or third lag as in
>> > instrument instead with xtivreg, but I do not have enough prior scores
>> > in the data).  Therefore I was hoping to use parent education level
>> > (time constant) as an instrument for the lagged test score.
>> >
>> > Is there any way I can use the time-constant parent education as an
>> > instrument for the time varying test score, while still controlling
>> > for individual effects?
>> >
>> > Thanks,
>> > Lisa
>> >
>> > On Fri, Apr 8, 2011 at 8:06 PM, Fernando Rios Avila <f.rios.a@gmail.com> wrote:
>> >> By default I don’t think  you can do such a thing.
>> >> I think, however, that you could check on xthtaylor. Which could to handle
>> >> that issue for invariant instruments.
>> >>
>> >> -----Original Message-----
>> >> From: owner-statalist@hsphsun2.harvard.edu
>> >> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Lisa Rickles
>> >> Sent: Friday, April 08, 2011 8:01 PM
>> >> To: statalist@hsphsun2.harvard.edu
>> >> Subject: st: Using time invariant variable as instrument in fixed effects?
>> >>
>> >> Dear Statalist,
>> >>
>> >> Is it possible for me to use a time invariant variable as an instrument in a
>> >> fixed effects regression?  I am currently using xtivreg, and it is demeaning
>> >> the instrument I want to use and causing the endogenous variable to drop out
>> >> of the regression.
>> >>
>> >> Thanks,
>> >> Lisa
>> >>
>> >> --
>> >> Lisa Rickles
>> >> College of Arts and Sciences
>> >> University of Pennsylvania, Class of 2011 lrickles@sas.upenn.edu
>> >>
>> >> *
>> >> *   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/
>> >>
>> >>
>> >
>> >
>> >
>> > --
>> > Lisa Rickles
>> > College of Arts and Sciences
>> > University of Pennsylvania, Class of 2011
>> > lrickles@sas.upenn.edu
>> > 201-274-9326
>> >
>>
>>
>>
>> --
>> Lisa Rickles
>> College of Arts and Sciences
>> University of Pennsylvania, Class of 2011
>> lrickles@sas.upenn.edu
>> 201-274-9326
>>
>> *
>> *   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/
>
>
>
> --
> Conor Hughes
> Mathematics and Economics
> University of Chicago 2011
>
> *
> *   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/
>

-- 
Lisa Rickles
College of Arts and Sciences
University of Pennsylvania, Class of 2011
lrickles@sas.upenn.edu
201-274-9326

*
*   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/