xtabond2 : Hansen test of overid. restrictions

[Ismaila Souare]

Hello Dear Stata user's,

I'm using STATA command xtabond2 and system GMM for my very first project. According to Arellano and Bond (1991), Arellano and Bover (1995) and Blundell and Bond (1998), two necessary tests (Sargan/Hansen and AR2) should be used.
Based on my reading, Sargan and Hansen are used to test the overall validity of the instruments.

My question is : if we rely on the results of the hansen test, can we validate the instruments?

The results for sargan and hansen tests are below:

.
Arellano-Bond test for AR(1) in first differences: z = -2.50 Pr > z = 0.012
Arellano-Bond test for AR(2) in first differences: z = -0.19 Pr > z = 0.849

Sargan test of overid. restrictions: chi2(137) = 183.79 Prob > chi2 = 0.005
(Not robust, but not weakened by many instruments.)

Hansen test of overid. restrictions: chi2(137) = 0.00 Prob > chi2 = 1.000
(Robust, but can be weakened by many instruments.)

Difference-in-Hansen tests of exogeneity of instrument subsets:
GMM instruments for levels

Hansen test excluding group: chi2(123) = 0.00 Prob > chi2 = 1.000
Difference (null H = exogenous): chi2(14) = 0.00 Prob > chi2 = 1.000

[Sebastian Kripfganz]

The p-value of your Hansen test (1.000) together with the large number of degrees of freedoms (137) indicates that your estimates are likely suffering from the too-many-instruments problem. Please read through the following paper to make yourself familiar with this and other potential pitfalls when estimating dynamic panel models with GMM:

• Roodman (2009). How to do xtabond2: An introduction to difference and system GMM in Stata. Stata Journal 9 (1), pp. 86-136.

[Ismaila Souare]

Hello Sebastian,

Thank you for your answer.

I 'll read it for better understanding.

Best regards.

[Sinem Ates]

Dear All,

I run xtabond2 command to solve endogeneity problem in my regression model. According to the results, there is no second order autocorrelation and difference Hansen tests for GMM and IV are confirming the exogeneity of instrument subsets. Also, the coefficients and significancy levels are in accordance with the theory. However, p value of Hansen test of overid. restrictions is 0.011. Could you please tell me possible ways to transform my model below to solve this problem?

xtabond2 TOBINSQ_w L.TOBINSQ_w ESGSCORE SIZE_w LEV_w ROA_w ICBIC2 ICBIC3 ICBIC4 ICBIC5 ICBIC6 ICBIC7 ICBIC8 ICBIC9 ICBIC10 ICBIC11 YEAR2 YEAR3 YEAR4 YEAR5 YEAR6 YEAR7 YEAR8 YEAR9 YEAR10, noconstant gmm(L.(TOBINSQ_w ESGSCORE SIZE_w LEV_w ROA_w), collapse) iv(ICBIC2 ICBIC3 ICBIC4 ICBIC5 ICBIC6 ICBIC7 ICBIC8 ICBIC9 ICBIC10 ICBIC11 YEAR2 YEAR3 YEAR4 YEAR5 YEAR6 YEAR7 YEAR8 YEAR9 YEAR10) twostep robust small orthogonal

