Hello Statalist
I have some problems with System GMM estimating.
The model be used is production model (dependent variable: lny=ln(sales), independent variables: ln(labor), lnk=ln(tangible capital), lno=ln(intangible capital), additional dummy variable is year*dummy.) Data be used is panel data with T（year）=23， N（id)=637
I used xtabond2 to estimate this model, however the output shows both AR(1) and AR(2) =0, Sargan test and Hansen test maybe have problems also.
The command I used is：xtabond2 lny lnl lnk lno year1992-year2012, gmm(lny lnl lnk lno, lag(2 .)) iv(year1992-year2012, eq(level)) robust
To be compared with System GMM's output, I put Fixed-Effect output here also.
I already red Roodman, D. 2009. How to Do xtabond2: An Introduction to "Difference" and "System" GMM in Stata but I still don't know how to decide instrument variables when I use System GMM. I really want to know how should I set instrument variables for this model or what kind of command I should use.
Thanks in advance.
I have some problems with System GMM estimating.
The model be used is production model (dependent variable: lny=ln(sales), independent variables: ln(labor), lnk=ln(tangible capital), lno=ln(intangible capital), additional dummy variable is year*dummy.) Data be used is panel data with T（year）=23， N（id)=637
I used xtabond2 to estimate this model, however the output shows both AR(1) and AR(2) =0, Sargan test and Hansen test maybe have problems also.
The command I used is：xtabond2 lny lnl lnk lno year1992-year2012, gmm(lny lnl lnk lno, lag(2 .)) iv(year1992-year2012, eq(level)) robust
| . xtabond2 lny lnl lnk lno year1992-year2012, gmm(lny lnl lnk lno, lag(2 .)) iv( |
| > year1992-year2012, eq(level)) robust |
| Favoring space over speed. To switch, type or click on mata: mata set matafavor |
| > speed, 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 robust weighting matrix for Hansen te |
| > st. |
| Difference-in-Sargan/Hansen statistics may be negative. |
| Dynamic panel-data estimation, one-step system GMM |
| Group variable: id Number of obs = 14651 |
| Time variable : year Number of groups = 637 |
| Number of instruments = 1030 Obs per group: min = 23 |
| Wald chi2(24) = 1.23e+06 avg = 23.00 |
| Prob > chi2 = 0.000 max = 23 |
| Robust |
| lny Coef. Std. Err. z P>z [95% Conf. Interval] |
| lnl .2311044 .063762 3.62 0.000 .1061332 .3560756 |
| lnk .3739435 .0431675 8.66 0.000 .2893367 .4585503 |
| lno .3977854 .0547142 7.27 0.000 .2905476 .5050232 |
| year1992 -.0420709 .0096513 -4.36 0.000 -.0609871 -.0231547 |
| year1993 -.1062366 .013575 -7.83 0.000 -.1328431 -.0796301 |
| year1994 -.1591675 .0151422 -10.51 0.000 -.1888456 -.1294894 |
| year1995 -.1438602 .0157278 -9.15 0.000 -.174686 -.1130343 |
| year1996 -.116888 .0173766 -6.73 0.000 -.1509454 -.0828305 |
| year1997 -.0839585 .0199827 -4.20 0.000 -.1231239 -.044793 |
| year1998 -.0706454 .0229383 -3.08 0.002 -.1156037 -.0256871 |
| year1999 -.1364316 .0249504 -5.47 0.000 -.1853334 -.0875297 |
| year2000 -.1282056 .0262582 -4.88 0.000 -.1796707 -.0767405 |
| year2001 -.0637555 .0284657 -2.24 0.025 -.1195473 -.0079636 |
| year2002 -.111371 .0312501 -3.56 0.000 -.1726201 -.0501219 |
| year2003 -.0865753 .0327983 -2.64 0.008 -.1508588 -.0222919 |
| year2004 -.05774 .032974 -1.75 0.080 -.1223678 .0068879 |
| year2005 -.0248328 .0337512 -0.74 0.462 -.090984 .0413183 |
| year2006 -.0146908 .0336846 -0.44 0.663 -.0807114 .0513298 |
| year2007 .009961 .034357 0.29 0.772 -.0573775 .0772995 |
| year2008 .0225851 .0363989 0.62 0.535 -.0487555 .0939256 |
| year2009 -.022184 .0384784 -0.58 0.564 -.0976002 .0532322 |
| year2010 -.1658045 .036336 -4.56 0.000 -.2370217 -.0945873 |
| year2011 -.10225 .0355019 -2.88 0.004 -.1718324 -.0326676 |
| year2012 4.546705 .036277 125.33 0.000 4.475604 4.617807 |
| _cons -.1015089 .4752674 -0.21 0.831 -1.033016 .829998 |
| Instruments for first differences equation |
| GMM-type (missing=0, separate instruments for each period unless collapsed) |
