Hi all
I am trying to estimate the impact of firm-level GVC (a binary variable) on firm env innovation (a binary variable) using WBES dataset of 41 countries and 3 years. The dataset is pooled cross-section and not panel. I am trying to run a probit model while accounting for country, year and industry fixed effects, but chi-square is reported as missing. If I understand correctly from other threads, that may be due to having lots of parameters in the model that absorbed much of the heterogeneity in the Y variable. My question is how to account for all these fixed effects in a probit model, Is there any code I can use to absorb them in probit or is there another model I can run to fit my data and the control of fixed effects? I have also read once that I can ignore the missing chi square but not that sure what can be the consequences of doing this? I have tried to run LPM with absorb fixed but at later stages of my analysis, I need to do instrumentation, which is by eprobit so I will have to go back to the probit model.
Thanks in advance !!
Here are the codes that I have used so far
probit innov GVC Xs i.year i.country i.industry_type [pw=wweak ], robust cluster (country)
reghdfe innov GVC Xs [pw=wweak ], absorb (year country industry_type) cluster (country)
Here are the results of probit
probit innov GVC_weak large_firm small_firm ln_age employee_training FC knowledge_dev NIQ ln_subsidy MANG_score_alt2 reg customer_certificate strategic_involovement i.year i.a1 i.industry_type [pw=wweak ], r
> obust cluster (a1)
Iteration 0: log pseudolikelihood = -2483275.1
Iteration 1: log pseudolikelihood = -1959170.7
Iteration 2: log pseudolikelihood = -1949453.4
Iteration 3: log pseudolikelihood = -1949444.4
Iteration 4: log pseudolikelihood = -1949444.4
Probit regression Number of obs = 19,509
Wald chi2(16) = .
Prob > chi2 = .
Log pseudolikelihood = -1949444.4 Pseudo R2 = 0.2150
(Std. err. adjusted for 35 clusters in a1)
--------------------------------------------------------------------------------------------
| Robust
innov | Coefficient std. err. z P>|z| [95% conf. interval]
---------------------------+----------------------------------------------------------------
GVC_weak | .3604266 .1203145 3.00 0.003 .1246145 .5962386
large_firm | .0645271 .0998798 0.65 0.518 -.1312336 .2602879
small_firm | .0471583 .1121527 0.42 0.674 -.172657 .2669735
ln_age | .168977 .0622416 2.71 0.007 .0469856 .2909683
employee_training | .0158048 .1107269 0.14 0.886 -.201216 .2328257
FC | -.0728579 .1086594 -0.67 0.503 -.2858264 .1401106
knowledge_dev | .4785452 .2641452 1.81 0.070 -.0391698 .9962602
NIQ | .4808054 .2457795 1.96 0.050 -.0009136 .9625245
ln_subsidy | .4180203 .2953128 1.42 0.157 -.1607822 .9968228
MANG_score_alt2 | 2.398323 .2487607 9.64 0.000 1.910761 2.885885
reg | .1319785 .1980638 0.67 0.505 -.2562194 .5201764
customer_certificate | .3304099 .0949439 3.48 0.001 .1443233 .5164965
strategic_involovement | .1718978 .1160677 1.48 0.139 -.0555907 .3993863
|
year |
2019 | .0518758 .0914128 0.57 0.570 -.1272901 .2310417
2020 | .5371449 .2227458 2.41 0.016 .1005711 .9737187
|
a1 |
Morocco | -.5929384 .5425835 -1.09 0.274 -1.656382 .4705057
Albania | 1.358481 .4062889 3.34 0.001 .5621698 2.154793
Georgia | 1.944364 1.07398 1.81 0.070 -.1605977 4.049326
Tajikistan | -.6343332 .1884926 -3.37 0.001 -1.003772 -.2648946
Turkey | -1.268857 .7705144 -1.65 0.100 -2.779037 .2413234
Ukraine | -.649828 .6056447 -1.07 0.283 -1.83687 .5372139
Uzbekistan | -.5938551 .8388376 -0.71 0.479 -2.237946 1.050236
Russia | -1.939941 1.388275 -1.40 0.162 -4.660911 .7810287
Poland | -.6432855 .6180704 -1.04 0.298 -1.854681 .5681102
Romania | -.8916952 .4496054 -1.98 0.047 -1.772906 -.0104848
Serbia | -.063677 .1290953 -0.49 0.622 -.3166992 .1893452
