Two-step System-GMM with xtabond2 - for growth model

[Gabor Racz]

Dear Statalist Users,

I am trying to do a system GMM estimation for my master thesis where I am trying to calculate the effect of income inequality on growth (with other control variables too) in 36 rich countries between 1970 and 2022 - I would like to ask for some help regarding the xtabond2 syntax (I'm using sata 17).

So the dependent variable is a cumulative 5-year growth of the log of GDP per capita (created by substracting the 5-year lag). The right-hand side of the eq. includes the lagged depvar the gini (gini_disp), gov. expenditures (gov_exp), inflation, average year of schooling (AYoS_15), and trade openness (trade). My first question with this is that if I want to to measure how the variables at the beginning of the 5-year period affect the 5-year cumulative growth, should I create 5-year lags for the variable and do the regression with them? If not, here would be my syntax that I am very unsure of:

 xtabond2 lnGDPpc_gr5 L.lnGDPpc_gr5 gini_disp gov_exp gcf AYoS_15 infl trade i.year i.groups, iv(i.year i.groups) gmm(L.lnGDPpc_gr5 gini_disp gov_exp gcf AYoS_15 infl trade, lag(2 4) collapse) two robust orthogonal

Here I added the lag(2 4) to avoid too many instruments, the orthogonal option is there because I have a lot of gaps in the data. Here I am not sure which of my variables are endogenous, predetermined and exogenous (expect that I know that the year and group dummies are exog. and the lagged depvar is endog.)

If I run this regression, most of my variables are insignificant (which can happen, but I'm more inclined to think that I am doing something wrong) and almost all of the years are ommitted due to collinearity (expect every 5-year like 1980, 1985, 1990, etc.) which is definitely seems wrong.

Here is the output:

