st: RE: xtabond2 and Sargan test

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

Olga,

> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu 
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of 
> Olga Gorbachev
> Sent: 05 October 2009 15:00
> To: statalist@hsphsun2.harvard.edu
> Subject: st: xtabond2 and Sargan test
> 
> Dear Listservers,
> 
> 
> I am running xtabond2 option.
> I get that Sargan test of overid. restrictions:
> chi2(188)  = 250.54  Prob > chi2 =  0.002
> 
> Hansen test of overid. restrictions:
> chi2(188)  = 175.52  Prob > chi2 =  0.734
> 
> Does this mean that I should reduce the number of instruments?
> Additionally, I have AR(2) in first differences and thus am 
> using gmm(l.Z, lag(3 .)) to specify the instruments to be used.
> 
> Could you please clarify whether the Hansen/Sargan tests are 
> valid under this specification?

For what it's worth...

You *could* argue that Sargan is rejecting and Hansen is failing to
reject because Hansen is more robust than Sargan.  For example, Sargan
is not distributed as chi-sq under heteroskedasticity and Hansen is, and
if this problem is present then it could cause Sargan to incorrectly
reject the null.

But... estimations with huge numbers of instruments like this are
susceptible to various problems.

Have you tried looking at David Roodman's "How to do xtabond2?"  He has
a discussion of this, I think.  If you're using a recent version of
xtabond2, the reference should be in the help file.

Cheers,
Mark

> 
> Thank you in advance,
> Olga
> 
> Here is the syntax of the command I ran and the output:
> 
> ***
> xtabond2 df age age2  ed12  nwe12 perd2 perd3 perd4 lnrtb3 
> dna dnk dms dhrsw dhrsh dyu2, gmm(L.(lnrtb3 dms dna dnk dfu 
> dyu2 dhrsh dhrsw),
> lag(3) collapse) iv(age age2  edCol edColp ednoHS) twostep robust 
> noconstant small   orthogonal  art(3)
> 
> Warning: Two-step estimated covariance matrix of moments is singular.
>    Using a generalized inverse to calculate optimal weighting 
> matrix for two-step estimation.
>    Difference-in-Sargan/Hansen statistics may be negative.
> 
> Dynamic panel-data estimation, two-step system GMM
> --------------------------------------------------------------
> ----------------
> Group variable: pid                Number of obs      =     31428
> Time variable : year               Number of groups   =      6715
> Number of instruments = 202        Obs per group: min =         1
> F(14, 6715)   =     12.30                         avg =      4.68
> Prob > F      =     0.000                         max =        23
> --------------------------------------------------------------
> ----------------
>               |              Corrected
>            df |      Coef.   Std. Err.      t    P>|t|     [95% Conf. 
> Interval]
> -------------+------------------------------------------------
> ----------
> -------------+------
>   age |   .0101859   .0072103     1.41   0.158    -.0039485   
>  .0243203
> age2 |  -.0001286   .0000823    -1.56   0.118    -.0002899    .0000327
> ed12 |  -.0650101   .0351083    -1.85   0.064    -.1338335    .0038133
> nwe12 |   .1055316   .0653455     1.61   0.106    -.0225664   
>  .2336296
> perd2 |  -.0762958   .1382154    -0.55   0.581    -.3472419   
>  .1946503
> perd3 |   .0147213   .1381243     0.11   0.915    -.2560462   
>  .2854888
>   perd4 |  -.1659716    .155361    -1.07   0.285    -.4705284 
>    .1385852
> lnrtb3 |   .8725896   .3702258     2.36   0.018     .1468295  
>    1.59835
>   dna |  -.0122323   .0858887    -0.14   0.887    -.1806013   
>  .1561368
> dnk |   .0865718   .0948575     0.91   0.361     -.099379    .2725226
>   dms |  -.0699546   .0793642    -0.88   0.378    -.2255337   
>  .0856244
>   dhrsw |   .0391827   .0248344     1.58   0.115    -.0095006 
>     .087866
>   dhrsh |   .0287372   .0310284     0.93   0.354    -.0320883 
>    .0895626
>   dyu2 |  -.2672077   .1173428    -2.28   0.023    -.4972369  
>  -.0371786
> --------------------------------------------------------------
> ----------------
> Instruments for orthogonal deviations equation
>    Standard
>      FOD.(age age2 edCol edColp ednoHS)
>    GMM-type (missing=0, separate instruments for each period unless
> collapsed)
>      L(3/.).(L.lnrtb3 L.dms L.dna L.dnk L.dfu L.dyu2 L.dhrsh 
> L.dhrsw) collapsed Instruments for levels equation
>    Standard
>      age age2 edCol edColp ednoHS
>    GMM-type (missing=0, separate instruments for each period unless
> collapsed)
>      DL2.(L.lnrtb3 L.dms L.dna L.dnk L.dfu L.dyu2 L.dhrsh 
> L.dhrsw) collapsed
> --------------------------------------------------------------
> ----------------
> Arellano-Bond test for AR(1) in first differences:
> z = -25.63  Pr > z =  0.000
> Arellano-Bond test for AR(2) in first differences:
> z =   8.11  Pr > z =  0.000
> Arellano-Bond test for AR(3) in first differences:
> z =   0.10  Pr > z =  0.919
> --------------------------------------------------------------
> ----------------
> Sargan test of overid. restrictions:
> chi2(188)  = 250.54  Prob > chi2 =  0.002
>    (Not robust, but not weakened by many instruments.) Hansen 
> test of overid. restrictions:
> chi2(188)  = 175.52  Prob > chi2 =  0.734
>    (Robust, but can be weakened by many instruments.)
> 
> Difference-in-Hansen tests of exogeneity of instrument subsets:
>    GMM instruments for levels
>      Hansen test excluding group:
> chi2(180)  = 169.29  Prob > chi2 =  0.706
>      Difference (null H = exogenous):
> chi2(8)    =   6.23  Prob > chi2 =  0.622
>    iv(age age2 edCol edColp ednoHS)
>      Hansen test excluding group:
> chi2(183)  = 164.35  Prob > chi2 =  0.835
>      Difference (null H = exogenous):
> chi2(5)    =  11.17  Prob > chi2 =  0.048
> 
> 
> --
> The University of Edinburgh is a charitable body, registered 
> in Scotland, with registration number SC005336.
> 
> 


-- 
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/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/