Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: xtabond and OLS for separate years


From   "R.E. De Hoyos" <redeho2@hotmail.com>
To   <statalist@hsphsun2.harvard.edu>
Subject   Re: st: xtabond and OLS for separate years
Date   Wed, 5 May 2004 11:42:07 +0100

Kate,

It seems that--as you pointed out--there is not enough within variation in
your data. That's why both the fixed effects and first-differences
estimators are barely significant and might show counter-intuitive signs.
The Arellano-Bond estimator tries to exploit the time (or within) variation
in the data correcting for autocorrelation. Therefore if you have a panel
formed by--let's say, countries and years, most of the variation will come
from differences between countries hence the Arellano-Bond estimator would
probably yield poor results.

Rafa

----- Original Message -----
From: "Kate Ivanova" <kivanova@usc.edu>
To: <statalist@hsphsun2.harvard.edu>
Sent: Wednesday, May 05, 2004 8:48 AM
Subject: RE: st: xtabond and OLS for separate years


> Thank you very much again, Mark and Antonio. I really appreciate all your
> comments.
>
> Now I could compare my OLS in cross section to between estimator, fixed
> effects, first differences [thanks to your help] and random effects. My
> income and income squared are significant in all OLS cross sections for 10
> years. They are also significant in the between-effects, random-effects
and
> fixed effects models. However, the signs on all my coefficients are wrong
in
> the fixed effects estimation. In the random effects model, the sign of the
> coefficient on my third variable, corruption, is wrong (though
significant).
> So I am basically left with the between estimator as it seems that I do
not
> have enough "within" variation in my explanatory variables. Does that mean
> that I should not use xtabond or is it possible to get round this problem?
>
> As for the first-differenced results, only two of the variables are barely
> significant (at the 10% level) but the signs are wrong for both of them.
>
> Please let me know if you have any suggestions.
>
> Kate
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Antonio
Rodrigues
> Andres
> Sent: Tuesday, May 04, 2004 5:12 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: RE: st: xtabond and OLS for separate years
>
> xtivreg using as instruments the same regressors gives the same result
> that doing it by hand
>
> Yes, you get the same results.
>
>
>  *FIRST DIFFERENCES
> . xtivreg lsrt ($xvars=income unempl), fd
>
> First-differenced IV regression                 Number of obs      =
>   276
> Group variable: country                         Number of groups   =
>    21
>
> R-sq:  within  = 0.0096                         Obs per group: min =
>     3
>        between = 0.8987                                        avg =
>  16.9
>        overall = 0.3945                                        max =
>    37
>
>                                                 chi2(2)            =
>  4.12
> corr(u_i, Xb)  = -0.5904                        Prob > chi2        =
> 0.1277
>
> --------------------------------------------------------------------------
--
> --
> d.lsrt       |      Coef.   Std. Err.      z    P>|z|     [95% Conf.
> Interval]
> -------------+------------------------------------------------------------
--
> --
> income       |
>           D1 |  -.0176139   .0088856    -1.98   0.047    -.0350292
> -.0001985
> unempl       |
>           D1 |  -.0041495   .0044982    -0.92   0.356    -.0129659
> .0046669
> _cons        |    .005815    .005735     1.01   0.311    -.0054254
> .0170553
> -------------+------------------------------------------------------------
--
> --
>      sigma_u |  .42097985
>      sigma_e |  .05951118
>          rho |  .98040789   (fraction of variance due to u_i)
> --------------------------------------------------------------------------
--
> --
> Instrumented:   income unempl
> Instruments:     income unempl
>
> . regress dlsrt dincome dunempl
>
>       Source |       SS       df       MS              Number of obs =
>   276
> -------------+------------------------------           F(  2,   273) =
>  2.06
>        Model |  .014576629     2  .007288314           Prob > F      =
> 0.1297
>     Residual |  .966851358   273   .00354158           R-squared     =
> 0.0149
> -------------+------------------------------           Adj R-squared =
> 0.0076
>        Total |  .981427987   275  .003568829           Root MSE      =
> .05951
>
> --------------------------------------------------------------------------
--
> --
>        dlsrt |      Coef.   Std. Err.      t    P>|t|     [95% Conf.
> Interval]
> -------------+------------------------------------------------------------
--
> --
>      dincome |  -.0176139   .0088856    -1.98   0.048    -.0351068
> -.0001209
>      dunempl |  -.0041495   .0044982    -0.92   0.357    -.0130051
> .0047061
>        _cons |    .005815    .005735     1.01   0.312    -.0054755
> .0171054
> --------------------------------------------------------------------------
--
> --
>
>
>
> >>> M.E.Schaffer@hw.ac.uk 05/05/04 12:27 AM >>>
> Kate,
>
> Quoting Kate Ivanova <kivanova@usc.edu>:
>
> > Hi Mark,
> >
