Hi

I run two step system GMM and the assumption that all the variable is exogenous; but might not strictly exogenous. Below is the output :

Code:
 xtabond2 gw L.gw L(0/1).( short_term long_term transient int_pct_tasset  tobinq yr_capex_growth b_seg  m_share
> _vol) event_num, gmm(L.(gw short_term long_term transient int_pct_tasset  tobinq yr_capex_growth b_seg  m_shar
> e_vol )) iv(event_num*, eq(level)) h(2) twostep pca
Favoring speed over space. To switch, type or click on mata: mata set matafavor space, 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.
  Difference-in-Sargan/Hansen statistics may be negative.

Dynamic panel-data estimation, two-step system GMM
------------------------------------------------------------------------------
Group variable: cid                             Number of obs      =       246
Time variable : event_num                       Number of groups   =        87
Number of instruments = 91                      Obs per group: min =      1201
Wald chi2(18) =  4.55e+06                                      avg =      2.83
Prob > chi2   =     0.000                                      max =      1201
---------------------------------------------------------------------------------
             gw |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
----------------+----------------------------------------------------------------
             gw |
            L1. |   .1489766   .0726771     2.05   0.040     .0065321    .2914212
                |
     short_term |
            --. |   10.31974   2.198471     4.69   0.000     6.010813    14.62866
            L1. |  -11.32196    2.33351    -4.85   0.000    -15.89556   -6.748367
                |
      long_term |
            --. |  -3.954428   2.070992    -1.91   0.056    -8.013499    .1046425
            L1. |   6.883481   3.553929     1.94   0.053    -.0820916    13.84905
                |
      transient |
            --. |   2.642767   3.810419     0.69   0.488    -4.825517    10.11105
            L1. |  -8.908993   4.557095    -1.95   0.051    -17.84074    .0227495
                |
 int_pct_tasset |
            --. |  -16.26055    17.4643    -0.93   0.352    -50.48995    17.96885
            L1. |   66.33505   16.10121     4.12   0.000     34.77727    97.89283
                |
         tobinq |
            --. |   .1755169   .3994388     0.44   0.660    -.6073688    .9584027
            L1. |   4.303243   1.165217     3.69   0.000      2.01946    6.587026
                |
yr_capex_growth |
            --. |  -.0600049   .0286807    -2.09   0.036     -.116218   -.0037917
            L1. |   .0196104   .0278989     0.70   0.482    -.0350705    .0742913
                |
          b_seg |
            --. |  -1.941495   1.435445    -1.35   0.176    -4.754915    .8719252
            L1. |   2.402056   1.219698     1.97   0.049     .0114919    4.792621
                |
    m_share_vol |
            --. |   .6679478   .1381052     4.84   0.000     .3972665     .938629
            L1. |   .1939754   .0791757     2.45   0.014     .0387939    .3491568
                |
      event_num |   .8060752   .3453696     2.33   0.020     .1291633    1.482987
          _cons |  -28.22342   11.88914    -2.37   0.018    -51.52571   -4.921134
---------------------------------------------------------------------------------
Warning: Uncorrected two-step standard errors are unreliable.

Instruments for first differences equation
  GMM-type (missing=0, separate instruments for each period unless collapsed)
    L(1/16).(L.gw L.short_term L.long_term L.transient L.int_pct_tasset
    L.tobinq L.yr_capex_growth L.b_seg L.m_share_vol)
Instruments for levels equation
  Standard
    event_num
    _cons
  GMM-type (missing=0, separate instruments for each period unless collapsed)
    D.(L.gw L.short_term L.long_term L.transient L.int_pct_tasset L.tobinq
    L.yr_capex_growth L.b_seg L.m_share_vol)
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z =  -3.06  Pr > z =  0.002
Arellano-Bond test for AR(2) in first differences: z =  -1.03  Pr > z =  0.302
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(72)   =  48.64  Prob > chi2 =  0.984
  (Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(72)   =  18.55  Prob > chi2 =  1.000
  (Robust, but weakened by many instruments.)
------------------------------------------------------------------------------
Extracted 89 principal components from GMM-style instruments
  Portion of variance explained by the components =  0.978
  Kaiser-Meyer-Olkin measure of sampling adequacy =  0.997
Firstly I do not have great understanding about GMM; but Hansen and Sargan statistics looks like problem here. Moreover, my original observation was 402 with 157 groups (unbalanced large N>T small). But the observation drop significantly. From my understanding as I am taking lag(0/1); so system might dropping the first observation of every group. However, there are lot of groups which have only one observation. So, dropping those observation might cause the reduction of the group number as well. Now, my question is that; statistically how accepted this model is??


