Regional dummies and interactions in xtabond2 two-step GMM
Dear All,
I am having an issue with using regional dummies in a two-step gmm procedure with xtabond2. Without the regional dummies, the results seem fine and in-line with what is found in the literature. However, when I include regional dummies and interactions, I get odd results. Is there something wrong with the code that I used?
Without dummy variables, and using:
I get the following:
Dynamic panel-data estimation, two-step system GMM
------------------------------------------------------------------------------
Group variable: country Number of obs = 1152
Time variable : year Number of groups = 112
Number of instruments = 33 Obs per group: min = 1
F(1, 111) = 17.74 avg = 10.29
Prob > F = 0.000 max = 29
------------------------------------------------------------------------------
| Corrected
grlhcountg~b | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
grlminc | -2.60158 .6177417 -4.21 0.000 -3.825676 -1.377483
_cons | .0160054 .0249184 0.64 0.522 -.0333721 .065383
------------------------------------------------------------------------------
Instruments for first differences equation
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(2/32).lminc collapsed
Instruments for levels equation
Standard
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
DL.lminc collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -3.65 Pr > z = 0.000
Arellano-Bond test for AR(2) in first differences: z = 1.07 Pr > z = 0.285
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(31) = 17.48 Prob > chi2 = 0.976
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(31) = 34.22 Prob > chi2 = 0.316
(Robust, but weakened by many instruments.)
Difference-in-Hansen tests of exogeneity of instrument subsets:
GMM instruments for levels
Hansen test excluding group: chi2(30) = 33.71 Prob > chi2 = 0.292
Difference (null H = exogenous): chi2(1) = 0.51 Prob > chi2 = 0.477
When using dummy variables with the code:
I get the following result:
Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.
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 Number of obs = 1152
Time variable : year Number of groups = 112
Number of instruments = 52 Obs per group: min = 1
F(25, 111) = 4.11 avg = 10.29
Prob > F = 0.000 max = 29
--------------------------------------------------------------------------------
| Corrected
grlhcountglob | Coef. Std. Err. t P>|t| [95% Conf. Interval]
---------------+----------------------------------------------------------------
grlminc | 49.83166 28.87619 1.73 0.087 -7.388451 107.0518
|
eap |
0 | 0 (empty)
1 | 6.888718 3.957638 1.74 0.085 -.9536063 14.73104
|
eca |
0 | 0 (empty)
1 | 6.977537 3.962253 1.76 0.081 -.8739327 14.82901
|
lca |
0 | 0 (empty)
1 | 7.027697 3.999272 1.76 0.082 -.8971276 14.95252
|
mena |
0 | 0 (empty)
1 | 6.821361 3.963337 1.72 0.088 -1.032257 14.67498
|
sa |
0 | 0 (empty)
1 | 7.015996 4.165934 1.68 0.095 -1.239081 15.27107
|
ssa |
0 | 0 (empty)
1 | 8.493254 6.202883 1.37 0.174 -3.798173 20.78468
|
eap#c.grlminc |
0 | 0 (empty)
1 | -50.19045 28.00473 -1.79 0.076 -105.6837 5.302786
|
eca#c.grlminc |
0 | 0 (empty)
1 | -50.40441 28.50497 -1.77 0.080 -106.8889 6.080084
|
lca#c.grlminc |
0 | 0 (empty)
1 | -51.13783 29.22885 -1.75 0.083 -109.0567 6.78108
|
mena#c.grlminc |
0 | 0 (empty)
1 | -50.83542 29.7939 -1.71 0.091 -109.874 8.203178
|
sa#c.grlminc |
0 | 0 (empty)
1 | -50.40036 27.55204 -1.83 0.070 -104.9966 4.195849
|
ssa#c.grlminc |
0 | 0 (empty)
1 | -42.59922 9.583961 -4.44 0.000 -61.59048 -23.60796
|
_cons | -7.029545 3.99622 -1.76 0.081 -14.94832 .8892326
--------------------------------------------------------------------------------