| L(2/22).(lny lnl lnk lno) |
| Instruments for levels equation |
| Standard |
| year1992 year1993 year1994 year1995 year1996 year1997 year1998 year1999 |
| year2000 year2001 year2002 year2003 year2004 year2005 year2006 year2007 |
| year2008 year2009 year2010 year2011 year2012 |
| _cons |
| GMM-type (missing=0, separate instruments for each period unless collapsed) |
| DL.(lny lnl lnk lno) |
| Arellano-Bond test for AR(1) in first differences: z = -2.96 Pr > z = 0.003 |
| Arellano-Bond test for AR(2) in first differences: z = -6.60 Pr > z = 0.000 |
| Sargan test of overid. restrictions: chi2(1005) =83847.66 Prob > chi2 = 0.000 |
| (Not robust, but not weakened by many instruments.) |
| Hansen test of overid. restrictions: chi2(1005) = 623.09 Prob > chi2 = 1.000 |
| (Robust, but weakened by many instruments.) |
| Difference-in-Hansen tests of exogeneity of instrument subsets: |
| GMM instruments for levels |
| Hansen test excluding group: chi2(921) = 628.22 Prob > chi2 = 1.000 |
| Difference (null H = exogenous): chi2(84) = -5.13 Prob > chi2 = 1.000 |
| iv(year1992 year1993 year1994 year1995 year1996 year1997 year1998 year1999 yea |
| > r2000 year2001 year2002 year2003 year2004 year2005 year2006 year2007 year2008 |
| > year2009 year2010 year2011 year2012, eq(level)) |
| Hansen test excluding group: chi2(984) = 625.84 Prob > chi2 = 1.000 |
| Difference (null H = exogenous): chi2(21) = -2.75 Prob > chi2 = 1.000 |
To be compared with System GMM's output, I put Fixed-Effect output here also.
| . xtreg lny lnl lnk lno year1991-year2012, fe |
| Fixed-effects (within) regression Number of obs = 14,651 |
| Group variable: id Number of groups = 637 |
| R-sq: Obs per group: |
| within = 0.9709 min = 23 |
| between = 0.9371 avg = 23.0 |
| overall = 0.9439 max = 23 |
| F(25,13989) = 18681.89 |
| corr(u_i, Xb) = 0.3448 Prob > F = 0.0000 |
| lny Coef. Std. Err. t P>t [95% Conf. Interval] |
| lnl .2645839 .0099081 26.70 0.000 .2451627 .284005 |
| lnk .1353052 .0074146 18.25 0.000 .1207715 .1498389 |
| lno .4795949 .0102243 46.91 0.000 .4595539 .4996359 |
| year1991 .0326442 .0094812 3.44 0.001 .0140599 .0512286 |
| year1992 .0121168 .009631 1.26 0.208 -.0067611 .0309948 |
| year1993 -.0378018 .0097561 -3.87 0.000 -.056925 -.0186786 |
| year1994 -.082716 .0098257 -8.42 0.000 -.1019757 -.0634564 |
| year1995 -.0642514 .0098764 -6.51 0.000 -.0836105 -.0448924 |
| year1996 -.0323739 .0099728 -3.25 0.001 -.0519219 -.012826 |
| year1997 .0065888 .0101065 0.65 0.514 -.0132214 .0263989 |
| year1998 .0268512 .0102659 2.62 0.009 .0067286 .0469738 |
| year1999 -.0323679 .0103545 -3.13 0.002 -.052664 -.0120717 |
| year2000 -.0220441 .0104418 -2.11 0.035 -.0425115 -.0015767 |
| year2001 .0457242 .0106252 4.30 0.000 .0248974 .066551 |
| year2002 .0002308 .0107995 0.02 0.983 -.0209378 .0213993 |
| year2003 .0239889 .0108953 2.20 0.028 .0026327 .0453451 |
| year2004 .0499187 .0109443 4.56 0.000 .0284664 .0713709 |
| year2005 .08174 .0110057 7.43 0.000 .0601673 .1033126 |
| year2006 .093895 .0110233 8.52 0.000 .0722879 .1155021 |
| year2007 .1219463 .0110828 11.00 0.000 .1002225 .1436701 |
| year2008 .1397135 .0112724 12.39 0.000 .1176182 .1618089 |
| year2009 .100775 .0114269 8.82 0.000 .0783768 .1231732 |
| year2010 -.0422209 .0112401 -3.76 0.000 -.064253 -.0201887 |
| year2011 .0202788 .0112097 1.81 0.070 -.0016937 .0422513 |
| year2012 4.668229 .0112566 414.71 0.000 4.646165 4.690294 |
| _cons .8495015 .1085321 7.83 0.000 .6367641 1.062239 |
| sigma_u .39135151 |
| sigma_e .16828035 |
| rho .84395432 (fraction of variance due to u_i) |
| F test that all u_i=0: F(636, 13989) = 89.31 Prob > F = 0.0000 |
I already red Roodman, D. 2009. How to Do xtabond2: An Introduction to "Difference" and "System" GMM in Stata but I still don't know how to decide instrument variables when I use System GMM. I really want to know how should I set instrument variables for this model or what kind of command I should use.
Thanks in advance.