Also I run a different model with same data; except where I did add the industry control variables. Below is the output:

Code:
 xtdpdgmm L(0/1).gw active_agg passive_agg  int_pct_tasset  tobinq yr_capex_growth b_seg  m_share_vol consumer_
> discretionary consumer_staples energy financial healthcare industrial information_technology materials real_es
> tate utilities, noserial gmmiv(L.gw, collapse model(difference)) iv( active_agg passive_agg int_pct_tasset  to
> binq yr_capex_growth b_seg  m_share_vol consumer_discretionary consumer_staples energy financial healthcare in
> dustrial information_technology materials real_estate utilities, difference model(difference)) twostep vce(rob
> ust)

Generalized method of moments estimation

Fitting full model:

Step 1:
initial:       f(b) =  44594.685
alternative:   f(b) =  43710.162
rescale:       f(b) =  5488.4395
Iteration 0:   f(b) =  5488.4395  
Iteration 1:   f(b) =  2723.4145  
Iteration 2:   f(b) =  116.75556  
Iteration 3:   f(b) =  85.806709  
Iteration 4:   f(b) =  84.388052  
Iteration 5:   f(b) =  82.676703  
Iteration 6:   f(b) =  82.606986  
Iteration 7:   f(b) =  82.589215  
Iteration 8:   f(b) =  82.584455  
Iteration 9:   f(b) =   82.58285  
Iteration 10:  f(b) =  82.582289  
Iteration 11:  f(b) =  82.582079  
Iteration 12:  f(b) =  82.581999  
Iteration 13:  f(b) =  82.581968  
Iteration 14:  f(b) =  82.581956  
Iteration 15:  f(b) =  82.581951  

Step 2:
Iteration 0:   f(b) =  .29690248  
Iteration 1:   f(b) =   .2426051  
Iteration 2:   f(b) =  .21019636  
Iteration 3:   f(b) =  .19037489  
Iteration 4:   f(b) =  .15706538  
Iteration 5:   f(b) =  .13085923  
Iteration 6:   f(b) =  .09782899  
Iteration 7:   f(b) =  .06635696  
Iteration 8:   f(b) =  .06635696  

Group variable: cid                          Number of obs         =       245
Time variable: event_num                     Number of groups      =        87

Moment conditions:     linear =      26      Obs per group:    min =         1
                    nonlinear =      13                        avg =  2.816092
                        total =      39                        max =        15

                                             (Std. Err. adjusted for 87 clusters in cid)
----------------------------------------------------------------------------------------
                       |              WC-Robust
                    gw |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-----------------------+----------------------------------------------------------------
                    gw |
                   L1. |  -.0080355   .2565298    -0.03   0.975    -.5108247    .4947537
                       |
            active_agg |    .378974   .2267992     1.67   0.095    -.0655443    .8234924
           passive_agg |  -1.489526   .8017774    -1.86   0.063    -3.060981    .0819289
        int_pct_tasset |  -101.3735   70.78429    -1.43   0.152    -240.1082    37.36117
                tobinq |   -.284594   .4127688    -0.69   0.491    -1.093606    .5244179
       yr_capex_growth |   -.033862   .0554779    -0.61   0.542    -.1425967    .0748728
                 b_seg |  -1.892156   2.286801    -0.83   0.408    -6.374203    2.589891
           m_share_vol |   .1213346   .0724229     1.68   0.094    -.0206116    .2632809
consumer_discretionary |  -426.5232   368.1016    -1.16   0.247    -1147.989    294.9427
      consumer_staples |  -512.5687    341.666    -1.50   0.134    -1182.222    157.0845
                energy |   -523.647   349.6281    -1.50   0.134    -1208.906    161.6116
             financial |   -271.938   304.6259    -0.89   0.372    -868.9937    325.1178
            healthcare |   -256.926   297.4885    -0.86   0.388    -839.9928    326.1409
            industrial |  -327.7467   646.1796    -0.51   0.612    -1594.235     938.742
information_technology |  -171.3303   131.7261    -1.30   0.193    -429.5087      86.848
             materials |   -534.384    587.235    -0.91   0.363    -1685.343    616.5753
           real_estate |  -348.9409   698.4176    -0.50   0.617    -1717.814    1019.932
             utilities |  -327.5032   691.5641    -0.47   0.636    -1682.944    1027.938
                 _cons |   434.0517   419.3341     1.04   0.301    -387.8279    1255.931
----------------------------------------------------------------------------------------
Instruments corresponding to the linear moment conditions:
 1, model(diff):
   L1.L.gw L2.L.gw L3.L.gw L4.L.gw L5.L.gw L6.L.gw L7.L.gw L8.L.gw L9.L.gw
   L10.L.gw L11.L.gw L12.L.gw L13.L.gw L14.L.gw
 2, model(diff):
   D.active_agg D.passive_agg D.int_pct_tasset D.tobinq D.yr_capex_growth
   D.b_seg D.m_share_vol D.consumer_discretionary D.consumer_staples
   D.financial D.real_estate
 3, model(level):
   _cons

. //porstestimation of GMM
. estat overid

Sargan-Hansen test of the overidentifying restrictions
H0: overidentifying restrictions are valid

2-step moment functions, 2-step weighting matrix       chi2(20)    =    5.7731
                                                       Prob > chi2 =    0.9992

2-step moment functions, 3-step weighting matrix       chi2(20)    =   27.0385
                                                       Prob > chi2 =    0.1342

. estat mmsc

Andrews-Lu model and moment selection criteria

       Model | ngroups          J  nmom  npar   MMSC-AIC   MMSC-BIC  MMSC-HQIC
-------------+----------------------------------------------------------------
           . |      87     5.7731    39    19   -34.2269   -83.5451   -54.6844

.
So, in both model the same number of observation was dropped. is it a matter of concern?? Also same as Sargan and Hansen statistics look suspicious. Can anyone please advise, where is the problem??

Thanks