Instruments for first differences equation
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(2/10).(0b.eap#c.lminc 1.eap#c.lminc 0b.eca#c.lminc 1.eca#c.lminc
0b.lca#c.lminc 1.lca#c.lminc 0b.mena#c.lminc 1.mena#c.lminc 0b.sa#c.lminc
1.sa#c.lminc 0b.ssa#c.lminc 1.ssa#c.lminc) collapsed
Instruments for levels equation
Standard
eap eca lca mena sa ssa
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
DL.(0b.eap#c.lminc 1.eap#c.lminc 0b.eca#c.lminc 1.eca#c.lminc
0b.lca#c.lminc 1.lca#c.lminc 0b.mena#c.lminc 1.mena#c.lminc 0b.sa#c.lminc
1.sa#c.lminc 0b.ssa#c.lminc 1.ssa#c.lminc) collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -3.52 Pr > z = 0.000
Arellano-Bond test for AR(2) in first differences: z = 0.53 Pr > z = 0.598
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(26) = 57.65 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(26) = 19.69 Prob > chi2 = 0.806
(Robust, but weakened by many instruments.)
Difference-in-Hansen tests of exogeneity of instrument subsets:
GMM instruments for levels
Hansen test excluding group: chi2(21) = 13.23 Prob > chi2 = 0.900
Difference (null H = exogenous): chi2(5) = 6.46 Prob > chi2 = 0.264
iv(eap eca lca mena sa ssa, eq(level))
Hansen test excluding group: chi2(20) = 13.10 Prob > chi2 = 0.873
Difference (null H = exogenous): chi2(6) = 6.59 Prob > chi2 = 0.361
Would anyone be able to provide a recommendation on how to proceed with this issue?
Thank you very much
I am having an issue with using regional dummies in a two-step gmm procedure with xtabond2. Without the regional dummies, the results seem fine and in-line with what is found in the literature. However, when I include regional dummies and interactions, I get odd results. Is there something wrong with the code that I used?
Without dummy variables, and using:
Code:
xtabond2 grlhcountglob grlmedinc, twostep gmm(lmedinc, lag(2 .) collapse) small robust
Dynamic panel-data estimation, two-step system GMM
------------------------------------------------------------------------------
Group variable: country Number of obs = 1152
Time variable : year Number of groups = 112
Number of instruments = 33 Obs per group: min = 1
F(1, 111) = 17.74 avg = 10.29
Prob > F = 0.000 max = 29
------------------------------------------------------------------------------
| Corrected
grlhcountg~b | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
grlminc | -2.60158 .6177417 -4.21 0.000 -3.825676 -1.377483
_cons | .0160054 .0249184 0.64 0.522 -.0333721 .065383
------------------------------------------------------------------------------
Instruments for first differences equation
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(2/32).lminc collapsed
Instruments for levels equation
Standard
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
DL.lminc collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -3.65 Pr > z = 0.000
Arellano-Bond test for AR(2) in first differences: z = 1.07 Pr > z = 0.285
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(31) = 17.48 Prob > chi2 = 0.976
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(31) = 34.22 Prob > chi2 = 0.316
(Robust, but weakened by many instruments.)
Difference-in-Hansen tests of exogeneity of instrument subsets:
GMM instruments for levels
Hansen test excluding group: chi2(30) = 33.71 Prob > chi2 = 0.292
Difference (null H = exogenous): chi2(1) = 0.51 Prob > chi2 = 0.477
When using dummy variables with the code:
Code:
xtabond2 grlhcountglob c.grlminc##i.(eap eca lca mena sa ssa), twostep robust small gmm(c.lminc#i.(eap eca lca mena sa ssa), lag(2 10) collapse) iv( eap eca lca mena sa ssa, equation(level))
Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.