Dynamic panel-data estimation, two-step system GMM
------------------------------------------------------------------------------
Group variable: ID                              Number of obs      =      3260
Time variable : YEAR                            Number of groups   =       625
Number of instruments = 64                      Obs per group: min =         1
F(24, 625)    =   1299.23                                      avg =      5.22
Prob > F      =     0.000                                      max =         9
------------------------------------------------------------------------------
             |              Corrected
   TOBINSQ_w |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
   TOBINSQ_w |
         L1. |   .6416167   .0633606    10.13   0.000     .5171913     .766042
             |
    ESGSCORE |   .0039574   .0014761     2.68   0.008     .0010586    .0068561
      SIZE_w |  -.0719947   .0274938    -2.62   0.009    -.1259861   -.0180033
       LEV_w |   .4354721   .2204205     1.98   0.049     .0026177    .8683265
       ROA_w |   .0295865   .0076507     3.87   0.000     .0145623    .0446107
      ICBIC2 |  -.1427394   .1762176    -0.81   0.418    -.4887897    .2033108
      ICBIC3 |  -.0375606   .1562704    -0.24   0.810    -.3444393    .2693181
      ICBIC4 |  -.1138193   .1956555    -0.58   0.561    -.4980411    .2704024
      ICBIC5 |  -.2603883    .173164    -1.50   0.133    -.6004421    .0796654
      ICBIC6 |  -.0878802   .1550526    -0.57   0.571    -.3923673     .216607
      ICBIC7 |   .1304957   .1575019     0.83   0.408    -.1788014    .4397928
      ICBIC8 |  -.1732675   .1665113    -1.04   0.298    -.5002568    .1537219
      ICBIC9 |   -.217849   .1653956    -1.32   0.188    -.5426475    .1069494
     ICBIC10 |  -.1670711   .1793827    -0.93   0.352    -.5193369    .1851947
     ICBIC11 |  -.1700764   .1832575    -0.93   0.354    -.5299514    .1897986
       YEAR2 |   1.254697   .3762052     3.34   0.001     .5159179    1.993477
       YEAR3 |   .9854767   .3700407     2.66   0.008     .2588031     1.71215
       YEAR4 |    1.12928   .3686532     3.06   0.002     .4053312    1.853229
       YEAR5 |   1.074168   .3645176     2.95   0.003     .3583407    1.789996
       YEAR6 |    1.11878   .3663088     3.05   0.002     .3994351    1.838125
       YEAR7 |   1.102081   .3682072     2.99   0.003     .3790082    1.825154
       YEAR8 |   1.075689   .3697049     2.91   0.004     .3496752    1.801703
       YEAR9 |    1.13531   .3676124     3.09   0.002     .4134052    1.857215
      YEAR10 |   1.024656   .3687495     2.78   0.006     .3005175    1.748794
------------------------------------------------------------------------------
Instruments for orthogonal deviations equation
  Standard
    FOD.(ICBIC2 ICBIC3 ICBIC4 ICBIC5 ICBIC6 ICBIC7 ICBIC8 ICBIC9 ICBIC10
    ICBIC11 YEAR2 YEAR3 YEAR4 YEAR5 YEAR6 YEAR7 YEAR8 YEAR9 YEAR10)
  GMM-type (missing=0, separate instruments for each period unless collapsed)
    L(1/9).(L.TOBINSQ_w L.ESGSCORE L.SIZE_w L.LEV_w L.ROA_w) collapsed
Instruments for levels equation
  Standard
    ICBIC2 ICBIC3 ICBIC4 ICBIC5 ICBIC6 ICBIC7 ICBIC8 ICBIC9 ICBIC10 ICBIC11
    YEAR2 YEAR3 YEAR4 YEAR5 YEAR6 YEAR7 YEAR8 YEAR9 YEAR10
  GMM-type (missing=0, separate instruments for each period unless collapsed)
    D.(L.TOBINSQ_w L.ESGSCORE L.SIZE_w L.LEV_w L.ROA_w) collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z =  -5.50  Pr > z =  0.000
Arellano-Bond test for AR(2) in first differences: z =  -1.81  Pr > z =  0.071
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(40)   =  86.19  Prob > chi2 =  0.000
  (Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(40)   =  63.33  Prob > chi2 =  0.011
  (Robust, but weakened by many instruments.)

Difference-in-Hansen tests of exogeneity of instrument subsets:
  GMM instruments for levels
    Hansen test excluding group:     chi2(35)   =  54.41  Prob > chi2 =  0.019
    Difference (null H = exogenous): chi2(5)    =   8.92  Prob > chi2 =  0.112
  iv(ICBIC2 ICBIC3 ICBIC4 ICBIC5 ICBIC6 ICBIC7 ICBIC8 ICBIC9 ICBIC10 ICBIC11 YEAR2 YEAR3 YEAR4 YEAR5 YEAR6 YEAR7 YEAR8 YEAR9 YEAR10)
    Hansen test excluding group:     chi2(21)   =  35.05  Prob > chi2 =  0.028
    Difference (null H = exogenous): chi2(19)   =  28.28  Prob > chi2 =  0.078

[Sebastian Kripfganz]

Note that there is a potentially severe bug in xtabond2 that produces incorrect estimates when used with the orthogonal option; see slides 70 and 71 of my 2019 London Stata Conference presentation. I recommend to use the xtdpdgmm command instead.

