Hi everyone,
I investigate the effect of an advertising ban on tobacco consumption for 20 countries over 23 year (1990-2012).Thus, my dependent variable is tobacco consumption (logcons) and my explanatory variables are advertising ban dummies (weak, limited and comprehensive) and my control variables are price (logprice), income (loggdp) and unemployment rate (logunemp).
I use a dynamic specification to account for the addictive nature of smoking estimating the following equation with the xtabond2 command
xtabond2 logcons L.logcons logprice loggdp logunemp lim compr, gmm(logcons, lag(2 3)) iv(logprice loggdp logunemp lim compr), robust twostep noleveleq
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 optimal weighting matrix for two-step estimation.
Difference-in-Sargan/Hansen statistics may be negative.
Dynamic panel-data estimation, two-step difference GMM
Group variable: Country Number of obs = 541
Time variable : year Number of groups = 28
Number of instruments = 46 Obs per group: min = 15
Wald chi2(6) = 331.93 avg = 19.32
Prob > chi2 = 0.000 max = 21
Corrected
logcons Coef. Std. Err. z P>z [95% Conf. Interval]
logcons
L1. .6668875 .3051204 2.19 0.029 .0688625 1.264913
logprice .0199656 .0785868 0.25 0.799 -.1340617 .1739929
loggdp -.2998901 .1614429 -1.86 0.063 -.6163124 .0165322
logunemp -.0914051 .0399553 -2.29 0.022 -.169716 -.0130942
lim -.0220455 .0282488 -0.78 0.435 -.0774121 .0333212
compr -.0676385 .0433735 -1.56 0.119 -.152649 .017372
Instruments for first differences equation
Standard
D.(logprice loggdp logunemp lim compr)
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(2/3).logcons
Arellano-Bond test for AR(1) in first differences: z = -1.92 Pr > z = 0.054
Arellano-Bond test for AR(2) in first differences: z = -0.90 Pr > z = 0.365
Sargan test of overid. restrictions: chi2(40) = 128.73 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(40) = 24.93 Prob > chi2 = 0.970
(Robust, but weakened by many instruments.)
Difference-in-Hansen tests of exogeneity of instrument subsets:
iv(logprice loggdp logunemp lim compr)
Hansen test excluding group: chi2(35) = 22.71 Prob > chi2 = 0.946
Difference (null H = exogenous): chi2(5) = 2.22 Prob > chi2 = 0.817
This results reveals, that the lagged consumption variable is statistically significant.
But when I inclde timme dummies, the results change remarkable
xtabond2 logcons L.logcons logprice loggdp logunemp lim compr $t, gmm(logcons, lag(2 3)) iv(logprice loggd
> p logunemp lim compr $t) robust twostep noleveleq
Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.
year1991 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 difference GMM
Group variable: Country Number of obs = 541
Time variable : year Number of groups = 28
Number of instruments = 67 Obs per group: min = 15
Wald chi2(27) = 303.44 avg = 19.32
Prob > chi2 = 0.000 max = 21
Corrected
logcons Coef. Std. Err. z P>z [95% Conf. Interval]
logcons
L1. .5063031 2.921375 0.17 0.862 -5.219487 6.232093
logprice .1175291 .205652 0.57 0.568 -.2855414 .5205996
loggdp -7.735883 11.64233 -0.66 0.506 -30.55444 15.08267
logunemp -.5194387 .7286962 -0.71 0.476 -1.947657 .9087796
lim .0603257 .1073641 0.56 0.574 -.150104 .2707554
compr .0999747 .1603905 0.62 0.533 -.2143849 .4143343
year1992 .0706594 .1682118 0.42 0.674 -.2590296 .4003484
year1993 .1206641 .3345694 0.36 0.718 -.5350798 .7764081
year1994 .365265 .697461 0.52 0.600 -1.001733 1.732263
year1995 .5822932 1.02011 0.57 0.568 -1.417085 2.581671
year1996 .771179 1.327287 0.58 0.561 -1.830256 3.372614
year1997 1.012911 1.705471 0.59 0.553 -2.329751 4.355573
year1998 1.190283 2.006007 0.59 0.553 -2.741419 5.121985
year1999 1.389821 2.3166 0.60 0.549 -3.150631 5.930273
year2000 1.703258 2.80897 0.61 0.544 -3.802222 7.208738
year2001 1.800673 3.015305 0.60 0.550 -4.109216 7.710563
year2002 2.018676 3.347925 0.60 0.547 -4.543136 8.580487
year2003 2.03479 3.412857 0.60 0.551 -4.654288 8.723867
