st: RE: RE: RE: FW: clustering in ivreg2

["Schaffer, Mark E" <M.E.Schaffer@hw.ac.uk>]

Sterling,

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-
> statalist@hsphsun2.harvard.edu] On Behalf Of HUANG Sterling
> Sent: 28 January 2013 14:16
> To: statalist@hsphsun2.harvard.edu
> Subject: st: RE: RE: FW: clustering in ivreg2
> 
> Thanks Mark.
> 
> Can we do this manually in two stages: I estimate two first stage probit/logit
> regressions for X and X*Y using instrument Z and Z*Y  with clustering at
> country level and then use predicted value of X and X*Y in the second stage
> for performance regression? Would that be equivalent to ivreg2 procedure?

That's not going to work.  (And clustering in the first stage is irrelevant since it affects only the SEs and not the fitted values.)  Jeff Wooldridge's "Econometric Analysis of Cross-Section and Panel Data" has a good discussion of an alternative procedure that has been discussed on Statalist before.  Have a look at, e.g.,

http://www.stata.com/statalist/archive/2004-09/msg00339.html
http://www.stata.com/statalist/archive/2007-04/msg00415.html
http://www.stata.com/statalist/archive/2008-01/msg00458.html

--Mark

> 
> Best,
> 
> Sterling
> 
> 
> 
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-
> statalist@hsphsun2.harvard.edu] On Behalf Of Schaffer, Mark E
> Sent: Monday, 28 January 2013 8:58 PM
> To: statalist@hsphsun2.harvard.edu
> Subject: st: RE: FW: clustering in ivreg2
> 
> Sterling,
> 
> > -----Original Message-----
> > From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-
> > statalist@hsphsun2.harvard.edu] On Behalf Of HUANG Sterling
> > Sent: 28 January 2013 09:53
> > To: statalist@hsphsun2.harvard.edu
> > Subject: st: FW: clustering in ivreg2
> >
> > Dear Statalister,
> >
> > I have problem with clustering option in ivreg2 and I hope to get a
> > better understanding.
> >
> > Here is the problem:
> > I am regressing firm level performance on variable X and
> > X*Y(interaction between X and Y). X is endogenous, Y is exogenous and
> > X*Y by construction is endogenous. I use instrument Z and Z*Y to
> > instrument both X and X*Y. The problem is that X is at firm level while Z is at
> country level.
> >
> > When I use ivreg2 with clustering option at country level,
> >                 Ivreg2 performance (X XY=Z ZY) control variables,
> > cluster(country_code)
> >
> > I keep getting following warning messages from STATA:
> >
> > Warning: estimated covariance matrix of moment conditions not of full
> rank.
> > standard errors and model tests should be interpreted with caution.
> > Possible causes: number of clusters insufficient to calculate robust
> > covariance matrix singleton dummy variable (dummy with one 1 and N-1
> > 0s or vice
> > versa) partial option may address problem.
> >
> > I also tried with partial(country_code) option in the ivreg2 command,
> > and I get following error message.
> > Error: country_code listed in partial() but not in list of regressors.
> > invalid syntax
> >
> > Questions:
> > 1. Does anyone know what these warning and error mean and how to fix
> it?
> 
> The messages are fairly self-explanatory.  I think the Baum et al. 2003 and/or
> 2007 papers cited in the ivreg2 help file briefly discuss the first problem.  For a
> Stata online source, see
> 
> help j_robustsingular
> 
> from within Stata.
> 
> Your problem with the partial option is also self-explanatory:
> 
> > Error: country_code listed in partial() but not in list of regressors.
> > invalid syntax
> 
> is an error because country_code is not a REGRESSOR in
> 
> >  ivreg2 performance (X XY=Z ZY) control variables,
> > cluster(country_code)
> 
> You need to partial out REGRESSORS, i.e.,
> 
> >  ivreg2 performance (X XY=Z ZY) control variables, cluster(selected
> > control variables)
> 
> See also the discussion in the ivreg2 help file on the partial option:
> 
> help ivreg2##s_partial
> 
> --Mark
> 
> > 2. Can we do this manually in two stages: I estimate two first stage
> > probit/logit regressions for X and X*Y using instrument Z and Z*Y
> > with clustering at country level and then use predicted value of X and
> > X*Y in the second stage for performance regression? I tried, it went
> > through, but I am not sure econometrically whether I can do this or not.
> >
> > Best,
> >
> > Sterling
> >
> >
> >
> > *
> > *   For searches and help try:
> > *   http://www.stata.com/help.cgi?search
> > *   http://www.stata.com/support/faqs/resources/statalist-faq/
> > *   http://www.ats.ucla.edu/stat/stata/
> 
> 
> -----
> Sunday Times Scottish University of the Year 2011-2013 Top in the UK for
> student experience Fourth university in the UK and top in Scotland (National
> Student Survey 2012)
> 
> We invite research leaders and ambitious early career researchers to join us
> in leading and driving research in key inter-disciplinary themes.
> Please see www.hw.ac.uk/researchleaders for further information and how
> to apply.
> 
> Heriot-Watt University is a Scottish charity registered under charity number
> SC000278.
> 
> 
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/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/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/


----- 
Sunday Times Scottish University of the Year 2011-2013
Top in the UK for student experience
Fourth university in the UK and top in Scotland (National Student Survey 2012)

We invite research leaders and ambitious early career researchers to 
join us in leading and driving research in key inter-disciplinary themes. 
Please see www.hw.ac.uk/researchleaders for further information and how
to apply.

Heriot-Watt University is a Scottish charity
registered under charity number SC000278.


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/