• Kripfganz, S. (2019). Generalized method of moments estimation of linear dynamic panel data models. Proceedings of the 2019 London Stata Conference.
• XTDPDGMM: new Stata command for GMM estimation of linear (dynamic) panel data models

[Sinem Ates]

Dear Sebastian,

First of all thank you very much for your fast reply. I have tried to replicate my model using xtdpdgmm instead of xtbond2 command by following the posts on XTDPDGMM: new Stata command for GMM estimation of linear (dynamic) panel data models. However, because there have been many updates over time, I could not achieve to find the right syntax. Is it possible you to write the right syntax with xtdpdgmm command for the model below? If not, I would also appreciate an example with xtdpdgmm command including year and industry dummies and collapse, twostep, robust, small, orthogonal options.

 
 xtabond2 TOBINSQ_w L.TOBINSQ_w ESGSCORE SIZE_w LEV_w ROA_w ICBIC2 ICBIC3 ICBIC4 ICBIC5 ICBIC6 ICBIC7 ICBIC8 ICBIC9 ICBIC10 ICBIC11 YEAR2 YEAR3 YEAR4 YEAR5 YEAR6 YEAR7 YEAR8 YEAR9 YEAR10, noconstant gmm(L.(TOBINSQ_w ESGSCORE SIZE_w LEV_w ROA_w), collapse) iv(ICBIC2 ICBIC3 ICBIC4 ICBIC5 ICBIC6 ICBIC7 ICBIC8 ICBIC9 ICBIC10 ICBIC11 YEAR2 YEAR3 YEAR4 YEAR5 YEAR6 YEAR7 YEAR8 YEAR9 YEAR10) twostep robust small orthogonal

ICBIC represents industry dummies and year represents year dummies, I dropped ICBIC1 and YEAR1 manually.

Thank you in advance

[Sinem Ates]

Dear All,

I would appreciate an example with xtdpdgmm command including year and industry dummies and collapse, twostep, robust, small, orthogonal options as in xtabond2 command.

[Eric de Souza]

See:https://www.stata.com/meeting/uk19/s..._kripfganz.pdf
Sebastian Kripfganz has posted this link many times

[Sinem Ates]

Dear Sebastian,
I have studied both your presentation and the posts on the link you sent. Now I am trying to apply the sequential model selection process in order to find the most suitable model specification. I have already applied the first 3 steps as you explained in your presentation. Before moving to the 4.step, I need your comment to be sure that I am in the right way.
I run 20 models with different lags of both dependent and independent variables. Then, I used the MMSC to compare the models which pass the specification tests. The result is as follows:

       Model | ngroups          J  nmom  npar   MMSC-AIC   MMSC-BIC  MMSC-HQIC
-------------+----------------------------------------------------------------
           . |     625    38.0683    49    18   -23.9317  -161.5020   -78.5413
     model10 |     454    43.6310    48    16   -20.3690  -152.1481   -73.4483
      model9 |     440    33.4082    47    18   -24.5918  -143.1082   -72.3942
      model8 |     440    29.7748    47    19   -26.2252  -140.6549   -72.3792
      model7 |     440    28.6277    47    20   -25.3723  -135.7152   -69.8780
      model6 |     440    28.6205    47    21   -23.3795  -129.6356   -66.2368
      model5 |     440    25.9056    47    22   -24.0944  -126.2638   -65.3034
      model4 |     440    25.5041    47    23   -22.4959  -120.5785   -62.0565
      model3 |     440    25.6459    47    24   -20.3541  -114.3499   -58.2664
      model2 |     440    26.1705    47    25   -17.8295  -107.7386   -54.0934
      model1 |     440    25.2280    47    26   -16.7720  -102.5943   -51.3876

Should I select the model10 although it does not have the lowest values for all criterias?

[Sebastian Kripfganz]

There is an important aspect about the use of the model selection criteria: You can only meaningfully compare models with the same number of groups. In your case, model10 and the last estimated model (indicated by the dot) have more groups than all the other models. A larger data set by itself can be useful, but for model comparison purposes you should restrict the estimation samples to be the same.

