Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: All two-step sysGMM coefficients are insignificant


From   [email protected]
To   [email protected]
Subject   st: All two-step sysGMM coefficients are insignificant
Date   Fri, 06 Mar 2009 09:59:52 +0100

Hello everybody,

I have a balanced panel of 19 countries over 24 time periods. The model is

    g_it= ß_0+δ*g_(it-1)+ß_1*x_it+ε_it

It is a income growth regression, where the regressor are population growth, institutions, the trade share and so on.

I used until now, the one-step system GMM estimator with robust standard errors.
xtdpd  g l.g FD CL I LQ RS RF  ddummy p,  dgmmiv(I LQ RS FD RF  ddummy l.g , lagrange(4)) lgmmiv(I LQ RS  ddummy FD RF l.g, lag(4)) iv(p)  vce(robust) artests(2)


Dynamic panel-data estimation                Number of obs         =       435
Group variable: id                           Number of groups      =        19
Time variable: time
                                             Obs per group:    min =        21
                                                               avg =  22.89474
                                                               max =        23

Number of instruments =    470               Wald chi2(9)          =     74.97
                                             Prob > chi2           =    0.0000
One-step results
------------------------------------------------------------------------------
             |               Robust
           g |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
           g |
         L1. |   .1320479   .0528852     2.50   0.013     .0283948     .235701
          FD |   .3835768   .1107316     3.46   0.001     .1665468    .6006068
          CL |  -.0000431   .0019382    -0.02   0.982     -.003842    .0037557
           I |  -.2958265   .0795866    -3.72   0.000    -.4518134   -.1398396
          LQ |  -.0002257   .0090368    -0.02   0.980    -.0179375    .0174861
          RS |   .0107728   .0041573     2.59   0.010     .0026247    .0189209
          RF |  -.0195168   .0047356    -4.12   0.000    -.0287984   -.0102352
      ddummy |   .0054591   .0041961     1.30   0.193    -.0027651    .0136834
           p |  -.0006131   .0002331    -2.63   0.009      -.00107   -.0001562
       _cons |   .0229008   .0155701     1.47   0.141     -.007616    .0534176
------------------------------------------------------------------------------
Instruments for differenced equation
        GMM-type: L(4/.).I L(4/.).LQ L(4/.).RS L(4/.).FD L(4/.).RF L(4/.).ddummy
                  L(4/.).L.g
        Standard: D.p
Instruments for level equation
        GMM-type: L4D.I L4D.LQ L4D.RS L4D.ddummy L4D.FD L4D.RF L5D.g
        Standard: p _cons


To compare the results, because I think the two step estimator can better deal 
with heteroskedasticity (Windmeijer robust errors), I want to employ the two step estimator.
However, all the coefficients become insignificant. How can I interpret these results?
And are the one step (with robust standard errors) estimates also robust to heteroskedasticity (across time and countries)?



xtdpd  g l.g FD CL I LQ RS RF  ddummy p, twostep  dgmmiv(I LQ RS FD RF  ddummy l.g , lagrange(4)) lgmmiv(I LQ RS  ddummy FD RF  l.g, lag(4)) iv(p)  vce(robust) artests(2)

Dynamic panel-data estimation                Number of obs         =       435
Group variable: id                           Number of groups      =        19
Time variable: time
                                             Obs per group:    min =        21
                                                               avg =  22.89474
                                                               max =        23

Number of instruments =    470               Wald chi2(8)          =      3.19
                                             Prob > chi2           =    0.9220
Two-step results
------------------------------------------------------------------------------
             |              WC-Robust
           g |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
           g |
         L1. |  -.1018478   3.185356    -0.03   0.974    -6.345031    6.141336
          FD |   .1034821   16.18238     0.01   0.995    -31.61339    31.82036
          CL |  -.0076123   .1367406    -0.06   0.956     -.275619    .2603944
           I |   -.073898    17.0281    -0.00   0.997    -33.44835    33.30056
          LQ |   .0169259   .6985014     0.02   0.981    -1.352112    1.385963
          RS |   .1058199   .7143544     0.15   0.882    -1.294289    1.505929
          RF |  -.1225682   .7167893    -0.17   0.864    -1.527449    1.282313
      ddummy |   .0806485   .2211386     0.36   0.715    -.3527752    .5140722
           p |  -.0019078   .0129519    -0.15   0.883    -.0272929    .0234774
       _cons |   .0281583   1.297514     0.02   0.983    -2.514923     2.57124
------------------------------------------------------------------------------
Instruments for differenced equation
        GMM-type: L(4/.).I L(4/.).LQ L(4/.).RS L(4/.).FD L(4/.).RF L(4/.).ddummy
                  L(4/.).L.g
        Standard: D.p
Instruments for level equation
        GMM-type: L4D.I L4D.LQ L4D.RS L4D.ddummy L4D.FD L4D.RF L5D.g
        Standard: p _cons




P.S. It does not matter, if I restrict the lagrange.

Hope somebody can help.

Best wishes

Markus
-- 
Psssst! Schon vom neuen GMX MultiMessenger gehört? Der kann`s mit allen: http://www.gmx.net/de/go/multimessenger01
*
*   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/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index