xtabond2 : Hansen test of overid. restrictions

Sinem Ates

Join Date: Mar 2018
Posts: 83

#46

25 Mar 2020, 11:09

Dear Sebastian,
I hope you are fine in these difficult days. I would really appreciate if you could answer my question below:
The only difference between the 2 models below is the number of lags of variable SIZE_w. When I decrease the number of lag of SIZE_w from 3 to 2, the incremental overidentificaton tests for exogeneous instruments (ICBIC represent INDUSTRY) produce unsatisfactory results. How can I explain this situation while interpreting my results?

Code:

xtdpdgmm L(0/2).TOBINSQ_w ESG CAPEXSALES_w  L(0/3).SIZE_w  ROA_w LEV_w i.ICBIC , model(fod) collapse gmm( TOBINSQ_w  , lag(1 .)) gmm(ESG , 
> lag(1.)) gmm( ROA_w  ,lag(1 .))gmm( SIZE_w  ,lag(1 .)) gmm( LEV_w  , lag(1 .)) gmm( CAPEXSALES_w  , lag(1 .)) gmm(LEV_w SIZE_w ROA_w CAPEXS
> ALES_w ESG, lag(0 0)) iv(i.ICBIC, model(level))  teffects two vce(r) overid

Generalized method of moments estimation

Fitting full model:
Step 1         f(b) =  .01303283
Step 2         f(b) =  .10188116

Fitting reduced model 1:
Step 1         f(b) =  .07870622

Fitting reduced model 2:
Step 1         f(b) =  .09621081

Fitting reduced model 3:
Step 1         f(b) =  .08887677

Fitting reduced model 4:
Step 1         f(b) =  .09052209

Fitting reduced model 5:
Step 1         f(b) =  .09879314

Fitting reduced model 6:
Step 1         f(b) =  .07764848

Fitting reduced model 7:
Step 1         f(b) =  .09606716

Fitting reduced model 8:
Step 1         f(b) =  .08450391

Fitting reduced model 9:
Step 1         f(b) =  .08450391

Fitting no-level model:
Step 1         f(b) =  .08450391

Group variable: ID                           Number of obs         =      2002
Time variable: YEAR                          Number of groups      =       413

Moment conditions:     linear =      70      Obs per group:    min =         1
                    nonlinear =       0                        avg =  4.847458
                        total =      70                        max =         7