 xtabond2 lnGDPpc_gr5 L.lnGDPpc_gr5 gini_disp gov_exp gcf AYoS_15 infl trade i.year i.groups, iv(i.year i.groups) gmm(L.lnGDP
> pc_gr5 gini_disp gov_exp gcf AYoS_15 infl trade, lag(2 4) collapse) two robust orthogonal
Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.
1970b.year dropped due to collinearity
1971.year dropped due to collinearity
1972.year dropped due to collinearity
1973.year dropped due to collinearity
1974.year dropped due to collinearity
1975.year dropped due to collinearity
1976.year dropped due to collinearity
1977.year dropped due to collinearity
1978.year dropped due to collinearity
1979.year dropped due to collinearity
1981.year dropped due to collinearity
1982.year dropped due to collinearity
1983.year dropped due to collinearity
1984.year dropped due to collinearity
1986.year dropped due to collinearity
1987.year dropped due to collinearity
1988.year dropped due to collinearity
1989.year dropped due to collinearity
1991.year dropped due to collinearity
1992.year dropped due to collinearity
1993.year dropped due to collinearity
1994.year dropped due to collinearity
1996.year dropped due to collinearity
1997.year dropped due to collinearity
1998.year dropped due to collinearity
1999.year dropped due to collinearity
2001.year dropped due to collinearity
2002.year dropped due to collinearity
2003.year dropped due to collinearity
2004.year dropped due to collinearity
2006.year dropped due to collinearity
2007.year dropped due to collinearity
2008.year dropped due to collinearity
2009.year dropped due to collinearity
2010.year dropped due to collinearity
2011.year dropped due to collinearity
2012.year dropped due to collinearity
2013.year dropped due to collinearity
2014.year dropped due to collinearity
2015.year dropped due to collinearity
2016.year dropped due to collinearity
2017.year dropped due to collinearity
2018.year dropped due to collinearity
2019.year dropped due to collinearity
2020.year dropped due to collinearity
2021.year dropped due to collinearity
2022.year dropped due to collinearity
1b.groups dropped due to collinearity
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: country_id                      Number of obs      =       176
Time variable : year                            Number of groups   =        34
Number of instruments = 36                      Obs per group: min =         2
Wald chi2(18) =   3870.31                                      avg =      5.18
Prob > chi2   =     0.000                                      max =         7
-----------------------------------------------------------------------------------
                  |              Corrected
      lnGDPpc_gr5 | Coefficient  std. err.      z    P>|z|     [95% conf. interval]
------------------+----------------------------------------------------------------
      lnGDPpc_gr5 |
              L1. |   1.014607   .1403153     7.23   0.000     .7395938     1.28962
                  |
        gini_disp |   .0055179   .0057715     0.96   0.339     -.005794    .0168299
          gov_exp |   .0107178   .0080529     1.33   0.183    -.0050656    .0265012
              gcf |   .0038121   .0024468     1.56   0.119    -.0009836    .0086078
          AYoS_15 |   .0386975   .0229345     1.69   0.092    -.0062533    .0836483
             infl |  -.0000766    .001517    -0.05   0.960      -.00305    .0028967
            trade |   -.000053   .0003429    -0.15   0.877    -.0007251     .000619
                  |
             year |
            1980  |   .1464282   .0715556     2.05   0.041     .0061818    .2866746
            1985  |   .1340342   .0634352     2.11   0.035     .0097034     .258365
            1990  |   .0899732   .0490963     1.83   0.067    -.0062537    .1862001
            1995  |   .0813435   .0458202     1.78   0.076    -.0084624    .1711495
            2000  |   .0766008   .0338248     2.26   0.024     .0103055    .1428961
            2005  |   .0248552   .0197054     1.26   0.207    -.0137668    .0634771
                  |
           groups |
   Mediterranean  |   .0823514   .0590913     1.39   0.163    -.0334655    .1981683
Nordic countries  |   .0153417   .0579169     0.26   0.791    -.0981733    .1288567
   North America  |  -.0622995   .0500158    -1.25   0.213    -.1603286    .0357296
  Post-Socialist  |   .0031748   .0511851     0.06   0.951    -.0971462    .1034959
  Western Europe  |   .0502238    .052768     0.95   0.341    -.0531997    .1536472
                  |
            _cons |  -.9419227   .4032259    -2.34   0.019    -1.732231   -.1516145
-----------------------------------------------------------------------------------
Instruments for orthogonal deviations equation
  Standard
    FOD.(1970b.year 1971.year 1972.year 1973.year 1974.year 1975.year
    1976.year 1977.year 1978.year 1979.year 1980.year 1981.year 1982.year
    1983.year 1984.year 1985.year 1986.year 1987.year 1988.year 1989.year
    1990.year 1991.year 1992.year 1993.year 1994.year 1995.year 1996.year
    1997.year 1998.year 1999.year 2000.year 2001.year 2002.year 2003.year
    2004.year 2005.year 2006.year 2007.year 2008.year 2009.year 2010.year
    2011.year 2012.year 2013.year 2014.year 2015.year 2016.year 2017.year
    2018.year 2019.year 2020.year 2021.year 2022.year 1b.groups 2.groups
    3.groups 4.groups 5.groups 6.groups)
  GMM-type (missing=0, separate instruments for each period unless collapsed)
    L(2/4).(L.lnGDPpc_gr5 gini_disp gov_exp gcf AYoS_15 infl trade) collapsed
Instruments for levels equation
  Standard
    1970b.year 1971.year 1972.year 1973.year 1974.year 1975.year 1976.year
    1977.year 1978.year 1979.year 1980.year 1981.year 1982.year 1983.year
    1984.year 1985.year 1986.year 1987.year 1988.year 1989.year 1990.year
    1991.year 1992.year 1993.year 1994.year 1995.year 1996.year 1997.year
    1998.year 1999.year 2000.year 2001.year 2002.year 2003.year 2004.year
    2005.year 2006.year 2007.year 2008.year 2009.year 2010.year 2011.year
    2012.year 2013.year 2014.year 2015.year 2016.year 2017.year 2018.year
    2019.year 2020.year 2021.year 2022.year 1b.groups 2.groups 3.groups
    4.groups 5.groups 6.groups
    _cons
  GMM-type (missing=0, separate instruments for each period unless collapsed)
    DL.(L.lnGDPpc_gr5 gini_disp gov_exp gcf AYoS_15 infl trade) collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z =      .  Pr > z =      .
Arellano-Bond test for AR(2) in first differences: z =      .  Pr > z =      .
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(17)   =  54.55  Prob > chi2 =  0.000
  (Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(17)   =  23.50  Prob > chi2 =  0.134
  (Robust, but weakened by many instruments.)

