Problem with number of instrument with xtabond2

[Marcus Romero]

Hello everyone,
I am just new to managing panel series and I am working on the investigation of financial stability. The model I am using is the two step GMM and I have a N = 11 and a T = 48, I use 1 dependent variable, 1 independent and 8 control. My problem arises when I run the model and it is that the Hansen test always gives me a p value of 1, in addition to that the number of instruments always comes out very high despite the fact that I only use the lag of the dependent as internal instruments and as external a variable related to the interest rate. If I try to use one more variable, the number of instruments sometimes increases to 130, that is what I cannot understand because even though I use 1 exogenous as an instrument, the number of instruments resulting from the test always gives me a very high value. The code I am using is the following:

xtabond2 Zscore l.Zscore l.LCTotal Capital C_I Deposit I_div size l.MarketShare l.PBI l.Inflación gmm(l.Zscore, collapse) iv(CD,equation(level)) nodiffsargan twostep robust orthogonal small


Dynamic panel-data estimation, two-step system GMM
------------------------------------------------------------------------------
Group variable: id Number of obs = 493
Time variable : qdate Number of groups = 11
Number of instruments = 49 Obs per group: min = 43
F(10, 10) = 96.59 avg = 44.82
Prob > F = 0.000 max = 45
------------------------------------------------------------------------------
| Corrected
Zscore | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
Zscore |
L1. | .2781406 .5680483 0.49 0.635 -.98755 1.543831
|
LCTotal |
L1. | 60.10113 53.74148 1.12 0.290 -59.64234 179.8446
|
Capital | 237.5321 202.9717 1.17 0.269 -214.7171 689.7812
C_I | -58.17815 48.10297 -1.21 0.254 -165.3583 49.00194
Deposit | -50.87249 40.33907 -1.26 0.236 -140.7536 39.00856
I_div | -37.41133 33.00418 -1.13 0.283 -110.9492 36.12657
size | -18.02906 15.16054 -1.19 0.262 -51.80884 15.75072
|
MarketShare |
L1. | -3.057071 2.683947 -1.14 0.281 -9.037278 2.923136
|
PBI |
L1. | -.2421628 .2463183 -0.98 0.349 -.7909941 .3066685
|
Inflación |
L1. | 1.944802 1.712228 1.14 0.283 -1.87028 5.759883
|
_cons | 321.238 269.9959 1.19 0.262 -280.3504 922.8263
------------------------------------------------------------------------------
Instruments for orthogonal deviations equation
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(1/.).L.Zscore collapsed
Instruments for levels equation
Standard
_cons
CD
GMM-type (missing=0, separate instruments for each period unless collapsed)
D.L.Zscore collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -0.70 Pr > z = 0.487
Arellano-Bond test for AR(2) in first differences: z = . Pr > z = .
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(38) = 41.33 Prob > chi2 = 0.327
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(38) = 0.00 Prob > chi2 = 1.000
(Robust, but can be weakened by many instruments.)


I would appreciate if you could explain to me the reason for the number of instruments so high and how I could overcome them

Best regards,
Marcus Romero

[Sebastian Kripfganz]

These GMM estimators are intended for panel models with large-N, small-T. You have the opposite situation. GMM-style instruments by default increase with the number of time periods.

The following presentation might be useful:

• Kripfganz, S. (2019). Generalized method of moments estimation of linear dynamic panel data models. Proceedings of the 2019 London Stata Conference.