Kazakhstan | -.9848711 .8556733 -1.15 0.250 -2.66196 .6922178
Moldova | .9573349 .4271374 2.24 0.025 .120161 1.794509
Bosnia and Herzegovina | 1.191216 .4027597 2.96 0.003 .4018213 1.98061
Azerbaijan | -.6784546 .504745 -1.34 0.179 -1.667737 .3108273
North Macedonia | -.0224012 .0504192 -0.44 0.657 -.1212211 .0764186
Armenia | 1.962945 .8488859 2.31 0.021 .299159 3.62673
Kyrgyz Republic | -.4837911 .2442355 -1.98 0.048 -.9624839 -.0050983
Mongolia | .8027477 .0549015 14.62 0.000 .6951428 .9103525
Estonia | 1.546961 .5356801 2.89 0.004 .4970475 2.596875
Czechia | .7792579 .0996308 7.82 0.000 .5839851 .9745306
Hungary | .7761078 .136825 5.67 0.000 .5079357 1.04428
Latvia | .7420941 .0935974 7.93 0.000 .5586466 .9255417
Lithuania | .3934754 .0966013 4.07 0.000 .2041402 .5828105
Slovak Republic | .7567628 .1112149 6.80 0.000 .5387856 .9747401
Slovenia | .8335099 .2871212 2.90 0.004 .2707627 1.396257
Bulgaria | -.3715321 .4377153 -0.85 0.396 -1.229438 .486374
Egypt | -1.876557 1.169642 -1.60 0.109 -4.169014 .4158996
Greece | -.008544 .5898928 -0.01 0.988 -1.164713 1.147625
Portugal | .4442348 .2110546 2.10 0.035 .0305753 .8578943
Lebanon | -.7562414 .6536081 -1.16 0.247 -2.03729 .5248069
Tunisia | -.207709 .2819669 -0.74 0.461 -.7603539 .3449359
Cyprus | 1.326577 .1531103 8.66 0.000 1.026486 1.626667
Italy | -1.164064 1.091522 -1.07 0.286 -3.303408 .9752809
|
industry_type |
Retail services | -.0696127 .0383865 -1.81 0.070 -.1448487 .0056234
Other services | -.0344265 .0412406 -0.83 0.404 -.1152566 .0464036
|
_cons | -3.703088 1.788344 -2.07 0.038 -7.208178 -.197997
--------------------------------------------------------------------------------------------
Here are the results of LPM with absorbed fixed effects
reghdfe innov GVC_weak large_firm small_firm ln_age employee_training FC knowledge_dev NIQ ln_subsidy MANG_score_alt2 reg customer_certificate strategic_involovement [pw=wweak ], absorb (a1 year industry_type
> ) cluster (a1)
(MWFE estimator converged in 6 iterations)
HDFE Linear regression Number of obs = 19,509
Absorbing 3 HDFE groups F( 13, 34) = 773.03
Statistics robust to heteroskedasticity Prob > F = 0.0000
R-squared = 0.2447
Adj R-squared = 0.2427
Within R-sq. = 0.2129
Number of clusters (a1) = 35 Root MSE = 0.4340
(Std. err. adjusted for 35 clusters in a1)
--------------------------------------------------------------------------------------------
| Robust
innov| Coefficient std. err. t P>|t| [95% conf. interval]
---------------------------+----------------------------------------------------------------
GVC_weak | .0854028 .0398036 2.15 0.039 .0045122 .1662935
large_firm | -.0088717 .0296829 -0.30 0.767 -.0691946 .0514512
small_firm | .0181906 .0349781 0.52 0.606 -.0528935 .0892746
ln_age | .0500536 .0198766 2.52 0.017 .0096594 .0904477
employee_training | .0025855 .0325173 0.08 0.937 -.0634976 .0686687
FC | -.023247 .0338457 -0.69 0.497 -.0920298 .0455358
knowledge_dev | .1120606 .0668292 1.68 0.103 -.0237526 .2478738
NIQ | .1591949 .0770409 2.07 0.046 .0026289 .315761
ln_subsidy | .0761537 .0738407 1.03 0.310 -.0739087 .226216
MANG_score_alt2 | .7037628 .0573147 12.28 0.000 .5872855 .8202402
reg | .0337931 .0594995 0.57 0.574 -.0871244 .1547105
customer_certificate | .0859756 .0292428 2.94 0.006 .0265471 .145404
strategic_involovement | .0223843 .0253451 0.88 0.383 -.0291231 .0738916
_cons | -.4777069 .6464486 -0.74 0.465 -1.791449 .8360347
--------------------------------------------------------------------------------------------
Absorbed degrees of freedom:
-------------------------------------------------------+
Absorbed FE | Categories - Redundant = Num. Coefs |
---------------+---------------------------------------|
a1 | 35 35 0 *|
year | 3 1 2 |
industry_type | 3 1 2 |
-------------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation
.