Difference-in-Hansen tests of exogeneity of instrument subsets:
  GMM instruments for levels
    Hansen test excluding group:     chi2(11)   =  18.20  Prob > chi2 =  0.077
    Difference (null H = exogenous): chi2(6)    =   5.29  Prob > chi2 =  0.507
  iv(1970b.year 1971.year 1972.year 1973.year 1974.year 1975.year 1976.year 1977.year 1978.year 1979.year 1980.year 1981.year
> 1982.year 1983.year 1984.year 1985.year 1986.year 1987.year 1988.year 1989.year 1990.year 1991.year 1992.year 1993.year 1994
> .year 1995.year 1996.year 1997.year 1998.year 1999.year 2000.year 2001.year 2002.year 2003.year 2004.year 2005.year 2006.yea
> r 2007.year 2008.year 2009.year 2010.year 2011.year 2012.year 2013.year 2014.year 2015.year 2016.year 2017.year 2018.year 20
> 19.year 2020.year 2021.year 2022.year 1b.groups 2.groups 3.groups 4.groups 5.groups 6.groups)
    Hansen test excluding group:     chi2(6)    =  10.21  Prob > chi2 =  0.116
    Difference (null H = exogenous): chi2(11)   =  13.29  Prob > chi2 =  0.275

So all in all, I have three main questions:

1. How do I correctly measure in stata the effect at the beginning of the period (ie. how gini in 1970 affects the cumulative growth of 1975)?
2. Which variables should I put in iv() and gmm()?
3. Why all those years are ommitted and what can I do against it?

Here is an example from the data I use:

* Example generated by -dataex-. For more info, type help dataex
clear
input str14 country long country_id int year double(gini_disp gov_exp gcf AYoS_15 infl trade) float lnGDPpc_gr5
"Australia" 1 1970 26.8     13.17 32.970001   9.7 3.4400001     26.15       .
"Australia" 1 1971 26.5      13.9 32.099998     . 6.1399999 25.549999       .
"Australia" 1 1972 26.2     14.41     30.43     .      6.02     24.76       .
"Australia" 1 1973 25.9     14.61 28.389999     . 9.0900002 25.129999       .
"Australia" 1 1974 25.6     14.45     30.57     .     15.42     26.32       .
"Australia" 1 1975 24.3 16.790001 26.809999 10.52     15.16 28.969999  .05187
"Australia" 1 1976 24.9 17.879999     26.43     .     13.32 26.860001  .06198
"Australia" 1 1977 25.5 17.459999 26.959999     .     12.31 28.629999  .06605
"Australia" 1 1978 26.1     17.83     25.82     .         8     28.18  .05274
"Australia" 1 1979 26.7 17.360001     27.85     . 9.1199999 29.620001  .06671
"Australia" 1 1980 27.3 17.209999     27.09  11.2     10.14 32.299999  .08346
"Australia" 1 1981 27.9 17.809999     28.68     . 9.4899998      31.6  .08515
"Australia" 1 1982 27.4     17.74     29.83     .     11.35     30.33  .07647
"Australia" 1 1983 27.5     18.65 25.299999     .     10.04     29.18  .04303
"Australia" 1 1984 27.5     18.32 26.629999     .      3.96 28.540001  .04702
"Australia" 1 1985 28.5     19.25        28  11.2      6.73 32.490002  .06742
"Australia" 1 1986 28.5     19.43     28.48     . 9.0500002 33.009998  .07463
"Australia" 1 1987 28.9 19.370001     27.35     . 8.5299997 32.509998  .06932
"Australia" 1 1988 29.4 18.459999 27.969999     . 7.2199998     32.57  .14511
"Australia" 1 1989 29.8 17.780001 29.610001     . 7.5300002     32.07  .13332
"Australia" 1 1990 29.2 17.639999 28.940001 11.18 7.3299999 32.150002  .11554
"Australia" 1 1991 29.1 18.629999 24.219999     . 3.1800001 32.189999  .07452
"Australia" 1 1992 29.1      19.5     22.35     .      1.01 33.040001  .05754
"Australia" 1 1993   29 19.379999      23.6     .      1.75 35.400002  .04871
"Australia" 1 1994 28.9 18.809999     24.27     .      1.97 36.459999  .05702
"Australia" 1 1995 30.1 18.639999        26  11.2 4.6300001 37.709999  .06369
"Australia" 1 1996 30.4     18.58 24.790001     . 2.6199999 38.240002  .10597
"Australia" 1 1997 30.7     18.26 24.860001     .       .22     37.98  .14045
"Australia" 1 1998   31     18.26 25.639999     . .86000001 39.990002  .14481
"Australia" 1 1999 31.3     18.67     26.16     .      1.48 39.060001  .15283
"Australia" 1 2000 31.6 18.469999     26.27 11.07      4.46 40.970001  .15266
"Australia" 1 2001 31.7     18.48     23.41     . 4.4099998     44.25  .13432
"Australia" 1 2002 31.7 18.290001     24.42     .      2.98 41.470001  .13455
"Australia" 1 2003 31.7 18.360001 25.950001     .      2.73 40.220001  .11855
"Australia" 1 2004 31.7     18.24     27.08     . 2.3399999 37.029999  .11173
"Australia" 1 2005 32.1     18.32 27.469999 11.38 2.6900001     39.18  .10377
"Australia" 1 2006 32.4     18.24     27.52     . 3.5599999     41.59  .10996
"Australia" 1 2007 32.7 18.110001     27.52     . 2.3299999 42.040001    .101
"Australia" 1 2008   33     18.02 28.610001     . 4.3499999 42.860001  .09688
"Australia" 1 2009   33     18.35 27.370001     .      1.77     45.75  .06419
"Australia" 1 2010   33     18.76 26.790001 11.54 2.9200001     40.52  .05159
"Australia" 1 2011 32.7     18.58 26.459999     .       3.3     41.84  .04778
"Australia" 1 2012 32.5     18.82 27.719999     .      1.76 43.169998  .04977
"Australia" 1 2013 32.6     18.77 27.879999     .      2.45     41.27    .043
"Australia" 1 2014 32.7 18.690001     26.73     .      2.49 42.470001  .05563
"Australia" 1 2015 32.7      19.1 26.280001     .      1.51 41.619999  .05626
"Australia" 1 2016 32.7 19.870001     25.42     .      1.28     40.82  .05784
"Australia" 1 2017 32.7     19.82 24.110001     .      1.95 41.950001  .04306
"Australia" 1 2018 32.7      19.9 24.559999     .      1.91 43.389999   .0482
"Australia" 1 2019 32.7     20.26 23.299999     .      1.61 45.830002  .04428
"Australia" 1 2020 32.6     21.74     22.25     . .85000002     44.23  .02451
"Australia" 1 2021    . 22.299999     22.76     . 2.8599999 39.869999  .03403
"Australia" 1 2022    .         .         .     .      6.59         .       .
"Austria"   2 1970    .     14.32 31.040001   7.4 4.3699999 54.860001       .
"Austria"   2 1971    .     14.39 31.030001     . 4.6999998 54.380001       .
"Austria"   2 1972    .     14.24        32     . 6.3600001 54.299999       .
"Austria"   2 1973    .     14.69 32.240002     . 7.5300002 54.459999       .
"Austria"   2 1974    .     15.35     32.52     . 9.5200005     59.57       .
"Austria"   2 1975    . 16.790001     27.16   7.7 8.4499998 56.599998 .178015
"Austria"   2 1976    .     17.18     28.51     . 7.3200002 59.919998 .179131
end
label values country_id country_id2
label def country_id2 1 "AUS", modify
label def country_id2 2 "AUT", modify

Thank you very much for your help in advance!

[Sebastian Kripfganz]

You variable AYoS_15 has values recorded only every 5 years, which explains why all the interim time dummies are dropped.

For more on dynamic panel GMM estimation, see the following presentation and the references therein:

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

At the end of the presentation, you can also find an illustration of a procedure for classifying your regressors as strictly exogenous, predetermined, or endogenous.

[Gabor Racz]

Dear Sebastian,

Thank you for your answer! Can this problem be solved by different model specification (different lags, different instrument used, etc.)? According to the literature, this metric is widely used to estimate human capital formation which is important to the model.

Thank you in advance!

[Sebastian Kripfganz]

If you do not want to lose the observations in between these 5-years interval, you need to interpolate the missing values of your variable.