year2004 2.212245 3.729325 0.59 0.553 -5.097098 9.521588
year2005 2.237491 3.845669 0.58 0.561 -5.299881 9.774863
year2006 2.515415 4.301992 0.58 0.559 -5.916335 10.94716
year2007 2.657412 4.555732 0.58 0.560 -6.271658 11.58648
year2008 2.727101 4.740238 0.58 0.565 -6.563594 12.0178
year2009 2.667281 4.694487 0.57 0.570 -6.533744 11.86831
year2010 2.775009 4.923766 0.56 0.573 -6.875394 12.42541
year2011 2.890987 5.179522 0.56 0.577 -7.26069 13.04266
year2012 2.884796 5.257103 0.55 0.583 -7.418937 13.18853
Instruments for first differences equation
Standard
D.(logprice loggdp logunemp lim compr year1991 year1992 year1993 year1994
year1995 year1996 year1997 year1998 year1999 year2000 year2001 year2002
year2003 year2004 year2005 year2006 year2007 year2008 year2009 year2010
year2011 year2012)
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(2/3).logcons
Arellano-Bond test for AR(1) in first differences: z = 0.04 Pr > z = 0.972
Arellano-Bond test for AR(2) in first differences: z = 0.22 Pr > z = 0.824
Sargan test of overid. restrictions: chi2(40) = 134.80 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(40) = 1.11 Prob > chi2 = 1.000
(Robust, but weakened by many instruments.)
Difference-in-Hansen tests of exogeneity of instrument subsets:
iv(logprice loggdp logunemp lim compr year1991 year1992 year1993 year1994 year1995 year1996 year1997 year1
> 998 year1999 year2000 year2001 year2002 year2003 year2004 year2005 year2006 year2007 year2008 year2009 yea
> r2010 year2011 year2012)
Hansen test excluding group: chi2(14) = 0.80 Prob > chi2 = 1.000
Difference (null H = exogenous): chi2(26) = 0.31 Prob > chi2 = 1.000
All variables turn insignificant - especially the lagged consumption variable L1.logcons.
And now I don't know how to interpret these results? Is it even possible or recommendable to include that many time dummies or is my specification of the model or the syntax wrong?
I really appreciate your help!
Thanks a lot
Louisa
I investigate the effect of an advertising ban on tobacco consumption for 20 countries over 23 year (1990-2012).Thus, my dependent variable is tobacco consumption (logcons) and my explanatory variables are advertising ban dummies (weak, limited and comprehensive) and my control variables are price (logprice), income (loggdp) and unemployment rate (logunemp).
I use a dynamic specification to account for the addictive nature of smoking estimating the following equation with the xtabond2 command
xtabond2 logcons L.logcons logprice loggdp logunemp lim compr, gmm(logcons, lag(2 3)) iv(logprice loggdp logunemp lim compr), robust twostep noleveleq
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 optimal weighting matrix for two-step estimation.
Difference-in-Sargan/Hansen statistics may be negative.
Dynamic panel-data estimation, two-step difference GMM
Group variable: Country Number of obs = 541
Time variable : year Number of groups = 28
Number of instruments = 46 Obs per group: min = 15
Wald chi2(6) = 331.93 avg = 19.32
Prob > chi2 = 0.000 max = 21
Corrected
logcons Coef. Std. Err. z P>z [95% Conf. Interval]
logcons
L1. .6668875 .3051204 2.19 0.029 .0688625 1.264913
logprice .0199656 .0785868 0.25 0.799 -.1340617 .1739929
loggdp -.2998901 .1614429 -1.86 0.063 -.6163124 .0165322
logunemp -.0914051 .0399553 -2.29 0.022 -.169716 -.0130942
lim -.0220455 .0282488 -0.78 0.435 -.0774121 .0333212
compr -.0676385 .0433735 -1.56 0.119 -.152649 .017372
Instruments for first differences equation
Standard
D.(logprice loggdp logunemp lim compr)
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(2/3).logcons
Arellano-Bond test for AR(1) in first differences: z = -1.92 Pr > z = 0.054
Arellano-Bond test for AR(2) in first differences: z = -0.90 Pr > z = 0.365
Sargan test of overid. restrictions: chi2(40) = 128.73 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(40) = 24.93 Prob > chi2 = 0.970
(Robust, but weakened by many instruments.)