[Sinem Ates]

Dear Sebastian,

Thank you for this warning. The first 9 models have the three lags of ESGSCORE variable so they have less number of groups. When I use MMSC to compare these 9 models, the result is as follows:

Models

xtdpdgmm L(0/3).TOBINSQ L(0/3).( ESGSCORE SIZE LEV ROA ), model(fod) collapse gmm( TOBINSQ , lag(1 .)) gmm( ESGSCORE , lag(1 .)) gmm( SIZE , lag(1 .)) gmm( LEV , lag(1 .)) gmm( ROA , lag(1 .)) teffects two vce(r) overid
xtdpdgmm L(0/2).TOBINSQ L(0/3).( ESGSCORE SIZE LEV ROA ), model(fod) collapse gmm( TOBINSQ , lag(1 .)) gmm( ESGSCORE , lag(1 .)) gmm( SIZE, lag(1 .)) gmm( LEV , lag(1 .)) gmm( ROA , lag(1 .)) teffects two vce(r) overid
xtdpdgmm L(0/2).TOBINSQ L(0/3).( ESGSCORE SIZE ROA ) L(0/2).LEV, model(fod) collapse gmm( TOBINSQ , lag(1 .)) gmm( ESGSCORE , lag(1 .)) gmm ( SIZE , lag(1 .)) gmm( LEV , lag(1 .)) gmm( ROA , lag(1 .)) teffects two vce(r) overid
xtdpdgmm L(0/2).TOBINSQ L(0/3).( ESGSCORE SIZE ROA ) L(0/1).LEV, model(fod) collapse gmm( TOBINSQ , lag(1 .)) gmm( ESGSCORE , lag(1 .)) gmm( SIZE , lag(1 .)) gmm( LEV , lag(1 .)) gmm( ROA , lag(1 .)) teffects two vce(r) overid
xtdpdgmm L(0/2).TOBINSQ L(0/3).( ESGSCORE SIZE) L(0/2).ROA L(0/1).LEV, model(fod) collapse gmm( TOBINSQ , lag(1 .)) gmm( ESGSCORE , lag(1 .)) gmm( SIZE , lag(1 .)) gmm( LEV , lag(1 .)) gmm( ROA , lag(1 .)) teffects two vce(r) overid
xtdpdgmm L(0/2).TOBINSQ L(0/3).( ESGSCORE) L(0/2).(SIZE ROA) L(0/1).LEV, model(fod) collapse gmm( TOBINSQ , lag(1 .)) gmm( ESGSCORE , lag(1 .)) gmm( SIZE , lag(1 .)) gmm( LEV , lag(1 .)) gmm( ROA , lag(1 .)) teffects two vce(r) overid
xtdpdgmm L(0/2).TOBINSQ L(0/3).( ESGSCORE) L(0/2).(ROA) L(0/1).(SIZE LEV), model(fod) collapse gmm( TOBINSQ , lag(1 .)) gmm( ESGSCORE , lag(1 .)) gmm( SIZE , lag(1 .)) gmm( LEV , lag(1 .)) gmm( ROA , lag(1 .)) teffects two vce(r) overid
xtdpdgmm L(0/2).TOBINSQ L(0/3).( ESGSCORE) L(0/2).(ROA) L(0/1).(LEV) SIZE, model(fod) collapse gmm( TOBINSQ , lag(1 .)) gmm( ESGSCORE , lag(1 .)) gmm( SIZE , lag(1 .)) gmm( LEV , lag(1 .)) gmm( ROA , lag(1 .)) teffects two vce(r) overid
xtdpdgmm L(0/2).TOBINSQ L(0/3).( ESGSCORE) L(0/2).(ROA) LEV SIZE, model(fod) collapse gmm( TOBINSQ , lag(1 .)) gmm( ESGSCORE , lag(1 .)) gmm( SIZE , lag(1 .)) gmm( LEV , lag(1 .)) gmm( ROA , lag(1 .)) teffects two vce(r) overid

 estat mmsc model9 model8 model7 model6 model5 model4 model3 model2 model1

Andrews-Lu model and moment selection criteria

       Model | ngroups          J  nmom  npar   MMSC-AIC   MMSC-BIC  MMSC-HQIC
-------------+----------------------------------------------------------------
           . |     440    33.4082    47    18   -24.5918  -143.1082   -72.3942
      model9 |     440    33.4082    47    18   -24.5918  -143.1082   -72.3942
      model8 |     440    29.7748    47    19   -26.2252  -140.6549   -72.3792
      model7 |     440    28.6277    47    20   -25.3723  -135.7152   -69.8780
      model6 |     440    28.6205    47    21   -23.3795  -129.6356   -66.2368
      model5 |     440    25.9056    47    22   -24.0944  -126.2638   -65.3034
      model4 |     440    25.5041    47    23   -22.4959  -120.5785   -62.0565
      model3 |     440    25.6459    47    24   -20.3541  -114.3499   -58.2664
      model2 |     440    26.1705    47    25   -17.8295  -107.7386   -54.0934
      model1 |     440    25.2280    47    26   -16.7720  -102.5943   -51.3876

The same question is valid, should I select the model 9 although it does not have the lowest values for all criterias?

