GMM Estimation

[Sarah Khalid]

I am running system GMM and having problem in my results. i gave the following command

. xtabond2 umr l.umr oda imn se urb gdp dem bss corp, gmm(umr se urb dem, lag(1 1)) iv(oda imn gdp bss corp) robust small

How to reduce the number of instruments as in my results no of instruments are greater than no of groups which should be less.

Dynamic panel-data estimation, one-step system GMM
------------------------------------------------------------------------------
Group variable: country Number of obs = 2350
Time variable : years Number of groups = 94
Number of instruments = 202 Obs per group: min = 25
F(9, 93) = 8282.05 avg = 25.00
Prob > F = 0.000 max = 25

The results are as follows

| Robust
umr | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
umr |
L1. | .9427665 .0094791 99.46 0.000 .9239429 .9615901
|
oda | -.0206139 .009431 -2.19 0.031 -.0393421 -.0018858
imn | -.0419936 .0117621 -3.57 0.001 -.0653507 -.0186365
se | -.0409261 .0182737 -2.24 0.027 -.077214 -.0046382
urb | -.7012923 .2047653 -3.42 0.001 -1.107916 -.2946691
gdp | -.055229 .0192167 -2.87 0.005 -.0933896 -.0170684
dem | .8264338 .2004452 4.12 0.000 .4283895 1.224478
bss | -.0416304 .0160471 -2.59 0.011 -.0734968 -.0097639
corp | 1.67886 .489032 3.43 0.001 .7077394 2.649981
_cons | 11.0575 3.104658 3.56 0.001 4.892268 17.22274
------------------------------------------------------------------------------

My AR2 value is less than 5% which should be greater as it says in my case there is second order autocorrelation. Kindly tell me how to remove that.

Arellano-Bond test for AR(1) in first differences: z = -1.82 Pr > z = 0.069
Arellano-Bond test for AR(2) in first differences: z = 2.19 Pr > z = 0.028

And value of Sargan test is also less than 5% which should be greater as in my case instruments are not valid.

Sargan test of overid. restrictions: chi2(192) =4944.62 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(192) = 86.87 Prob > chi2 = 1.000
(Robust, but weakened by many instruments.)

Kindly tell me what should i do to correct these things. Thanks in advance.

[Sebastian Kripfganz]

The easiest way to reduce the number of instruments is the collapse suboption of the gmm() option. You should first reestimate the model with a reasonably small number of instruments before looking at the test statistics.

[Sarah Khalid]

Thank you for replying.
By using collapse option, number of instruments are reduced but now variales become insignificant

[Sarah Khalid]

Sebastian Kripfganz