xtabond2 omitting indep. variable

[Giorgio Carlini]

Hello, I'm using xtabond2 command to estimate TFP growth determinants at regional level.

This is the code

xtabond2 gtfp l(1).gtfp l(1).( c.dettotoccrfls##c.dettotoccrfls netais PSIis netaes PSIes corrups branchvas rspub rspri spesatotvasn) tau1995-tau2012 if cod_reg<21 , ///
iv(l(1).( c.dettotoccrfls##c.dettotoccrfls netais PSIis netaes PSIes corrups branchvas rspub rspri spesatotvasn) tau1995-tau2012, eq(diff)) ///
iv(l(1).( c.dettotoccrfls##c.dettotoccrfls netais PSIis netaes PSIes corrups branchvas rrspub spri spesatotvasn) tau1995-tau2012 , eq(lev)) ///
gmm(l(1).gtfp , laglimits(2 3) collapse eq(diff) ) ///
gmm(l(1).gtfp , laglimits(1 2) collapse eq(lev) ) ///
h(3) rob ar(3) two

And this is the result:
Dynamic panel-data estimation, two-step system GMM
------------------------------------------------------------------------------
Group variable: cod_reg Number of obs = 302
Time variable : anno Number of groups = 20
Number of instruments = 47 Obs per group: min = 8
Wald chi2(27) = 5884.43 avg = 15.10
Prob > chi2 = 0.000 max = 16
---------------------------------------------------------------------------------------------------
| Corrected
gtfp | Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------------------------+----------------------------------------------------------------
gtfp |
L1. | -.3486568 .2051498 -1.70 0.089 -.750743 .0534294
|
dettotoccrfls |
L1. | -.2904224 .0420268 -6.91 0.000 -.3727935 -.2080514
|
cL.dettotoccrfls#cL.dettotoccrfls | .0012302 .0001987 6.19 0.000 .0008408 .0016197
|
netais |
L1. | -.1445761 .0552673 -2.62 0.009 -.2528981 -.0362542
|
PSIis |
L1. | .0871704 .0501783 1.74 0.082 -.0111772 .185518
|
netaes |
L1. | -.0161285 .1417915 -0.11 0.909 -.2940347 .2617777
|
PSIes |
L1. | .1127257 .1227786 0.92 0.359 -.1279158 .3533673
|
corrups |
L1. | -.1038076 .0363532 -2.86 0.004 -.1750585 -.0325566
|
branchvas |
L1. | .3335042 .0761757 4.38 0.000 .1842024 .4828059
|
rspub |
L1. | 8.881163 2.170696 4.09 0.000 4.626676 13.13565
|
rspri |
L1. | 2.137914 1.485923 1.44 0.150 -.7744416 5.050269
|
spesatotvasn |
L1. | -.001382 .0030787 -0.45 0.653 -.0074161 .004652
|
tau1997 | 0 (omitted)
tau1998 | -3.219385 1.161367 -2.77 0.006 -5.495622 -.9431483
tau1999 | 0 (omitted)
tau2000 | 1.56857 .2815227 5.57 0.000 1.016795 2.120344
tau2001 | 0 (omitted)
tau2002 | .5246916 .9521124 0.55 0.582 -1.341414 2.390798
tau2003 | -2.179011 2.134557 -1.02 0.307 -6.362665 2.004643
tau2004 | -.8417027 1.577904 -0.53 0.594 -3.934337 2.250931
tau2005 | 0 (omitted)
tau2006 | 0 (omitted)
tau2007 | .0393123 1.647896 0.02 0.981 -3.190505 3.269129
tau2008 | -5.877881 2.511571 -2.34 0.019 -10.80047 -.9552915
tau2009 | 0 (omitted)
tau2010 | 0 (omitted)
tau2011 | 2.930558 1.428196 2.05 0.040 .1313464 5.72977
_cons | 0 (omitted)
---------------------------------------------------------------------------------------------------
Instruments for first differences equation
Standard
D.(L.dettotoccrfls cL.dettotoccrfls#cL.dettotoccrfls L.netais L.PSIis
L.netaes L.PSIes L.corrups L.branchvas L.rspub L.rspri L.spesatotvasn
tau1995 tau1996 tau1997 tau1998 tau1999 tau2000 tau2001 tau2002 tau2003
tau2004 tau2005 tau2006 tau2007 tau2008 tau2009 tau2010 tau2011 tau2012)
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(2/3).L.gtfp collapsed
Instruments for levels equation
Standard
L.dettotoccrfls cL.dettotoccrfls#cL.dettotoccrfls L.netais L.PSIis
L.netaes L.PSIes L.corrups L.branchvas L.rspub L.rspri L.spesatotvasn
tau1995 tau1996 tau1997 tau1998 tau1999 tau2000 tau2001 tau2002 tau2003
tau2004 tau2005 tau2006 tau2007 tau2008 tau2009 tau2010 tau2011 tau2012
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
DL(1/2).L.gtfp collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -2.29 Pr > z = 0.022
Arellano-Bond test for AR(2) in first differences: z = 1.02 Pr > z = 0.309
Arellano-Bond test for AR(3) in first differences: z = -0.93 Pr > z = 0.353
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(19) = 81.27 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(19) = 0.00 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(17) = 0.00 Prob > chi2 = 1.000
Difference (null H = exogenous): chi2(2) = -0.00 Prob > chi2 = 1.000
gmm(L.gtfp, collapse eq(diff) lag(2 3))
Hansen test excluding group: chi2(17) = 0.00 Prob > chi2 = 1.000
Difference (null H = exogenous): chi2(2) = -0.00 Prob > chi2 = 1.000
gmm(L.gtfp, collapse eq(level) lag(1 2))
Hansen test excluding group: chi2(17) = 0.00 Prob > chi2 = 1.000
Difference (null H = exogenous): chi2(2) = -0.00 Prob > chi2 = 1.000
iv(L.dettotoccrfls cL.dettotoccrfls#cL.dettotoccrfls L.netais L.PSIis L.netaes L.PSIes L.corrups L.branchvas L.rspub L.rspri L.spesatotvasn t
> au1995 tau1996 tau1997 tau1998 tau1999 tau2000 tau2001 tau2002 tau2003 tau2004 tau2005 tau2006 tau2007 tau2008 tau2009 tau2010 tau2011 tau201
> 2, eq(level))
Hansen test excluding group: chi2(0) = 0.00 Prob > chi2 = .
Difference (null H = exogenous): chi2(19) = -0.00 Prob > chi2 = 1.000