                                   (Std. Err. adjusted for 413 clusters in ID)
------------------------------------------------------------------------------
             |              WC-Robust
   TOBINSQ_w |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
   TOBINSQ_w |
         L1. |   .6874987    .074449     9.23   0.000     .5415812    .8334161
         L2. |  -.0839133   .0426733    -1.97   0.049    -.1675514   -.0002752
             |
         ESG |  -.0058779   .0024605    -2.39   0.017    -.0107004   -.0010555
CAPEXSALES_w |  -.0069368   .0032667    -2.12   0.034    -.0133395   -.0005342
             |
      SIZE_w |
         --. |   -.071505   .0689636    -1.04   0.300    -.2066712    .0636612
         L1. |   .0819699   .0629519     1.30   0.193    -.0414136    .2053534
         L2. |  -.0250898   .0451815    -0.56   0.579    -.1136439    .0634643
         L3. |   .0293345   .0430134     0.68   0.495    -.0549703    .1136393
             |
       ROA_w |   .0239917   .0058741     4.08   0.000     .0124787    .0355047
       LEV_w |   .2045407   .3936637     0.52   0.603     -.567026    .9761075
             |
       ICBIC |
         15  |  -.2333013   .2420391    -0.96   0.335    -.7076891    .2410866
         20  |   -.057852    .203594    -0.28   0.776    -.4568888    .3411849
         30  |  -.3756351   .2549801    -1.47   0.141    -.8753869    .1241167
         35  |  -.3308295    .264687    -1.25   0.211    -.8496066    .1879475
         40  |  -.0721328   .1996534    -0.36   0.718    -.4634463    .3191808
         45  |   .2810171   .2315749     1.21   0.225    -.1728615    .7348957
         50  |  -.2654361   .2397228    -1.11   0.268    -.7352841    .2044119
         55  |   -.223023   .2460639    -0.91   0.365    -.7052995    .2592534
         60  |  -.3237489   .2395391    -1.35   0.177     -.793237    .1457392
         65  |  -.3710317   .2522621    -1.47   0.141    -.8654564    .1233929
             |
        YEAR |
       2013  |  -.0355795   .0313526    -1.13   0.256    -.0970295    .0258704
       2014  |   .0415234   .0332967     1.25   0.212     -.023737    .1067837
       2015  |    .017638    .037172     0.47   0.635    -.0552178    .0904939
       2016  |   .0042694   .0354168     0.12   0.904    -.0651463    .0736851
       2017  |   .0565129   .0345025     1.64   0.101    -.0111107    .1241366
       2018  |  -.0040439   .0366809    -0.11   0.912    -.0759371    .0678493
             |
       _cons |   .6553736   .5429982     1.21   0.227    -.4088834    1.719631
------------------------------------------------------------------------------
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.ESG L2.ESG L3.ESG L4.ESG L5.ESG L6.ESG L7.ESG L8.ESG
 3, 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
 4, 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
 5, 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
 6, model(fodev):
   L1.CAPEXSALES_w L2.CAPEXSALES_w L3.CAPEXSALES_w L4.CAPEXSALES_w
   L5.CAPEXSALES_w L6.CAPEXSALES_w L7.CAPEXSALES_w L8.CAPEXSALES_w
 7, model(fodev):
   LEV_w SIZE_w ROA_w CAPEXSALES_w ESG
 8, model(level):
   15bn.ICBIC 20.ICBIC 30.ICBIC 35.ICBIC 40.ICBIC 45.ICBIC 50.ICBIC 55.ICBIC
   60.ICBIC 65.ICBIC
 9, model(level):
   2013bn.YEAR 2014.YEAR 2015.YEAR 2016.YEAR 2017.YEAR 2018.YEAR
 10, 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 =   -3.7580   Prob > |z|  =    0.0002
H0: no autocorrelation of order 2:     z =   -1.0592   Prob > |z|  =    0.2895
H0: no autocorrelation of order 3:     z =    0.6213   Prob > |z|  =    0.5344

. estat overid

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

2-step moment functions, 2-step weighting matrix       chi2(43)    =   42.0769
                                                       Prob > chi2 =    0.5112

2-step moment functions, 3-step weighting matrix       chi2(43)    =   43.5511
                                                       Prob > chi2 =    0.4479

. 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) |    32.5057     35    0.5891 |      9.5713      8    0.2964
  2, model(fodev) |    39.7351     35    0.2672 |      2.3419      8    0.9687
  3, model(fodev) |    36.7061     35    0.3897 |      5.3708      8    0.7173
  4, model(fodev) |    37.3856     35    0.3601 |      4.6913      8    0.7900
  5, model(fodev) |    40.8016     35    0.2305 |      1.2754      8    0.9958
  6, model(fodev) |    32.0688     35    0.6104 |     10.0081      8    0.2645
  7, model(fodev) |    39.6757     38    0.3952 |      2.4012      5    0.7913
  8, model(level) |    34.9001     37    0.5679 |      7.1768      6    0.3048
  9, model(level) |    34.9001     37    0.5679 |      7.1768      6    0.3048
     model(fodev) |          .    -10         . |           .      .         .
     model(level) |    34.9001     27    0.1414 |      7.1768     16    0.9697

. xtdpdgmm L(0/2).TOBINSQ_w ESG CAPEXSALES_w  L(0/2).SIZE_w  ROA_w LEV_w i.ICBIC , model(fod) collapse gmm( TOBINSQ_w  , lag(1 .)) gmm(ESG , 
> lag(1.)) gmm( ROA_w  ,lag(1 .))gmm( SIZE_w  ,lag(1 .)) gmm( LEV_w  , lag(1 .)) gmm( CAPEXSALES_w  , lag(1 .)) gmm(LEV_w SIZE_w ROA_w CAPEXS
> ALES_w ESG, lag(0 0)) iv(i.ICBIC, model(level))  teffects two vce(r) overid

Generalized method of moments estimation

Fitting full model:
Step 1         f(b) =  .01556283
Step 2         f(b) =  .12316875