[Gabor Racz]

Dear Sebastian,

Thank you for your answer! Interpolating did solve the problems of dropped years but I'm still very unsure of my model specification and code. After interpolating the schooling data, I run the following code:

  xtabond2 gr5 L.gr5 L5.(gini_disp gov_exp ip_sch_15 gcf infl trade) i.year i.groups, ///
    iv(i.year i.groups, equation(level)) ///
    gmm(L.gr5, lag(3 4) collapse) ///
    gmm(L5.(gini_disp gov_exp ip_sch_15 gcf infl trade), lag(2 3) collapse) ///
    twostep robust orthogonal small

But it produces weird results. The 5-year lags are important to the model, everything else I am higly unsure of. (I use higher lags for instrumenting, otherwise the lagged dependent variable sometimes gets ommitted from the model). The result of the Hansen test also doesn't seem to be too convincing as I suppose it shouldn't really be exactly 1.000

  xtabond2 gr5 L.gr5 L5.(gini_disp gov_exp ip_sch_15 gcf infl trade) i.year i.groups, ///
>         iv(i.year i.groups, equation(level)) ///
>         gmm(L.gr5, lag(3 4) collapse) ///
>         gmm(L5.(gini_disp gov_exp ip_sch_15 gcf infl trade), lag(2 3) collapse) ///
>         twostep robust orthogonal small
Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.
1970b.year dropped due to collinearity
1971.year dropped due to collinearity
1972.year dropped due to collinearity
1973.year dropped due to collinearity
1974.year dropped due to collinearity
1975.year dropped due to collinearity
2015.year dropped due to collinearity
2016.year dropped due to collinearity
2017.year dropped due to collinearity
2018.year dropped due to collinearity
2019.year dropped due to collinearity
2020.year dropped due to collinearity
2021.year dropped due to collinearity
2022.year dropped due to collinearity
1b.groups dropped due to collinearity
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: country_id                      Number of obs      =      1042
Time variable : year                            Number of groups   =        36
Number of instruments = 66                      Obs per group: min =        12
F(51, 35)     =   2580.16                                      avg =     28.94
Prob > F      =     0.000                                      max =        40
-----------------------------------------------------------------------------------
                  |              Corrected
              gr5 | Coefficient  std. err.      t    P>|t|     [95% conf. interval]
------------------+----------------------------------------------------------------
              gr5 |
              L1. |   1.145021   .2013876     5.69   0.000      .736183     1.55386
                  |
        gini_disp |
              L5. |   -.002623   .0029595    -0.89   0.381    -.0086311    .0033851
                  |
          gov_exp |
              L5. |   .0029712      .0021     1.41   0.166     -.001292    .0072343
                  |
        ip_sch_15 |
              L5. |   .0012795   .0067917     0.19   0.852    -.0125084    .0150675
                  |
              gcf |
              L5. |   .0002879   .0015819     0.18   0.857    -.0029236    .0034994
                  |
             infl |
              L5. |   .0000369   .0002025     0.18   0.856    -.0003741    .0004479
                  |
            trade |
              L5. |   .0004803   .0003651     1.32   0.197    -.0002609    .0012214
                  |
             year |
            1976  |   -.011436   .0216351    -0.53   0.600    -.0553576    .0324856
            1977  |          0  (omitted)
            1978  |          0  (omitted)
            1979  |          0  (omitted)
            1980  |  -.0645869    .064948    -0.99   0.327    -.1964383    .0672645
            1981  |   -.016445   .0796844    -0.21   0.838    -.1782128    .1453229
            1982  |  -.0334431   .0205171    -1.63   0.112     -.075095    .0082087
            1983  |          0  (omitted)
            1984  |          0  (omitted)
            1985  |          0  (omitted)
            1986  |          0  (omitted)
            1987  |          0  (omitted)
            1988  |  -.0049377   .0231373    -0.21   0.832    -.0519089    .0420335
            1989  |          0  (omitted)
            1990  |   -.187855   .1225121    -1.53   0.134    -.4365677    .0608578
            1991  |  -.0518657   .0277375    -1.87   0.070    -.1081759    .0044445
