System GMM with panel data

[Guest]

Hi

I had run a system GMM panel data regression on STATA 14 and I got

Variance-covariance matrix of the two-step estimator is not full rank.
Two-step estimator is not available. One-step estimator is available and variance-covariance matrix provides correct coverage.

Is running system GMM twostep impossible in this case, however my sample is very small ( Number of obs 157)? And how can I solve this problem?

[Sven-Kristjan Bormann]

Welcome to the forum,
to get an answer in this forum, it helps to get familiar with the FAQs http://www.statalist.org/forums/help
To help you, we need more information about your dataset and the exact command that you have run so far. Stata has different options to run System GMM, therefore it is not clear which commands did you use and with which options.

[Guest]

eststo:xtabond2 roa vaic bsize bind bmeet rcind acind acsize acmeet con duality femaledirectors natin edubackground lnbanksize blr bage bankcomplexity mbv ato gdp infl baseli
> ii car beta i.year, gmm(roa vaic) iv(lnbanksize blr bage bankcomplexity mbv ato gdp infl baseliii car beta, equation(level)) nodiffsargan twostep robust orthogonal small
Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.
Warning: Number of instruments may be large relative to number of observations.
Warning: Two-step estimated covariance matrix of moments is singular.
Using a generalized inverse to calculate optimal weighting matrix for two-step estimation.

Dynamic panel-data estimation, two-step system GMM
------------------------------------------------------------------------------
Group variable: bankname1 Number of obs = 178
Time variable : year Number of groups = 25
Number of instruments = 82 Obs per group: min = 4
F(32, 24) = 135.12 avg = 7.12
Prob > F = 0.000 max = 8
---------------------------------------------------------------------------------
| Corrected
roa | Coef. Std. Err. t P>|t| [95% Conf. Interval]
----------------+----------------------------------------------------------------
vaic | .0091604 .0144536 0.63 0.532 -.0206703 .0389912
bsize | -.091867 .1033516 -0.89 0.383 -.3051741 .1214402
bind | 0 (omitted)
bmeet | -.0196124 .026448 -0.74 0.466 -.0741983 .0349735
rcind | .1122069 .0999533 1.12 0.273 -.0940866 .3185004
acind | 0 (omitted)
acsize | .1084461 .1370031 0.79 0.436 -.1743145 .3912067
acmeet | .0072846 .0246211 0.30 0.770 -.0435309 .0581001
con | .0417183 .1653702 0.25 0.803 -.2995891 .3830257
duality | 0 (omitted)
femaledirectors | .1353364 .2394799 0.57 0.577 -.3589259 .6295987
natin | -.026946 .0568616 -0.47 0.640 -.1443027 .0904106
edubackground | 0 (omitted)
lnbanksize | .0212808 .0181405 1.17 0.252 -.0161593 .0587209
blr | .0513118 .0406106 1.26 0.219 -.0325044 .135128
bage | -.0024112 .0054 -0.45 0.659 -.0135562 .0087338
bankcomplexity | .0324853 .033139 0.98 0.337 -.0359103 .1008809
mbv | .0048779 .0174684 0.28 0.782 -.0311752 .0409309
ato | 0 (omitted)
gdp | -.0028912 .0048245 -0.60 0.555 -.0128484 .007066
infl | .0076665 .00464 1.65 0.112 -.0019099 .0172429
baseliii | .0101087 .0788272 0.13 0.899 -.1525826 .1728001
car | 0 (omitted)
beta | .0610936 .0855293 0.71 0.482 -.1154302 .2376173
|
year |
2009 | 0 (empty)
2010 | .0511826 .0737254 0.69 0.494 -.1009792 .2033444
2011 | -.0153034 .0983908 -0.16 0.878 -.218372 .1877652
2012 | -.0232941 .0661021 -0.35 0.728 -.1597221 .1131338
2013 | -.0354119 .0897522 -0.39 0.697 -.2206513 .1498275
2014 | -.0122543 .114754 -0.11 0.916 -.2490948 .2245862
2015 | -.0292923 .1340039 -0.22 0.829 -.3058626 .2472781
2016 | -.0214804 .1404789 -0.15 0.880 -.3114146 .2684539
|
_cons | 0 (omitted)
---------------------------------------------------------------------------------
Instruments for orthogonal deviations equation
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(1/7).(roa vaic)
Instruments for levels equation
Standard
lnbanksize blr bage bankcomplexity mbv ato gdp infl baseliii car beta
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
D.(roa vaic)
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -0.98 Pr > z = 0.326
Arellano-Bond test for AR(2) in first differences: z = -0.41 Pr > z = 0.685
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(49) = 68.70 Prob > chi2 = 0.033
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(49) = 0.00 Prob > chi2 = 1.000
(Robust, but weakened by many instruments.)

[Guest]

Thank you so much Sven-Kristjan for your response
Is there any problem with no.of instruments > no.of group? !!!

[Dario Maimone Ansaldo Patti]

Yes. The Hansen test has a p value equal to 1. As the results warn you, its validity can be compromised by too many instruments and this is the case. You should reduce the number of instruments to get more meaningful results