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 Number of obs = 1152
Time variable : year Number of groups = 112
Number of instruments = 52 Obs per group: min = 1
F(25, 111) = 4.11 avg = 10.29
Prob > F = 0.000 max = 29
--------------------------------------------------------------------------------
| Corrected
grlhcountglob | Coef. Std. Err. t P>|t| [95% Conf. Interval]
---------------+----------------------------------------------------------------
grlminc | 49.83166 28.87619 1.73 0.087 -7.388451 107.0518
|
eap |
0 | 0 (empty)
1 | 6.888718 3.957638 1.74 0.085 -.9536063 14.73104
|
eca |
0 | 0 (empty)
1 | 6.977537 3.962253 1.76 0.081 -.8739327 14.82901
|
lca |
0 | 0 (empty)
1 | 7.027697 3.999272 1.76 0.082 -.8971276 14.95252
|
mena |
0 | 0 (empty)
1 | 6.821361 3.963337 1.72 0.088 -1.032257 14.67498
|
sa |
0 | 0 (empty)
1 | 7.015996 4.165934 1.68 0.095 -1.239081 15.27107
|
ssa |
0 | 0 (empty)
1 | 8.493254 6.202883 1.37 0.174 -3.798173 20.78468
|
eap#c.grlminc |
0 | 0 (empty)
1 | -50.19045 28.00473 -1.79 0.076 -105.6837 5.302786
|
eca#c.grlminc |
0 | 0 (empty)
1 | -50.40441 28.50497 -1.77 0.080 -106.8889 6.080084
|
lca#c.grlminc |
0 | 0 (empty)
1 | -51.13783 29.22885 -1.75 0.083 -109.0567 6.78108
|
mena#c.grlminc |
0 | 0 (empty)
1 | -50.83542 29.7939 -1.71 0.091 -109.874 8.203178
|
sa#c.grlminc |
0 | 0 (empty)
1 | -50.40036 27.55204 -1.83 0.070 -104.9966 4.195849
|
ssa#c.grlminc |
0 | 0 (empty)
1 | -42.59922 9.583961 -4.44 0.000 -61.59048 -23.60796
|
_cons | -7.029545 3.99622 -1.76 0.081 -14.94832 .8892326
--------------------------------------------------------------------------------
Instruments for first differences equation
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(2/10).(0b.eap#c.lminc 1.eap#c.lminc 0b.eca#c.lminc 1.eca#c.lminc
0b.lca#c.lminc 1.lca#c.lminc 0b.mena#c.lminc 1.mena#c.lminc 0b.sa#c.lminc
1.sa#c.lminc 0b.ssa#c.lminc 1.ssa#c.lminc) collapsed
Instruments for levels equation
Standard
eap eca lca mena sa ssa
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
DL.(0b.eap#c.lminc 1.eap#c.lminc 0b.eca#c.lminc 1.eca#c.lminc
0b.lca#c.lminc 1.lca#c.lminc 0b.mena#c.lminc 1.mena#c.lminc 0b.sa#c.lminc
1.sa#c.lminc 0b.ssa#c.lminc 1.ssa#c.lminc) collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -3.52 Pr > z = 0.000
Arellano-Bond test for AR(2) in first differences: z = 0.53 Pr > z = 0.598
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(26) = 57.65 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(26) = 19.69 Prob > chi2 = 0.806
(Robust, but weakened by many instruments.)
Difference-in-Hansen tests of exogeneity of instrument subsets:
GMM instruments for levels
Hansen test excluding group: chi2(21) = 13.23 Prob > chi2 = 0.900
Difference (null H = exogenous): chi2(5) = 6.46 Prob > chi2 = 0.264
iv(eap eca lca mena sa ssa, eq(level))
Hansen test excluding group: chi2(20) = 13.10 Prob > chi2 = 0.873
Difference (null H = exogenous): chi2(6) = 6.59 Prob > chi2 = 0.361
Would anyone be able to provide a recommendation on how to proceed with this issue?
Thank you very much