            1992  |  -.1902306   .0941287    -2.02   0.051    -.3813221    .0008608
            1993  |          0  (omitted)
            1994  |          0  (omitted)
            1995  |  -.0153175   .0142215    -1.08   0.289    -.0441886    .0135537
            1996  |   .0031573   .0153127     0.21   0.838    -.0279292    .0342437
            1997  |   .0037137   .0593979     0.06   0.951    -.1168705    .1242979
            1998  |  -.0032725   .0194306    -0.17   0.867    -.0427187    .0361736
            1999  |   .2288079   .0868057     2.64   0.012      .052583    .4050327
            2000  |   -.121513    .118709    -1.02   0.313     -.362505    .1194789
            2001  |  -.0061415    .022372    -0.27   0.785    -.0515591     .039276
            2002  |  -.0446254    .020591    -2.17   0.037    -.0864274   -.0028234
            2003  |  -.2795529    .171137    -1.63   0.111    -.6269796    .0678737
            2004  |  -.0258896   .0191072    -1.35   0.184    -.0646793    .0129002
            2005  |          0  (omitted)
            2006  |  -.0190221   .0183614    -1.04   0.307    -.0562977    .0182534
            2007  |          0  (omitted)
            2008  |  -.0560815   .0226772    -2.47   0.018    -.1021186   -.0100444
            2009  |  -.1252733   .0199084    -6.29   0.000    -.1656895   -.0848572
            2010  |  -.0414425   .0166508    -2.49   0.018    -.0752454   -.0076397
            2011  |          0  (omitted)
            2012  |   .1358538   .1031792     1.32   0.197    -.0736111    .3453188
            2013  |  -.0183525    .027428    -0.67   0.508    -.0740344    .0373293
            2014  |  -.0898898   .0621802    -1.45   0.157    -.2161224    .0363427
                  |
           groups |
   Mediterranean  |  -.0068462   .0400306    -0.17   0.865    -.0881128    .0744203
Nordic countries  |  -.0464442   .0246133    -1.89   0.067    -.0964118    .0035234
   North America  |          0  (omitted)
  Post-Socialist  |  -.0488101   .0244393    -2.00   0.054    -.0984246    .0008043
  Western Europe  |  -.0361439   .0301651    -1.20   0.239    -.0973824    .0250946
                  |
            _cons |          0  (omitted)
-----------------------------------------------------------------------------------
Instruments for orthogonal deviations equation
  GMM-type (missing=0, separate instruments for each period unless collapsed)
    L(2/3).(L5.gini_disp L5.gov_exp L5.ip_sch_15 L5.gcf L5.infl L5.trade)
    collapsed
    L(3/4).L.gr5 collapsed
Instruments for levels equation
  Standard
    1970b.year 1971.year 1972.year 1973.year 1974.year 1975.year 1976.year
    1977.year 1978.year 1979.year 1980.year 1981.year 1982.year 1983.year
    1984.year 1985.year 1986.year 1987.year 1988.year 1989.year 1990.year
    1991.year 1992.year 1993.year 1994.year 1995.year 1996.year 1997.year
    1998.year 1999.year 2000.year 2001.year 2002.year 2003.year 2004.year
    2005.year 2006.year 2007.year 2008.year 2009.year 2010.year 2011.year
    2012.year 2013.year 2014.year 2015.year 2016.year 2017.year 2018.year
    2019.year 2020.year 2021.year 2022.year 1b.groups 2.groups 3.groups
    4.groups 5.groups 6.groups
    _cons
  GMM-type (missing=0, separate instruments for each period unless collapsed)
    DL.(L5.gini_disp L5.gov_exp L5.ip_sch_15 L5.gcf L5.infl L5.trade)
    collapsed
    DL2.L.gr5 collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z =  -2.72  Pr > z =  0.007
Arellano-Bond test for AR(2) in first differences: z =  -1.36  Pr > z =  0.174
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(14)   =  73.49  Prob > chi2 =  0.000
  (Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(14)   =   0.00  Prob > chi2 =  1.000
  (Robust, but weakened by many instruments.)

At this point, I still not sure if I specify the model correctly. Should I include eq() suboptions to the gmmstyle, should I put them in one bracket using the same instruments for all the non-dummy variables?

Generally, as my T is pretty large (with the N too), do you think that I should switch using other approaches like xtivreg?

Thank you in advance!

[Sebastian Kripfganz]

Your N is very small (36), especially relative to T. These types of estimators do not work very well in such a situation. Switching to a simpler estimation approach would be indicated here.