Problem in reporting Arellano-Bond test for auto-correlation after xtabond2o

[Amaury Barra]

Hello dear Stata Users,
Using xtabond2 in stata 11, I am estimating a dynamical panel (N=94 ; T=8) where my dependent var is the number of the Ministry of Defence duty personel (at logarithm) and and my independents aren the lagged dependent, the log-revenue for male workers, the unemployment rate, the share of industry in the economy and the log-density of population.
I use the following command : xtabond2 lqtym_ unemp_ L.lqtym_ ldensity_ lmev_ part_indus , gmmstyle(L.lqtym_ , laglimit(2 .)) iv(unemp_ ldensity_ lmrev_ part_indus, mz) twostep robust
Stata reports correctly the Sargan/Hansen tests but do report for the Arellano-Bond test for autocorellation

Arellano-Bond test for AR(1) in first differences: z = . Pr > z = .
Arellano-Bond test for AR(2) in first differences: z = . Pr > z = .

My panel is not balanced and I have missing values in ldensity_ and lmrev_ can it be a reason ? I thought that not, as missing values are changed to 0.

Does anyone have an idea ? I am stuck here.

Thank you all in advance,

Amaury

[Amaury Barra]

Post scrip-tom for greater understanding I attached the code and Stata output
xtabond2 lqtym_ txchom_ L.lqtym_ L2.lqtym_ ldensite_ lrevh_ part_indus , gmmstyle(L.lqtym_ L2.lqtym_ , laglimit(1 .)) iv(txchom_ ldensite_ lrevh_ part_indus , mz eq
> (diff)) twostep orthogonal
Favoring speed over space. To switch, type or click on mata: mata set matafavor space, perm.
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: _ID Number of obs = 188
Time variable : year Number of groups = 94
Number of instruments = 11 Obs per group: min = 2
Wald chi2(6) = 82.77 avg = 2.00
Prob > chi2 = 0.000 max = 2

lqtym_ Coef. Std. Err. z P>z [95% Conf. Interval]

txchom_ -.0597908 .1841901 -0.32 0.745 -.4207967 .301215

lqtym_
L1. .6519922 .2997847 2.17 0.030 .0644251 1.239559
L2. .523822 .3906448 1.34 0.180 -.2418276 1.289472

ldensite_ -.8942009 2.095412 -0.43 0.670 -5.001133 3.212732
lrevh_ -1.973115 1.62359 -1.22 0.224 -5.155293 1.209063
part_indus 8.749925 5.304268 1.65 0.099 -1.64625 19.1461
_cons 13.64138 15.87719 0.86 0.390 -17.47734 44.7601

Warning: Uncorrected two-step standard errors are unreliable.

Instruments for orthogonal deviations equation
Standard
FOD.(txchom_ ldensite_ lrevh_ part_indus), missing recoded as zero
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(1/7).(L.lqtym_ L2.lqtym_)
Instruments for levels equation
Standard
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
D.(L.lqtym_ L2.lqtym_)

Arellano-Bond test for AR(1) in first differences: z = . Pr > z = .
Arellano-Bond test for AR(2) in first differences: z = . Pr > z = .

Sargan test of overid. restrictions: chi2(4) = 20.70 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(4) = 8.28 Prob > chi2 = 0.082
(Robust, but weakened by many instruments.)

Difference-in-Hansen tests of exogeneity of instrument subsets:
GMM instruments for levels
Hansen test excluding group: chi2(0) = 0.33 Prob > chi2 = .
Difference (null H = exogenous): chi2(4) = 7.95 Prob > chi2 = 0.094
iv(txchom_ ldensite_ lrevh_ part_indus, mz eq(diff))
Hansen test excluding group: chi2(1) = 0.25 Prob > chi2 = 0.617
Difference (null H = exogenous): chi2(3) = 8.03 Prob > chi2 = 0.045


.
end of do-file

Sorry for the variable name in french,

Many thanks,

Amaury

[Sebastian Kripfganz]

It indeed seems to be a problem with missing values. They are changed to zero for the instruments but not for the regressors. The reason why your Arellano-Bond tests do not report results is that your minimum number of observations is 2, but you need at least 3 observations to compute the test statistic for the AR(1) test, 4 observations for the AR(2) test, and so on.

[Amaury Barra]

Dear Sebastian,
Now that you make the remark it sounds obvious, I was so focused on the instruments that I forgot to address the problem in the regressors !

Many thanks !

Amaury