Fitting reduced model 1:
Step 1         f(b) =  .10200775

Fitting reduced model 2:
Step 1         f(b) =  .11302198

Fitting reduced model 3:
Step 1         f(b) =  .10304694

Fitting reduced model 4:
Step 1         f(b) =  .10207448

Fitting reduced model 5:
Step 1         f(b) =  .11697899

Fitting reduced model 6:
Step 1         f(b) =  .09436803

Fitting reduced model 7:
Step 1         f(b) =  .11246203

Fitting reduced model 8:
Step 1         f(b) =   .0839411

Fitting reduced model 9:
Step 1         f(b) =   .0839411

Fitting no-level model:
Step 1         f(b) =   .0839411

Group variable: ID                           Number of obs         =      2429
Time variable: YEAR                          Number of groups      =       427

Moment conditions:     linear =      71      Obs per group:    min =         1
                    nonlinear =       0                        avg =  5.688525
                        total =      71                        max =         8

                                   (Std. Err. adjusted for 427 clusters in ID)
------------------------------------------------------------------------------
             |              WC-Robust
   TOBINSQ_w |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
   TOBINSQ_w |
         L1. |   .6661253   .0799538     8.33   0.000     .5094187    .8228318
         L2. |   -.060883   .0515144    -1.18   0.237    -.1618495    .0400835
             |
         ESG |  -.0037064   .0018913    -1.96   0.050    -.0074132    4.24e-07
CAPEXSALES_w |  -.0032306   .0027834    -1.16   0.246    -.0086859    .0022247
             |
      SIZE_w |
         --. |  -.0764026     .05956    -1.28   0.200     -.193138    .0403329
         L1. |   .0397292   .0570908     0.70   0.486    -.0721667    .1516251
         L2. |   .0226691   .0403963     0.56   0.575    -.0565062    .1018444
             |
       ROA_w |   .0266766   .0056555     4.72   0.000     .0155921    .0377612
       LEV_w |   .6162228   .3148211     1.96   0.050    -.0008152    1.233261
             |
       ICBIC |
         15  |   -.347145   .2236941    -1.55   0.121    -.7855774    .0912874
         20  |  -.0545094   .1791532    -0.30   0.761    -.4056431    .2966244
         30  |  -.4764755   .2200233    -2.17   0.030    -.9077132   -.0452377
         35  |  -.4591035   .2486139    -1.85   0.065    -.9463777    .0281707
         40  |  -.1824917   .1771518    -1.03   0.303    -.5297029    .1647195
         45  |   .1921836   .2145475     0.90   0.370    -.2283218     .612689
         50  |  -.3859718   .2165313    -1.78   0.075    -.8103653    .0384217
         55  |   -.335951   .2277819    -1.47   0.140    -.7823952    .1104933
         60  |  -.3697299   .2278514    -1.62   0.105    -.8163104    .0768506
         65  |  -.4428622   .2398108    -1.85   0.065    -.9128828    .0271584
             |
        YEAR |
       2012  |   .1369935   .0342361     4.00   0.000     .0698921    .2040949
       2013  |    .058375   .0339537     1.72   0.086     -.008173     .124923
       2014  |   .1311551   .0380246     3.45   0.001     .0566283    .2056819
       2015  |   .1222051   .0349892     3.49   0.000     .0536275    .1907827
       2016  |    .094505   .0343991     2.75   0.006      .027084    .1619261
       2017  |   .1535793   .0355352     4.32   0.000     .0839315    .2232271
       2018  |   .0885713     .03502     2.53   0.011     .0199334    .1572092
             |
       _cons |    .692059   .5476086     1.26   0.206     -.381234    1.765352
------------------------------------------------------------------------------
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.ESG L2.ESG L3.ESG L4.ESG L5.ESG L6.ESG L7.ESG L8.ESG
 3, 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
 4, 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
 5, 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
 6, model(fodev):
   L1.CAPEXSALES_w L2.CAPEXSALES_w L3.CAPEXSALES_w L4.CAPEXSALES_w
   L5.CAPEXSALES_w L6.CAPEXSALES_w L7.CAPEXSALES_w L8.CAPEXSALES_w
 7, model(fodev):
   LEV_w SIZE_w ROA_w CAPEXSALES_w ESG
 8, model(level):
   15bn.ICBIC 20.ICBIC 30.ICBIC 35.ICBIC 40.ICBIC 45.ICBIC 50.ICBIC 55.ICBIC
   60.ICBIC 65.ICBIC
 9, model(level):
   2012bn.YEAR 2013.YEAR 2014.YEAR 2015.YEAR 2016.YEAR 2017.YEAR 2018.YEAR
 10, 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 =   -4.1569   Prob > |z|  =    0.0000
H0: no autocorrelation of order 2:     z =   -1.2273   Prob > |z|  =    0.2197
H0: no autocorrelation of order 3:     z =   -1.0013   Prob > |z|  =    0.3167