Difference-in-Hansen tests of exogeneity of instrument subsets:
iv(logprice loggdp logunemp lim compr)
Hansen test excluding group: chi2(35) = 22.71 Prob > chi2 = 0.946
Difference (null H = exogenous): chi2(5) = 2.22 Prob > chi2 = 0.817
This results reveals, that the lagged consumption variable is statistically significant.
But when I inclde timme dummies, the results change remarkable
xtabond2 logcons L.logcons logprice loggdp logunemp lim compr $t, gmm(logcons, lag(2 3)) iv(logprice loggd
> p logunemp lim compr $t) robust twostep noleveleq
Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.
year1991 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 difference GMM
Group variable: Country Number of obs = 541
Time variable : year Number of groups = 28
Number of instruments = 67 Obs per group: min = 15
Wald chi2(27) = 303.44 avg = 19.32
Prob > chi2 = 0.000 max = 21
Corrected
logcons Coef. Std. Err. z P>z [95% Conf. Interval]
logcons
L1. .5063031 2.921375 0.17 0.862 -5.219487 6.232093
logprice .1175291 .205652 0.57 0.568 -.2855414 .5205996
loggdp -7.735883 11.64233 -0.66 0.506 -30.55444 15.08267
logunemp -.5194387 .7286962 -0.71 0.476 -1.947657 .9087796
lim .0603257 .1073641 0.56 0.574 -.150104 .2707554
compr .0999747 .1603905 0.62 0.533 -.2143849 .4143343
year1992 .0706594 .1682118 0.42 0.674 -.2590296 .4003484
year1993 .1206641 .3345694 0.36 0.718 -.5350798 .7764081
year1994 .365265 .697461 0.52 0.600 -1.001733 1.732263
year1995 .5822932 1.02011 0.57 0.568 -1.417085 2.581671
year1996 .771179 1.327287 0.58 0.561 -1.830256 3.372614
year1997 1.012911 1.705471 0.59 0.553 -2.329751 4.355573
year1998 1.190283 2.006007 0.59 0.553 -2.741419 5.121985
year1999 1.389821 2.3166 0.60 0.549 -3.150631 5.930273
year2000 1.703258 2.80897 0.61 0.544 -3.802222 7.208738
year2001 1.800673 3.015305 0.60 0.550 -4.109216 7.710563
year2002 2.018676 3.347925 0.60 0.547 -4.543136 8.580487
year2003 2.03479 3.412857 0.60 0.551 -4.654288 8.723867
year2004 2.212245 3.729325 0.59 0.553 -5.097098 9.521588
year2005 2.237491 3.845669 0.58 0.561 -5.299881 9.774863
year2006 2.515415 4.301992 0.58 0.559 -5.916335 10.94716
year2007 2.657412 4.555732 0.58 0.560 -6.271658 11.58648
year2008 2.727101 4.740238 0.58 0.565 -6.563594 12.0178
year2009 2.667281 4.694487 0.57 0.570 -6.533744 11.86831
year2010 2.775009 4.923766 0.56 0.573 -6.875394 12.42541
year2011 2.890987 5.179522 0.56 0.577 -7.26069 13.04266
year2012 2.884796 5.257103 0.55 0.583 -7.418937 13.18853
Instruments for first differences equation
Standard
D.(logprice loggdp logunemp lim compr year1991 year1992 year1993 year1994
year1995 year1996 year1997 year1998 year1999 year2000 year2001 year2002
year2003 year2004 year2005 year2006 year2007 year2008 year2009 year2010
year2011 year2012)
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(2/3).logcons
Arellano-Bond test for AR(1) in first differences: z = 0.04 Pr > z = 0.972
Arellano-Bond test for AR(2) in first differences: z = 0.22 Pr > z = 0.824
Sargan test of overid. restrictions: chi2(40) = 134.80 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(40) = 1.11 Prob > chi2 = 1.000
(Robust, but weakened by many instruments.)
Difference-in-Hansen tests of exogeneity of instrument subsets:
iv(logprice loggdp logunemp lim compr year1991 year1992 year1993 year1994 year1995 year1996 year1997 year1
> 998 year1999 year2000 year2001 year2002 year2003 year2004 year2005 year2006 year2007 year2008 year2009 yea
> r2010 year2011 year2012)
Hansen test excluding group: chi2(14) = 0.80 Prob > chi2 = 1.000
Difference (null H = exogenous): chi2(26) = 0.31 Prob > chi2 = 1.000
All variables turn insignificant - especially the lagged consumption variable L1.logcons.
And now I don't know how to interpret these results? Is it even possible or recommendable to include that many time dummies or is my specification of the model or the syntax wrong?
I really appreciate your help!
Thanks a lot
Louisa