[Sinem Ates]

Dear Sebastian,

Please ignore my question in the post #11. When I use the winsorized variables in my model and apply the same steps, selection of model11 is clear according to the results below. However I have another problem now; when I try to run the underidentication test for the model11, I get the error: "command underid is unrecognized". I could not achieve the installation of this command via help menu.I use Stata/MP 14.2. How can I solve the issue?

Andrews-Lu model and moment selection criteria
 
       Model | ngroups          J  nmom  npar   MMSC-AIC   MMSC-BIC  MMSC-HQIC
-------------+----------------------------------------------------------------
           . |     440    23.2936    31    14   -10.7064   -80.1815   -38.7285
     model13 |     440    23.2936    31    14   -10.7064   -80.1815   -38.7285
     model12 |     440    22.7521    39    15   -25.2479  -123.3305   -64.8085
     model11 |     440    27.3939    47    16   -34.6061  -161.2962   -85.7053
     model10 |     440    25.6474    47    17   -34.3526  -156.9559   -83.8034
      model9 |     440    25.5839    47    18   -32.4161  -150.9326   -80.2186
      model8 |     440    24.8981    47    19   -31.1019  -145.5316   -77.2559
      model7 |     440    22.8425    47    20   -31.1575  -141.5004   -75.6632
      model6 |     440    21.8887    47    21   -30.1113  -136.3674   -72.9686
      model5 |     440    21.2686    47    22   -28.7314  -130.9008   -69.9404
      model4 |     440    20.9577    47    23   -27.0423  -125.1249   -66.6029
      model3 |     440    20.9466    47    24   -25.0534  -119.0492   -62.9656
      model2 |     440    20.7716    47    25   -23.2284  -113.1374   -59.4923
      model1 |     440    20.4736    47    26   -21.5264  -107.3487   -56.1419

[Sebastian Kripfganz]

The underid command is not an official Stata command. It seems that it is not yet publicly released by its author. So, I am afraid you cannot yet use it.

[Sinem Ates]

Thank you. So, could you please check my model below whether the syntax is correct and specification tests are satisfying?

xtdpdgmm L(0/2).TOBINSQ_w  L(0/3).( ESGSCORE ) SIZE_w ROA_w  LEV_w, model(fod) collapse gmm( TOBINSQ_w  , lag(1 .)) gmm( ESGSCORE , lag(1 .)) gmm( SIZE_w  , lag(1 .)) gmm( LEV_w  , lag(1 .)) gmm( ROA_w  , lag(1 .)) teffects two vce(r) overid

Generalized method of moments estimation
 
Fitting full model:
Step 1         f(b) =  .00473743
Step 2         f(b) =  .06225878
 
Fitting reduced model 1:
Step 1         f(b) =  .03609266
 
Fitting reduced model 2:
Step 1         f(b) =  .04614479
 
Fitting reduced model 3:
Step 1         f(b) =  .05093924
 
Fitting reduced model 4:
Step 1         f(b) =  .05709632
 
Fitting reduced model 5:
Step 1         f(b) =  .04657066
 
Fitting reduced model 6:
Step 1         f(b) =   .0505345
 
Group variable: ID                           Number of obs         =      2164
Time variable: YEAR                          Number of groups      =       440
 
Moment conditions:     linear =      47      Obs per group:    min =         1
                    nonlinear =       0                        avg =  4.918182
                        total =      47                        max =         7
 