. estat overid

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

2-step moment functions, 2-step weighting matrix       chi2(44)    =   52.5931
                                                       Prob > chi2 =    0.1756

2-step moment functions, 3-step weighting matrix       chi2(44)    =   55.8861
                                                       Prob > chi2 =    0.1079

. 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) |    43.5573     36    0.1808 |      9.0357      8    0.3393
  2, model(fodev) |    48.2604     36    0.0832 |      4.3327      8    0.8259
  3, model(fodev) |    44.0010     36    0.1690 |      8.5920      8    0.3779
  4, model(fodev) |    43.5858     36    0.1800 |      9.0073      8    0.3417
  5, model(fodev) |    49.9500     36    0.0610 |      2.6430      8    0.9547
  6, model(fodev) |    40.2951     36    0.2859 |     12.2979      8    0.1384
  7, model(fodev) |    48.0213     39    0.1524 |      4.5718      5    0.4703
  8, model(level) |    35.8429     37    0.5232 |     16.7502      7    0.0191
  9, model(level) |    35.8429     37    0.5232 |     16.7502      7    0.0191
     model(fodev) |          .     -9         . |           .      .         .
     model(level) |    35.8429     27    0.1188 |     16.7502     17    0.4714

Comment

Sebastian Kripfganz

Join Date: May 2014

Posts: 2576
#47

25 Mar 2020, 12:57

Hard to say. In principle, L3.Size_w could be seen as an omitted variable in the second regression. If it is correlated with those level instruments, the latter could become invalid. However, in the first regression the coefficient of L3.Size_w is highly statistically insignificant such that omitting this regressor should not matter.

The other effect that removing L3.Size_w has is an increase of the sample size (1 time period more per unit). This makes it difficult to directly compare the two regressions. It becomes highly speculative what exactly causes the differences in the incremental Hansen tests. With one additional time period, it might become more difficult to reliably estimate the weighting matrix. On the other side, more observations should increase the overall efficiency of the model estimation. There could also be some outliers in the extra observations.

Given that the moment conditions 8 and 9 refer to time and industry dummy variables, I would not worry too much about those incremental tests, as we would usually treat those dummies as exogenous and most of the time do not want to interpret their coefficients. (They are just controls.)

https://www.kripfganz.de/stata/
Comment
Sinem Ates

Join Date: Mar 2018

Posts: 83
#48

25 Mar 2020, 13:32

Dear Sebastian,

Thank you very much for your fast reply.
Comment

Sinem Ates

Join Date: Mar 2018
Posts: 83

#49

25 Mar 2020, 23:53

Dear Sebastian,

I have a similar question as in post #47. This time I removed the L3.ESG from the first model below, this caused unsatisfactory result for moment condition 4. Does this mean that the second model produces unreliable results?

Code:

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

Generalized method of moments estimation

Fitting full model:
Step 1         f(b) =  .00581307
Step 2         f(b) =  .06913361

Fitting reduced model 1:
Step 1         f(b) =  .04596722

Fitting reduced model 2:
Step 1         f(b) =  .05467238

Fitting reduced model 3:
Step 1         f(b) =  .05680316

Fitting reduced model 4:
Step 1         f(b) =  .05645681

Fitting reduced model 5:
Step 1         f(b) =   .0654722

Fitting reduced model 6:
Step 1         f(b) =  .05698076

Fitting reduced model 7:
Step 1         f(b) =  .06513794

Group variable: ID                           Number of obs         =      1966
Time variable: YEAR                          Number of groups      =       406