I am trying to estimate the impact of firm-level GVC (a binary variable) on firm env innovation (a binary variable) using WBES dataset of 41 countries and 3 years. The dataset is pooled cross-section and not panel. I am trying to run a probit model while accounting for country, year and industry fixed effects, but chi-square is reported as missing. If I understand correctly from other threads, that may be due to having lots of parameters in the model that absorbed much of the heterogeneity in the Y variable. My question is how to account for all these fixed effects in a probit model, Is there any code I can use to absorb them in probit or is there another model I can run to fit my data and the control of fixed effects? I have also read once that I can ignore the missing chi square but not that sure what can be the consequences of doing this? I have tried to run LPM with absorb fixed but at later stages of my analysis, I need to do instrumentation, which is by eprobit so I will have to go back to the probit model.
Thanks in advance !!
Here are the codes that I have used so far
probit innov GVC Xs i.year i.country i.industry_type [pw=wweak ], robust cluster (country)
reghdfe innov GVC Xs [pw=wweak ], absorb (year country industry_type) cluster (country)
Here are the results of probit
probit innov GVC_weak large_firm small_firm ln_age employee_training FC knowledge_dev NIQ ln_subsidy MANG_score_alt2 reg customer_certificate strategic_involovement i.year i.a1 i.industry_type [pw=wweak ], r
> obust cluster (a1)
Iteration 0: log pseudolikelihood = -2483275.1
Iteration 1: log pseudolikelihood = -1959170.7
Iteration 2: log pseudolikelihood = -1949453.4
Iteration 3: log pseudolikelihood = -1949444.4
Iteration 4: log pseudolikelihood = -1949444.4
Probit regression Number of obs = 19,509
Wald chi2(16) = .
Prob > chi2 = .
Log pseudolikelihood = -1949444.4 Pseudo R2 = 0.2150
(Std. err. adjusted for 35 clusters in a1)
--------------------------------------------------------------------------------------------
| Robust
innov | Coefficient std. err. z P>|z| [95% conf. interval]
---------------------------+----------------------------------------------------------------
GVC_weak | .3604266 .1203145 3.00 0.003 .1246145 .5962386
large_firm | .0645271 .0998798 0.65 0.518 -.1312336 .2602879
small_firm | .0471583 .1121527 0.42 0.674 -.172657 .2669735
ln_age | .168977 .0622416 2.71 0.007 .0469856 .2909683
employee_training | .0158048 .1107269 0.14 0.886 -.201216 .2328257
FC | -.0728579 .1086594 -0.67 0.503 -.2858264 .1401106
knowledge_dev | .4785452 .2641452 1.81 0.070 -.0391698 .9962602
NIQ | .4808054 .2457795 1.96 0.050 -.0009136 .9625245
ln_subsidy | .4180203 .2953128 1.42 0.157 -.1607822 .9968228
MANG_score_alt2 | 2.398323 .2487607 9.64 0.000 1.910761 2.885885
reg | .1319785 .1980638 0.67 0.505 -.2562194 .5201764
customer_certificate | .3304099 .0949439 3.48 0.001 .1443233 .5164965
strategic_involovement | .1718978 .1160677 1.48 0.139 -.0555907 .3993863
|
year |
2019 | .0518758 .0914128 0.57 0.570 -.1272901 .2310417
2020 | .5371449 .2227458 2.41 0.016 .1005711 .9737187
|
a1 |
Morocco | -.5929384 .5425835 -1.09 0.274 -1.656382 .4705057
Albania | 1.358481 .4062889 3.34 0.001 .5621698 2.154793
Georgia | 1.944364 1.07398 1.81 0.070 -.1605977 4.049326
Tajikistan | -.6343332 .1884926 -3.37 0.001 -1.003772 -.2648946
Turkey | -1.268857 .7705144 -1.65 0.100 -2.779037 .2413234
Ukraine | -.649828 .6056447 -1.07 0.283 -1.83687 .5372139
Uzbekistan | -.5938551 .8388376 -0.71 0.479 -2.237946 1.050236
Russia | -1.939941 1.388275 -1.40 0.162 -4.660911 .7810287
Poland | -.6432855 .6180704 -1.04 0.298 -1.854681 .5681102
Romania | -.8916952 .4496054 -1.98 0.047 -1.772906 -.0104848
Serbia | -.063677 .1290953 -0.49 0.622 -.3166992 .1893452
Kazakhstan | -.9848711 .8556733 -1.15 0.250 -2.66196 .6922178