                                   (Std. Err. adjusted for 440 clusters in ID)
------------------------------------------------------------------------------
             |              WC-Robust
   TOBINSQ_w |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
   TOBINSQ_w |
         L1. |   .8185818   .0891309     9.18   0.000     .6438884    .9932751
         L2. |  -.0972134   .0515191    -1.89   0.059     -.198189    .0037622
             |
    ESGSCORE |
         --. |  -.0111641   .0084997    -1.31   0.189    -.0278231    .0054949
         L1. |   .0055477   .0083977     0.66   0.509    -.0109116     .022007
         L2. |   .0001915   .0015369     0.12   0.901    -.0028208    .0032039
         L3. |   -.002003   .0010463    -1.91   0.056    -.0040537    .0000477
             |
      SIZE_w |  -.0018058    .029179    -0.06   0.951    -.0589956    .0553839
       ROA_w |   .0133053   .0084792     1.57   0.117    -.0033136    .0299243
       LEV_w |   .3625575   .4295074     0.84   0.399    -.4792616    1.204377
             |
        YEAR |
       2013  |  -.0217133   .0317829    -0.68   0.494    -.0840066      .04058
       2014  |   .0509819   .0329594     1.55   0.122    -.0136174    .1155812
       2015  |   .0297979   .0391076     0.76   0.446    -.0468516    .1064475
       2016  |   .0373139   .0400087     0.93   0.351    -.0411017    .1157295
       2017  |   .1008531   .0383089     2.63   0.008      .025769    .1759373
       2018  |   .0255624   .0472693     0.54   0.589    -.0670838    .1182085
             |
       _cons |   .4858946    .490693     0.99   0.322     -.475846    1.447635
------------------------------------------------------------------------------
Instruments corresponding to the linear moment conditions:
 1, model(fodev):
   L1.TOBINSQ_w L2.TOBINSQ_w L3.TOBINSQ_w L4.TOBINSQ_w L5.TOBINSQ_w
   L6.TOBINSQ_w L7.TOBINSQ_w L8.TOBINSQ_w
 2, model(fodev):
   L1.ESGSCORE L2.ESGSCORE L3.ESGSCORE L4.ESGSCORE L5.ESGSCORE L6.ESGSCORE
   L7.ESGSCORE L8.ESGSCORE
 3, model(fodev):
   L1.SIZE_w L2.SIZE_w L3.SIZE_w L4.SIZE_w L5.SIZE_w L6.SIZE_w L7.SIZE_w
   L8.SIZE_w
 4, model(fodev):
   L1.LEV_w L2.LEV_w L3.LEV_w L4.LEV_w L5.LEV_w L6.LEV_w L7.LEV_w L8.LEV_w
 5, model(fodev):
   L1.ROA_w L2.ROA_w L3.ROA_w L4.ROA_w L5.ROA_w L6.ROA_w L7.ROA_w L8.ROA_w
 6, model(level):
   2013bn.YEAR 2014.YEAR 2015.YEAR 2016.YEAR 2017.YEAR 2018.YEAR
 7, model(level):
   _cons

estat serial, ar(1/3)
 
Arellano-Bond test for autocorrelation of the first-differenced residuals
H0: no autocorrelation of order 1:     z =   -5.2641   Prob > |z|  =    0.0000
H0: no autocorrelation of order 2:     z =    0.3312   Prob > |z|  =    0.7405
H0: no autocorrelation of order 3:     z =    0.5569   Prob > |z|  =    0.5776

. estat overid
 
Sargan-Hansen test of the overidentifying restrictions
H0: overidentifying restrictions are valid
 
2-step moment functions, 2-step weighting matrix       chi2(31)    =   27.3939
                                                       Prob > chi2 =    0.6523
 
2-step moment functions, 3-step weighting matrix       chi2(31)    =   29.8561
                                                       Prob > chi2 =    0.5247

. estat overid, difference
 
Sargan-Hansen (difference) test of the overidentifying restrictions
H0: (additional) overidentifying restrictions are valid
 
2-step weighting matrix from full model
 
                  | Excluding                   | Difference                 
Moment conditions |       chi2     df         p |        chi2     df         p
------------------+-----------------------------+-----------------------------
  1, model(fodev) |    15.8808     23    0.8603 |     11.5131      8    0.1743
  2, model(fodev) |    20.3037     23    0.6235 |      7.0902      8    0.5269
  3, model(fodev) |    22.4133     23    0.4954 |      4.9806      8    0.7596
  4, model(fodev) |    25.1224     23    0.3440 |      2.2715      8    0.9715
  5, model(fodev) |    20.4911     23    0.6121 |      6.9028      8    0.5472
  6, model(level) |    22.2352     25    0.6221 |      5.1587      6    0.5236
     model(fodev) |          .     -9         . |           .      .         .

[Sebastian Kripfganz]

The specification looks reasonable indeed and none of the specification tests raises any concern.