Moment conditions:     linear =      55      Obs per group:    min =         1
                    nonlinear =       0                        avg =  4.842365
                        total =      55                        max =         7

                                   (Std. Err. adjusted for 406 clusters in ID)
------------------------------------------------------------------------------
             |              WC-Robust
   TOBINSQ_w |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
   TOBINSQ_w |
         L1. |   .8368863   .0945805     8.85   0.000      .651512    1.022261
         L2. |  -.0870895   .0596461    -1.46   0.144    -.2039938    .0298148
             |
         ESG |
         --. |  -.0089572   .0083547    -1.07   0.284     -.025332    .0074177
         L1. |   .0011553   .0085646     0.13   0.893    -.0156309    .0179415
         L2. |  -.0007376   .0016784    -0.44   0.660    -.0040272     .002552
         L3. |  -.0013731   .0011262    -1.22   0.223    -.0035805    .0008343
             |
      SIZE_w |
         --. |  -.2488051    .188912    -1.32   0.188    -.6190659    .1214556
         L1. |   .2744836   .1822084     1.51   0.132    -.0826384    .6316056
             |
CAPEXSALES_w |
         --. |   .0024956   .0067776     0.37   0.713    -.0107882    .0157794
         L1. |  -.0059505   .0037306    -1.60   0.111    -.0132623    .0013613
             |
       ROA_w |   .0100532   .0106795     0.94   0.347    -.0108782    .0309846
       LEV_w |   .2016963   .4635453     0.44   0.663    -.7068357    1.110228
             |
        YEAR |
       2013  |  -.0391881   .0440057    -0.89   0.373    -.1254376    .0470614
       2014  |   .0608903   .0421915     1.44   0.149    -.0218036    .1435841
       2015  |   .0338224   .0442332     0.76   0.444    -.0528731     .120518
       2016  |   .0532811   .0441278     1.21   0.227    -.0332078      .13977
       2017  |    .086227   .0448253     1.92   0.054    -.0016291     .174083
       2018  |   .0397917   .0570502     0.70   0.485    -.0720247    .1516081
             |
       _cons |   .3072037   .5593658     0.55   0.583    -.7891331     1.40354
------------------------------------------------------------------------------
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.ESG L2.ESG L3.ESG L4.ESG L5.ESG L6.ESG L7.ESG L8.ESG
 3, 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
 4, 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
 5, 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
 6, model(fodev):
   L1.CAPEXSALES_w L2.CAPEXSALES_w L3.CAPEXSALES_w L4.CAPEXSALES_w
   L5.CAPEXSALES_w L6.CAPEXSALES_w L7.CAPEXSALES_w L8.CAPEXSALES_w
 7, model(level):
   2013bn.YEAR 2014.YEAR 2015.YEAR 2016.YEAR 2017.YEAR 2018.YEAR
 8, 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 =   -4.9375   Prob > |z|  =    0.0000
H0: no autocorrelation of order 2:     z =    0.5237   Prob > |z|  =    0.6005
H0: no autocorrelation of order 3:     z =    0.0839   Prob > |z|  =    0.9332

. estat overid

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

2-step moment functions, 2-step weighting matrix       chi2(36)    =   28.0682
                                                       Prob > chi2 =    0.8248

2-step moment functions, 3-step weighting matrix       chi2(36)    =   32.8432
                                                       Prob > chi2 =    0.6195

. 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) |    18.6627     28    0.9082 |      9.4056      8    0.3092
  2, model(fodev) |    22.1970     28    0.7721 |      5.8713      8    0.6617
  3, model(fodev) |    23.0621     28    0.7299 |      5.0062      8    0.7569
  4, model(fodev) |    22.9215     28    0.7370 |      5.1468      8    0.7418
  5, model(fodev) |    26.5817     28    0.5411 |      1.4865      8    0.9929
  6, model(fodev) |    23.1342     28    0.7263 |      4.9341      8    0.7646
  7, model(level) |    26.4460     30    0.6522 |      1.6222      6    0.9510
     model(fodev) |          .    -12         . |           .      .         .

