Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: RE: st: predetermined variables with xtabond2
From
"Justina Fischer" <[email protected]>
To
[email protected]
Subject
Re: RE: st: predetermined variables with xtabond2
Date
Sat, 16 Jun 2012 19:16:41 +0200
Dear Soren,
the quote says that why x is an instrument for y, D.x is not a valid instrument for D.y.
I think you have to a) understand better the model (exogenous, endogenous factors) you will estimate better and b) know what GMM is doing, what kind of transformation it does to the regression model 'you typed in for Stata'.
Is the estimator for the original equation (a) calculated or does the programm estimate some transformation of it (that affects the validity of x, and hence the model you have to specify in (a))?
best
Justina
on: "Søren Møller-Larsson" <[email protected]>
> An: [email protected]
> Betreff: RE: st: predetermined variables with xtabond2
> Thanks for answering Justina. However, I am afraid that it does not solve
> my problem.
> I am estimating a system of both difference equations and levels equations
> that will need proper instruments.
>
> The standard treatment of a predetermined independent variable, x, is
> instrumenting it GMM-style with lag 1 and longer. However, the quote from the
> Stata help-file seems to describe some special case of instrumenting a
> predetermined variable, x, using ivstyle(x, equation(level)). Since these two
> methods of instrumenting predetermined variables are quite different, my
> question is, when is it preferable to use the ivstyle(x, eq(level)) option for
> instrumenting predetermined independent variables?
>
> quote from xtabond2.hlp
>
> "equation() is useful for proper handling of predetermined variables used
> as IV-style instruments in system GMM. For example, if x is
> predetermined, it is a valid instrument for the levels equation because it is assumed to
> be uncorrelated with the contemporaneous error term. However, x becomes
> endogenous in first-differences, so D.x is not a valid instrument for the
> transformed equation. ivstyle(x) would therefore be inappropriate. The use
> of x as an IV-style instrument in levels only could be specified by
> ivstyle(x, equation(level))."
>
> Thanks again
> Soren
>
> ----------------------------------------
> > Date: Fri, 15 Jun 2012 22:23:52 +0200
> > From: [email protected]
> > Subject: Re: RE: st: predetermined variables with xtabond2
> > To: [email protected]
> >
> > Hi
> >
> > not being an expert, you appear to have to decide whether you estimate
> first differences or not. First differences (al a Dy = da + db + e2) lead to
> endogeneity of otherwise exogenous (instrumental) variables if you
> estimated levels instead (y = a +b + + e1).
> >
> > I think this is what the quote says.
> >
> > Best
> >
> > Justina
> > -------- Original-Nachricht --------
> > > Datum: Fri, 15 Jun 2012 21:15:11 +0200
> > > Von: "Søren Møller-Larsson" <[email protected]>
> > > An: [email protected]
> > > Betreff: RE: st: predetermined variables with xtabond2
> >
> > > Dear all
> > >
> > > I have put this question forward ones before as seen below, but I
> still do
> > > not quite understand.
> > >
> > > Looking at the user-written Stata command -xtabond2- with the
> two-stage
> > > system gmm option. Considering treatment of a predetermined
> independent
> > > variable call it x. I have read the "How to Xtabond..." by Roodman
> (2009).
> > >
> > > Standard treatment is the gmm(x, laglimit(1 .)) option with lag 1 and
> > > longer.
> > >
> > > Now looking at the -xtabond2- help-file in Stata it reads the
> following:
> > >
> > > "equation() is useful for proper handling of predetermined variables
> used
> > > as IV-style instruments in system GMM. For example, if x is
> > > predetermined, it is a valid instrument for the levels equation since
> it is assumed to
> > > be uncorrelated with the contemporaneous error term. However, x
> becomes
> > > endogenous in first differences, so D.x is not a valid instrument for
> the
> > > transformed equation. ivstyle(x) would therefore be inappropriate.
> The use
> > > of x as an IV-style instrument in levels only could be specified by
> iv(x,
> > > eq(level))."
> > >
> > > So is this a special case for predetermined variables? Can anybody
> explain
> > > when to use the iv(x, eq(level)) for predetermined variables?
> > >
> > > Thank you and enjoy the weekend.
> > >
> > > Regards Soren
> > > Aarhus university
> > >
> > >
> > > ----------------------------------------
> > > > Date: Mon, 28 May 2012 09:30:53 -0400
> > > > Subject: Re: st: predetermined variables with xtabond2
> > > > From: [email protected]
> > > > To: [email protected]
> > > >
> > > > Dear Soren,
> > > > I believe that you have got it right. pp.124 of the paper contains
> the
> > > > relevant information to your question. The iv( . ) option is for
> > > > strictly exogenous regressors, while the gmm( . ) option is for
> > > > predtermined or suspected endogenous variables. Thus, e.g.
> > > >
> > > > " If w1 is strictly exogenous, w2 is predetermined but not strictly
> > > > exogenous, and w3 is endogenous, then
> > > > - xtabond2 y L.y w1 w2 w3 i.t, gmmstyle(L.y w2 L.w3) ivstyle(i.t w1)
> > > > twostep robust small orthogonal -
> > > > would fit the model with the standard choices of instruments—here
> with
> > > > two-step system GMM, Windmeijer-corrected standard errors,
> > > > small-sample adjustments, and orthogonal deviations." (pp. 127)
> > > >
> > > > The - lag - option is helpful for controlling the number of
> > > > instruments- another issue of significance that has also been
> > > > discussed by Roodman in his other paper (Roodman, D. M. 2009. A note
> > > > on the theme of too many instruments. Oxford Bulletin
> > > > of Economics and Statistics 71: 135–158).