> > I tried to run xtreg, fd (first-differencing without instrumenting)
> > as you
> > suggested but I could not find this option in the xtreg command.
> > There is
> > xtivreg, fd but then this is for estimation with instrumental
> > variables. Is
> > there any other way to estimate a first-differenced model without
> > instrumenting?
>
> Funny, I thought it would be there.
>
> I suppose you either have to try to trick xtivreg into running an
> uninstrumented equation by specifying the same instruments as regressors
> -
> and it might be too clever to be fooled - or you have to first
> difference
> by hand.
>
> --Mark
>
> > Thanks!
> >
> > Kate
> >
> > -----Original Message-----
> > From: owner-statalist@hsphsun2.harvard.edu
> > [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Kate
> > Ivanova
> > Sent: Monday, May 03, 2004 10:02 AM
> > To: statalist@hsphsun2.harvard.edu
> > Subject: RE: st: xtabond and OLS for separate years
> >
> > Mark,
> >
> > Thank you very much for your suggestions. They are very helpful.
> > Yes, I did
> > mean OLS in cross-section so I'll now compare it to the estimators
> > you
> > specified below. I'll get back again when I have the results. Thanks
> > a lot!
> >
> > Kate
> >
> > -----Original Message-----
> > From: owner-statalist@hsphsun2.harvard.edu
> > [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Mark
> > Schaffer
> > Sent: Monday, May 03, 2004 4:09 AM
> > To: statalist@hsphsun2.harvard.edu
> > Subject: Re: st: xtabond and OLS for separate years
> >
> > Kate,
> >
> > Date sent:      Sat, 01 May 2004 17:59:00 -0700
> > From:           Kate Ivanova <kivanova@usc.edu>
> > Subject:        st: xtabond and OLS for separate years
> > To:             statalist@hsphsun2.harvard.edu
> > Send reply to:  statalist@hsphsun2.harvard.edu
> >
> > > Hi,
> > >
> > > I am confused by the results I get using xtabond. I have a panel
> > of 116
> > > countries over 10 years and when I run separate regressions for
> > each year
> > > using OLS, my variables (income and income squared) are highly
> > significant
> > > (at a 0.001 level). But when I run xtabond with one lag of the
> > dependent
> > > variable, they are not significant at all. I wonder why I have
> > such a
> > > difference between the results. Any help, any ideas would be
> > greatly
> > > appreciated.
> >
> > It's hard to tell exactly what's going on from the info you've
> > provided, but there are at least two possibilities:
> >
> > 1. You are comparing OLS in cross-section (yes?) and xtabond, which
> >
> > you can think of as a first-difference estimator with instrumenting.
> >
> > Each of your cross-sections uses the cross-sectional variation
> > across
> > 116 countries in any year; xtabond using only the time-series
> > variation within countries.
> >
> > A better comparison would be to leave the instrumenting out of it
> > for
> > the moment, and compare:
> >
> > OLS period-by-period (uses only cross-sectional variation)
> > Between estimator (also uses only cross-sectional variation)
> > Fixed effects (uses only "within", i.e., time-series, variation)
> > First differences (also uses only "within" variation)
> > Random effects (uses both "within" and "between" variation)
> >
> > 2.  xtabond is an IV estimator, and the results you get will depend
> >
> > on the instrumenting.  You can compare the xtabond results with the
> >
> > results from first-differencing without instrumenting (xtreg, fd),
> >
> > for example, and see what happens.
> >
> > Hope this helps.
> >
> > --Mark
> >
> > >
> > > Kate
> > >
> > >
> >
> >
> > Prof. Mark E. Schaffer
> > Director
> > Centre for Economic Reform and Transformation
> > Department of Economics
> > School of Management & Languages
> > Heriot-Watt University, Edinburgh EH14 4AS  UK
> > 44-131-451-3494 direct
> > 44-131-451-3008 fax
> > 44-131-451-3485 CERT administrator
> > http://www.som.hw.ac.uk/cert
> > *
> > *   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/
> >
>
>
>
> Prof. Mark Schaffer
> Director, CERT
> Department of Economics
> School of Management & Languages
> Heriot-Watt University, Edinburgh EH14 4AS
> tel +44-131-451-3494 / fax +44-131-451-3008
> email: m.e.schaffer@hw.ac.uk
> web: http://www.sml.hw.ac.uk/ecomes
> ________________________________________________________________
>
> DISCLAIMER:
>
> This e-mail and any files transmitted with it are confidential
> and intended solely for the use of the individual or entity to
> whom it is addressed.  If you are not the intended recipient
> you are prohibited from using any of the information contained
> in this e-mail.  In such a case, please destroy all copies in
> your possession and notify the sender by reply e-mail.  Heriot
> Watt University does not accept liability or responsibility
> for changes made to this e-mail after it was sent, or for
> viruses transmitted through this e-mail.  Opinions, comments,
> conclusions and other information in this e-mail that do not
> relate to the official business of Heriot Watt University are
> not endorsed by it.
> ________________________________________________________________
> *
> *   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/
>
*
*   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