Moldova | .9573349 .4271374 2.24 0.025 .120161 1.794509
Bosnia and Herzegovina | 1.191216 .4027597 2.96 0.003 .4018213 1.98061
Azerbaijan | -.6784546 .504745 -1.34 0.179 -1.667737 .3108273
North Macedonia | -.0224012 .0504192 -0.44 0.657 -.1212211 .0764186
Armenia | 1.962945 .8488859 2.31 0.021 .299159 3.62673
Kyrgyz Republic | -.4837911 .2442355 -1.98 0.048 -.9624839 -.0050983
Mongolia | .8027477 .0549015 14.62 0.000 .6951428 .9103525
Estonia | 1.546961 .5356801 2.89 0.004 .4970475 2.596875
Czechia | .7792579 .0996308 7.82 0.000 .5839851 .9745306
Hungary | .7761078 .136825 5.67 0.000 .5079357 1.04428
Latvia | .7420941 .0935974 7.93 0.000 .5586466 .9255417
Lithuania | .3934754 .0966013 4.07 0.000 .2041402 .5828105
Slovak Republic | .7567628 .1112149 6.80 0.000 .5387856 .9747401
Slovenia | .8335099 .2871212 2.90 0.004 .2707627 1.396257
Bulgaria | -.3715321 .4377153 -0.85 0.396 -1.229438 .486374
Egypt | -1.876557 1.169642 -1.60 0.109 -4.169014 .4158996
Greece | -.008544 .5898928 -0.01 0.988 -1.164713 1.147625
Portugal | .4442348 .2110546 2.10 0.035 .0305753 .8578943
Lebanon | -.7562414 .6536081 -1.16 0.247 -2.03729 .5248069
Tunisia | -.207709 .2819669 -0.74 0.461 -.7603539 .3449359
Cyprus | 1.326577 .1531103 8.66 0.000 1.026486 1.626667
Italy | -1.164064 1.091522 -1.07 0.286 -3.303408 .9752809
|
industry_type |
Retail services | -.0696127 .0383865 -1.81 0.070 -.1448487 .0056234
Other services | -.0344265 .0412406 -0.83 0.404 -.1152566 .0464036
|
_cons | -3.703088 1.788344 -2.07 0.038 -7.208178 -.197997
--------------------------------------------------------------------------------------------
Here are the results of LPM with absorbed fixed effects
reghdfe innov GVC_weak large_firm small_firm ln_age employee_training FC knowledge_dev NIQ ln_subsidy MANG_score_alt2 reg customer_certificate strategic_involovement [pw=wweak ], absorb (a1 year industry_type
> ) cluster (a1)
(MWFE estimator converged in 6 iterations)
HDFE Linear regression Number of obs = 19,509
Absorbing 3 HDFE groups F( 13, 34) = 773.03
Statistics robust to heteroskedasticity Prob > F = 0.0000
R-squared = 0.2447
Adj R-squared = 0.2427
Within R-sq. = 0.2129
Number of clusters (a1) = 35 Root MSE = 0.4340
(Std. err. adjusted for 35 clusters in a1)
--------------------------------------------------------------------------------------------
| Robust
innov| Coefficient std. err. t P>|t| [95% conf. interval]
---------------------------+----------------------------------------------------------------
GVC_weak | .0854028 .0398036 2.15 0.039 .0045122 .1662935
large_firm | -.0088717 .0296829 -0.30 0.767 -.0691946 .0514512
small_firm | .0181906 .0349781 0.52 0.606 -.0528935 .0892746
ln_age | .0500536 .0198766 2.52 0.017 .0096594 .0904477
employee_training | .0025855 .0325173 0.08 0.937 -.0634976 .0686687
FC | -.023247 .0338457 -0.69 0.497 -.0920298 .0455358
knowledge_dev | .1120606 .0668292 1.68 0.103 -.0237526 .2478738
NIQ | .1591949 .0770409 2.07 0.046 .0026289 .315761
ln_subsidy | .0761537 .0738407 1.03 0.310 -.0739087 .226216
MANG_score_alt2 | .7037628 .0573147 12.28 0.000 .5872855 .8202402
reg | .0337931 .0594995 0.57 0.574 -.0871244 .1547105
customer_certificate | .0859756 .0292428 2.94 0.006 .0265471 .145404
strategic_involovement | .0223843 .0253451 0.88 0.383 -.0291231 .0738916
_cons | -.4777069 .6464486 -0.74 0.465 -1.791449 .8360347
--------------------------------------------------------------------------------------------
Absorbed degrees of freedom:
-------------------------------------------------------+
Absorbed FE | Categories - Redundant = Num. Coefs |
---------------+---------------------------------------|
a1 | 35 35 0 *|
year | 3 1 2 |
industry_type | 3 1 2 |
-------------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation
.
Comment