> > > >
> > > > Overall, I believe that you are on track!
> > > >
> > > > Best wishes,
> > > > Suryadipta.
> > > >
> > > > On Sun, May 27, 2012 at 10:37 AM, Søren Møller-Larsson
> > > > <[email protected]> wrote:
> > > > > Dear Suryadipta
> > > > >
> > > > > Thanks you for your answer. Yes I did indeed. I am aware that the
> > > standard treatment of predetermined variables in system GMM is
> gmmstyle() and
> > > to use lags 1 and longer. Please correct me if I am wrong. In "how to
> > > xtabond2.." (2009) p. 124 it reads:
> > > > >
> > > > >
> > > > >
> > > > >
> > > > >
> > > > >
> > > > >
> > > > >
> > > > > "ivstyle() also generates one column per variable in
> > > System
> > > > > GMM, following (26). The patterns in (27) can be requested using
> the
> > > equation suboption, as in: iv(w1
> > > > > w2, eq(level)) and the compound iv(w1 w2, eq(diff)) iv(w1 w2,
> > > eq(level))."
> > > > > So is it just two different methods to treat predetermined
> variables?
> > > If so is there a reason to use one over the other
> > > > > Thanks againRegards
> > > > > Soren
> > > > >
> > > > >
> > > > >
> > > > >
> > > > > ----------------------------------------
> > > > >> Date: Sun, 27 May 2012 10:10:30 -0400
> > > > >> Subject: Re: st: predetermined variables with xtabond2
> > > > >> From: [email protected]
> > > > >> To: [email protected]
> > > > >>
> > > > >> Soren,
> > > > >> Did you read the following paper (The Stata Journal Volume 9
> Number
> > > 1:
> > > > >> pp. 86-136) from the author of this code:
> > > > >> http://www.stata-journal.com/article.html?article=st0159
> > > > >>
> > > > >> I believe that you wil get the answers to your queries in the
> paper.
> > > > >>
> > > > >> Best wishes,
> > > > >> Suryadipta.
> > > > >>
> > > > >> On Sat, May 26, 2012 at 4:49 AM, Søren Møller-Larsson
> > > > >> <[email protected]> wrote:
> > > > >> > Dear all
> > > > >> >
> > > > >> > in the Stata help file of xtabond2 it reads the following:
> > > > >> > "
> > > > >> > y_it = x_it * b_1 + w_it * b_2 + u_it ...
> > > > >> >
> > > > >> > x_it is a vector of strictly exogenous covariates (ones
> dependent
> > > on
> > > > >> > neither current nor past e_it);
> > > > >> >
> > > > >> > w_it is a vector of predetermined covariates (which may include
> the
> > > lag of
> > > > >> > y) and endogenous covariates, all of which may be
> > > correlated with
> > > > >> > the v_i (Predetermined variables are potentially
> > > correlated with
> > > > >> > past errors. Endogenous ones are potentially
> > > correlated with past
> > > > >> > and present errors.);
> > > > >> > "
> > > > >> > So to me it looks like predetermined variables are part of the
> > > gmmstyle() instruments.
> > > > >> >
> > > > >> > However further down the text it is explained how predetermined
> > > variables are treated in the ivstyle() instrument matrix:
> > > > >> >
> > > > >> >
> > > > >> > ..."equation() is useful for proper handling of predetermined
> > > variables
> > > > >> > used as IV-style instruments in system GMM. For example, if x
> > > is
> > > > >> > predetermined, it is a valid instrument for the levels
> > > equation since
> > > > >> > it is assumed to be uncorrelated with the contemporaneous
> > > error term.
> > > > >> > However, x becomes endogenous in first differences, so D.x is
> > > not a
> > > > >> > valid instrument for the transformed equation. ivstyle(x)
> > > would
> > > > >> > therefore be inappropriate. The use of x as an IV-style
> > > instrument in
> > > > >> > levels only could be specified by iv(x, eq(level))."
> > > > >> >
> > > > >> > So my question is, when do I use gmm(h, laglimits(1 .)) and
> when do
> > > I use iv(h, equation(level)) for predetermined h, and what is the
> > > difference?
> > > > >> >
> > > > >> > Kind regard
> > > > >> > Soren
> > > > >> > Aarhus university, Denmark
> > > > >> >
> > > > >> > *
> > > > >> > * 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/
> > > > >
> > > > > *
> > > > > * 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/
> > >
> > > *
> > > * 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/
> >
> > --
> > Justina AV Fischer, PhD
> > COFIT Fellow
> > World Trade Institute
> > University of Bern
> >
> > homepage: http://www.justinaavfischer.de/
> > e-mail: [email protected]. [email protected]
> > papers: http://ideas.repec.org/e/pfi55.html
> >
> >
> > *
> > * 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/
--
Justina AV Fischer, PhD
COFIT Fellow
World Trade Institute
University of Bern
homepage: http://www.justinaavfischer.de/
e-mail: [email protected]. [email protected]
papers: http://ideas.repec.org/e/pfi55.html
*
* 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/