However, when I drop the variable rspub I got this result



Dynamic panel-data estimation, two-step system GMM
------------------------------------------------------------------------------
Group variable: cod_reg Number of obs = 312
Time variable : anno Number of groups = 20
Number of instruments = 43 Obs per group: min = 11
Wald chi2(26) = 7280.04 avg = 15.60
Prob > chi2 = 0.000 max = 16
---------------------------------------------------------------------------------------------------
| Corrected
gtfp | Coef. Std. Err. z P>|z| [95% Conf. Interval]
----------------------------------+----------------------------------------------------------------
gtfp |
L1. | -.0509799 .7195486 -0.07 0.944 -1.461269 1.359309
|
dettotoccrfls |
L1. | -.0758702 .0272724 -2.78 0.005 -.1293232 -.0224172
|
cL.dettotoccrfls#cL.dettotoccrfls | .0002493 .000101 2.47 0.014 .0000515 .0004472
|
netais |
L1. | -.1384859 .1186769 -1.17 0.243 -.3710885 .0941166
|
PSIis |
L1. | .1137223 .1094074 1.04 0.299 -.1007123 .3281568
|
netaes |
L1. | .2859073 .0985538 2.90 0.004 .0927455 .4790692
|
PSIes |
L1. | -.2711722 .0839869 -3.23 0.001 -.4357835 -.1065609
|
corrups |
L1. | -.0133743 .0566096 -0.24 0.813 -.124327 .0975784
|
branchvas |
L1. | .0588094 .0292706 2.01 0.045 .00144 .1161788
|
rspri |
L1. | 0 (omitted)
|
spesatotvasn |
L1. | .0059124 .0031741 1.86 0.063 -.0003087 .0121336
|
tau1997 | 0 (omitted)
tau1998 | -2.429709 2.815155 -0.86 0.388 -7.947311 3.087894
tau1999 | -3.935108 2.373173 -1.66 0.097 -8.586441 .7162255
tau2000 | -1.189491 2.323914 -0.51 0.609 -5.744278 3.365297
tau2001 | 1.805959 .7105831 2.54 0.011 .4132417 3.198676
tau2002 | 0 (omitted)
tau2003 | 0 (omitted)
tau2004 | 1.323694 1.462585 0.91 0.365 -1.54292 4.190307
tau2005 | 1.085146 1.797455 0.60 0.546 -2.437802 4.608094
tau2006 | .9815203 2.026142 0.48 0.628 -2.989645 4.952686
tau2007 | 2.923707 1.86269 1.57 0.117 -.7270979 6.574511
tau2008 | 0 (omitted)
tau2009 | -5.558402 1.229401 -4.52 0.000 -7.967984 -3.14882
tau2010 | 0 (omitted)
tau2011 | .8804618 2.105211 0.42 0.676 -3.245676 5.0066
_cons | 0 (omitted)
---------------------------------------------------------------------------------------------------
Instruments for first differences equation
Standard
D.(L.dettotoccrfls cL.dettotoccrfls#cL.dettotoccrfls L.netais L.PSIis
L.netaes L.PSIes L.corrups L.branchvas L.rspri L.spesatotvasn tau1995
tau1996 tau1997 tau1998 tau1999 tau2000 tau2001 tau2002 tau2003 tau2004
tau2005 tau2006 tau2007 tau2008 tau2009 tau2010 tau2011 tau2012)
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(2/3).L.gtfp collapsed
Instruments for levels equation
Standard
L.dettotoccrfls cL.dettotoccrfls#cL.dettotoccrfls L.netais L.PSIis
L.netaes L.PSIes L.corrups L.branchvas L.rspri L.spesatotvasn tau1995
tau1996 tau1997 tau1998 tau1999 tau2000 tau2001 tau2002 tau2003 tau2004
tau2005 tau2006 tau2007 tau2008 tau2009 tau2010 tau2011 tau2012