Code:

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

Generalized method of moments estimation

Fitting full model:
Step 1         f(b) =  .00887583
Step 2         f(b) =   .1079143

Fitting reduced model 1:
Step 1         f(b) =  .08823401

Fitting reduced model 2:
Step 1         f(b) =  .08209872

Fitting reduced model 3:
Step 1         f(b) =  .08999702

Fitting reduced model 4:
Step 1         f(b) =   .0761198

Fitting reduced model 5:
Step 1         f(b) =  .09922221

Fitting reduced model 6:
Step 1         f(b) =  .08624098

Fitting reduced model 7:
Step 1         f(b) =  .08769461

Group variable: ID                           Number of obs         =      2399
Time variable: YEAR                          Number of groups      =       424

Moment conditions:     linear =      56      Obs per group:    min =         1
                    nonlinear =       0                        avg =  5.658019
                        total =      56                        max =         8

                                   (Std. Err. adjusted for 424 clusters in ID)
------------------------------------------------------------------------------
             |              WC-Robust
   TOBINSQ_w |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
   TOBINSQ_w |
         L1. |   .8018932   .0896627     8.94   0.000     .6261576    .9776288
         L2. |   -.028647   .0589736    -0.49   0.627    -.1442331    .0869391
             |
         ESG |
         --. |   .0057379   .0095117     0.60   0.546    -.0129048    .0243806
         L1. |  -.0094316   .0092781    -1.02   0.309    -.0276163    .0087531
         L2. |  -.0018889   .0017897    -1.06   0.291    -.0053967    .0016189
             |
      SIZE_w |
         --. |  -.4341186   .2528541    -1.72   0.086    -.9297034    .0614662
         L1. |   .4359773   .2516762     1.73   0.083     -.057299    .9292536
             |
CAPEXSALES_w |
         --. |   .0000857    .006617     0.01   0.990    -.0128834    .0130548
         L1. |  -.0047081   .0041305    -1.14   0.254    -.0128037    .0033876
             |
       ROA_w |   .0125186   .0102974     1.22   0.224    -.0076639     .032701
       LEV_w |   .5072269   .4085789     1.24   0.214    -.2935731    1.308027
             |
        YEAR |
       2012  |   .1527758   .0422686     3.61   0.000     .0699309    .2356208
       2013  |   .0551001   .0539448     1.02   0.307    -.0506298    .1608299
       2014  |   .1995252   .0533505     3.74   0.000     .0949601    .3040903
       2015  |   .1275927   .0451641     2.83   0.005     .0390727    .2161126
       2016  |   .1379098   .0462403     2.98   0.003     .0472804    .2285392
       2017  |   .1648767   .0477703     3.45   0.001     .0712487    .2585048
       2018  |   .1308579   .0554476     2.36   0.018     .0221826    .2395333
             |
       _cons |   .1107581   .5192834     0.21   0.831    -.9070187    1.128535
------------------------------------------------------------------------------
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.ESG L2.ESG L3.ESG L4.ESG L5.ESG L6.ESG L7.ESG L8.ESG
 3, 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
 4, 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
 5, 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
 6, model(fodev):
   L1.CAPEXSALES_w L2.CAPEXSALES_w L3.CAPEXSALES_w L4.CAPEXSALES_w
   L5.CAPEXSALES_w L6.CAPEXSALES_w L7.CAPEXSALES_w L8.CAPEXSALES_w
 7, model(level):
   2012bn.YEAR 2013.YEAR 2014.YEAR 2015.YEAR 2016.YEAR 2017.YEAR 2018.YEAR
 8, 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 =   -4.8169   Prob > |z|  =    0.0000
H0: no autocorrelation of order 2:     z =   -1.0429   Prob > |z|  =    0.2970
H0: no autocorrelation of order 3:     z =   -0.8830   Prob > |z|  =    0.3772

. estat overid

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

2-step moment functions, 2-step weighting matrix       chi2(37)    =   45.7557
                                                       Prob > chi2 =    0.1531

2-step moment functions, 3-step weighting matrix       chi2(37)    =   45.4456
                                                       Prob > chi2 =    0.1606

. 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) |    37.4112     29    0.1360 |      8.3444      8    0.4006
  2, model(fodev) |    34.8099     29    0.2109 |     10.9458      8    0.2048
  3, model(fodev) |    38.1587     29    0.1188 |      7.5969      8    0.4738
  4, model(fodev) |    32.2748     29    0.3079 |     13.4809      8    0.0963
  5, model(fodev) |    42.0702     29    0.0553 |      3.6854      8    0.8843
  6, model(fodev) |    36.5662     29    0.1576 |      9.1895      8    0.3266
  7, model(level) |    37.1825     30    0.1719 |      8.5731      7    0.2848
     model(fodev) |          .    -11         . |           .      .         .