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
DL(1/2).L.gtfp collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -0.19 Pr > z = 0.846
Arellano-Bond test for AR(2) in first differences: z = -2.37 Pr > z = 0.018
Arellano-Bond test for AR(3) in first differences: z = -3.90 Pr > z = 0.000
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(16) = 49.10 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(16) = 0.00 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(14) = 0.00 Prob > chi2 = 1.000
Difference (null H = exogenous): chi2(2) = 0.00 Prob > chi2 = 1.000
gmm(L.gtfp, collapse eq(diff) lag(2 3))
Hansen test excluding group: chi2(14) = 0.00 Prob > chi2 = 1.000
Difference (null H = exogenous): chi2(2) = -0.00 Prob > chi2 = 1.000
gmm(L.gtfp, collapse eq(level) lag(1 2))
Hansen test excluding group: chi2(14) = 0.00 Prob > chi2 = 1.000
Difference (null H = exogenous): chi2(2) = 0.00 Prob > chi2 = 1.000
iv(L.dettotoccrfls cL.dettotoccrfls#cL.dettotoccrfls L.netais L.PSIis L.netaes L.PSIes L.corrups L.branchvas L.rspri L.spesatotvasn tau1995 t
> au1996 tau1997 tau1998 tau1999 tau2000 tau2001 tau2002 tau2003 tau2004 tau2005 tau2006 tau2007 tau2008 tau2009 tau2010 tau2011 tau2012, eq(le
> vel))
Hansen test excluding group: chi2(0) = 0.00 Prob > chi2 = .
Difference (null H = exogenous): chi2(16) = 0.00 Prob > chi2 = 1.000



In other words, why now rspri is omitted? Please note that rspub is R&D expenditure by the government, while rspri is R&D expenditure by public and private firms. I've checked for possible collinearity, but there is not.
Thanks for help.
Gc

[Sebastian Kripfganz]

There must be quite a bit of collinearity because so many of the time dummies are omitted. From merely looking at the regression results, it is however hardly possible to speculate about the reason.

More on GMM estimation of linear dynamic panel models:
XTDPDGMM: new Stata command for efficient GMM estimation of linear (dynamic) panel models with nonlinear moment conditions

[Giorgio Carlini]

Thanks a lot Sebastian for your help.
I'm trying to send the dataset with dataex but the file is too big. Is there a way for you to have a look at the data?
This can possibly be useful for some more suggestions.
Thanks again.
Gc

[Sebastian Kripfganz]

Sorry, I currently do not have the time to analyze your data file in detail. But maybe someone else can chip in here.

[Giorgio Carlini]

Hello Sebastian,
yes sure. Thanks anyway for your suggestions.
I've a final comment. I presume that collineraity should arise when you add a variable to the model estimation. However in my case, Stata omits the variable "rspri" when I drop the variable "rspub". It seems strange to me, isn't it? Does this imply that rspri has some collinearity with the remaining independent variables?
Thanks for your help.
Gc

[Nariman Sayed]

Hi Giorgio,
have you found a solution for the omtied indep variable pleas?