Comment

Sebastian Kripfganz

Join Date: May 2014

Posts: 2576
#50

26 Mar 2020, 06:43

I am afraid I cannot say more than in my previous post.

https://www.kripfganz.de/stata/
Comment

lal mohan kumar

Join Date: May 2019
Posts: 265

#51

27 Dec 2020, 09:57

Dear Stata members
I am trying with system GMM and based on some videos and stataforums, I tried the following code

Code:

xi i.year
xtabond2 dividends_ta_w L.dividends_ta_w ncfo_ta_w avgq_w  avgsg_w avgsize_w avg_lev_w avgoppro_w logage_w, gmm(l.dividends_ta_w , lag(5 6)) ///
iv(ncfo_ta_w avgsg_w avgsize_w avg_lev_w avgoppro_w logage_w _Iyear_2004-_Iyear_2019)
 
. xtabond2 dividends_ta_w L.dividends_ta_w ncfo_ta_w avgq_w  avgsg_w avgsize_w avg_lev_w avgoppro_w logage_w, gmm(l.dividends_ta_w , lag(5
> 6)) ///
> iv(ncfo_ta_w avgsg_w avgsize_w avg_lev_w avgoppro_w logage_w _Iyear_2004-_Iyear_2019)
Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.

Dynamic panel-data estimation, one-step system GMM
------------------------------------------------------------------------------
Group variable: id                              Number of obs      =     13323
Time variable : year                            Number of groups   =      1867
Number of instruments = 42                      Obs per group: min =         1
Wald chi2(8)  =   6980.13                                      avg =      7.14
Prob > chi2   =     0.000                                      max =        13
--------------------------------------------------------------------------------
dividends_ta_w |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------+----------------------------------------------------------------
dividends_ta_w |
           L1. |   .4294056   .0293894    14.61   0.000     .3718034    .4870078
               |
     ncfo_ta_w |   .0149341   .0012993    11.49   0.000     .0123874    .0174807
        avgq_w |     -.0051   .0008365    -6.10   0.000    -.0067394   -.0034605
       avgsg_w |   .0145548   .0010516    13.84   0.000     .0124936    .0166159
     avgsize_w |   .0011807   .0001133    10.42   0.000     .0009587    .0014028
     avg_lev_w |  -.0270888   .0016501   -16.42   0.000     -.030323   -.0238547
    avgoppro_w |   .0117715   .0011511    10.23   0.000     .0095153    .0140277
      logage_w |   .0006755   .0002689     2.51   0.012     .0001486    .0012025
         _cons |   .0055626   .0010242     5.43   0.000     .0035552    .0075699
--------------------------------------------------------------------------------
Instruments for first differences equation
  Standard
    D.(ncfo_ta_w avgsg_w avgsize_w avg_lev_w avgoppro_w logage_w _Iyear_2004
    _Iyear_2005 _Iyear_2006 _Iyear_2007 _Iyear_2008 _Iyear_2009 _Iyear_2010
    _Iyear_2011 _Iyear_2012 _Iyear_2013 _Iyear_2014 _Iyear_2015 _Iyear_2016
    _Iyear_2017 _Iyear_2018 _Iyear_2019)
  GMM-type (missing=0, separate instruments for each period unless collapsed)
    L(5/6).L.dividends_ta_w
Instruments for levels equation
  Standard
    ncfo_ta_w avgsg_w avgsize_w avg_lev_w avgoppro_w logage_w _Iyear_2004
    _Iyear_2005 _Iyear_2006 _Iyear_2007 _Iyear_2008 _Iyear_2009 _Iyear_2010
    _Iyear_2011 _Iyear_2012 _Iyear_2013 _Iyear_2014 _Iyear_2015 _Iyear_2016
    _Iyear_2017 _Iyear_2018 _Iyear_2019
    _cons
  GMM-type (missing=0, separate instruments for each period unless collapsed)
    DL4.L.dividends_ta_w
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -24.15  Pr > z =  0.000
Arellano-Bond test for AR(2) in first differences: z =  10.59  Pr > z =  0.000
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(33)   = 637.94  Prob > chi2 =  0.000
  (Not robust, but not weakened by many instruments.)

Difference-in-Sargan tests of exogeneity of instrument subsets:
  GMM instruments for levels
    Sargan test excluding group:     chi2(25)   = 604.51  Prob > chi2 =  0.000
    Difference (null H = exogenous): chi2(8)    =  33.43  Prob > chi2 =  0.000
  gmm(L.dividends_ta_w, lag(5 6))
    Sargan test excluding group:     chi2(10)   = 411.56  Prob > chi2 =  0.000
    Difference (null H = exogenous): chi2(23)   = 226.37  Prob > chi2 =  0.000
  iv(ncfo_ta_w avgsg_w avgsize_w avg_lev_w avgoppro_w logage_w _Iyear_2004 _Iyear_2005 _Iyear_2006 _Iyear_2007 _Iyear_2008 _Iyear_2009 _Iye
> ar_2010 _Iyear_2011 _Iyear_2012 _Iyear_2013 _Iyear_2014 _Iyear_2015 _Iyear_2016 _Iyear_2017 _Iyear_2018 _Iyear_2019)
    Sargan test excluding group:     chi2(14)   = 193.14  Prob > chi2 =  0.000
    Difference (null H = exogenous): chi2(19)   = 444.79  Prob > chi2 =  0.000

I realized that there is Autocorrelation and My instruments failed sargan tests. I don't think it is advisable to go for more lags. Also how do I qualify for Sargan tests for instruments.
Sebastian Kripfganz. Sorry for marking you. But I would like to get your comments in this if possible. Also I have seen people playing with gmm & iv part

Code:

gmm(l.dividends_ta_w , lag(5 6)) ///
iv(ncfo_ta_w avgsg_w avgsize_w avg_lev_w avgoppro_w logage_w _Iyear_2004-_Iyear_2019)

by using different specifications. Are there any strict guidelines? I know it is better to learn GMM from scratch but given the paucity of time are there any quick references that are very practical

Comment

Eric de Souza

Join Date: Mar 2014
Posts: 587

#52

27 Dec 2020, 11:59

On Edit: the post below should be ignored (I think). I thought I saw a post saying that -underid- had not yet been released but cannot find it.
-ssc describe underid- gives the following information:

Code:

      Requires: Stata version 13.1 and ranktest from SSC (q.v.)
      
      Distribution-Date: 20200929
      
      Author: Mark E Schaffer, Heriot-Watt University
      Support: email [email protected]
      
      Author: Frank Windmeijer, University of Bristol
      Support: email [email protected]

Last edited by Eric de Souza; 27 Dec 2020, 12:05.

Comment

Sebastian Kripfganz

Join Date: May 2014

Posts: 2576
#53

28 Dec 2020, 06:47

lal mohan kumar
With your large number of observations, specification tests might already detect small deviations from the null hypothesis. Opting for deeper lags to address a rejection of the Arellano-Bond serial correlation test is usually not an ideal approach. Instead, it might be more promising to directly account for the serial correlation by adding further lags of the dependent variable and lags of the independent variables to the model. See for example:
Kiviet, J. F. (2020). Microeconometric dynamic panel data methods: Model specification and selection issues. Econometrics and Statistics 13, 16-45.

Also, for all variables in the iv() option, you are essentially assuming that they are not only strictly exogenous with respect to the idiosyncratic error term but also uncorrelated with the unit-specific effects. This is effectively a random-effects assumption which might be too strong in many applications.

A useful quick reference might be my 2019 London Stata Conference presentation:
Kripfganz, S. (2019). Generalized method of moments estimation of linear dynamic panel data models. Proceedings of the 2019 London Stata Conference.

https://www.kripfganz.de/stata/
Comment
lal mohan kumar

Join Date: May 2019

Posts: 265
#54

28 Dec 2020, 09:32

Thanks @Sebastian Kripfganz. Will try to learn those and then apply
Comment

Announcement

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment