APPENDIX I
a. Results for the sample of manufacturing firms
. reg rota cul_dis eco_dis ins_dis exp_yrs No_sub No_employee
Source | SS df MS Number of obs = 152 ---+--- F( 6, 145) = 4.32 Model | 3672.87309 6 612.145514 Prob > F = 0.0005 Residual | 20541.4844 145 141.665409 R-squared = 0.1517 ---+--- Adj R-squared = 0.1166 Total | 24214.3574 151 160.359983 Root MSE = 11.902 rota | Coef. Std. Err. t P>|t| [95% Conf. Interval] cul_dis | -.556176 1.239551 -0.45 0.654 -3.006098 1.893746 eco_dis | -11.44244 4.458519 -2.57 0.011 -20.25452 -2.630354 ins_dis | 4.68552 2.169108 2.16 0.032 .3983657 8.972674 exp_yrs | -.0131333 .077726 -0.17 0.866 -.1667555 .140489 No_sub | .0122145 .0036739 3.32 0.001 .0049531 .0194759 No_employee | -.0741967 .1868884 -0.40 0.692 -.443574 .2951807 _cons | 21.79704 8.022274 2.72 0.007 5.941336 37.65273 ---. . predict yhat
(option xb assumed; fitted values) .
. predict ehat, residuals .
. histogram ehat, percent
(bin=12, start=-42.108971, width=7.3823741) .
. di "Jarque-Bera Statistic = " jb Jarque-Bera Statistic = 26.2276 .
. scalar define chic = invchi2tail(2,.05) .
. di "Chi-square(2) 95th percentile = " chic Chi-square(2) 95th percentile = 5.9914645 .
. scalar define pvalue = chi2tail(2,jb) .
. di "Jarque-Bera p-value = " pvalue Jarque-Bera p-value = 2.017e-06
b. Results for the sample of non-manufacturing firms
. reg rota cul_dis eco_dis ins_dis exp_yrs No_sub No_employee
---.
. predict yhat
(option xb assumed; fitted values) .
. predict ehat, residuals .
. histogram ehat, percent
(bin=13, start=-23.4648, width=5.2476265) .
. sum ehat, detail
r(p5) = -16.22714805603027 r(p10) = -10.31724834442139 r(p25) = -5.747159242630005 r(p50) = -.6076636016368866 r(p75) = 4.600624799728394 r(p90) = 12.24035167694092 r(p95) = 17.01251792907715 r(p99) = 29.99980735778809 . . scalar jb = (184/6)*((0.66^2) + (5.30-3)^2/4) . . di "Jarque-Bera Statistic = " jb Jarque-Bera Statistic = 53.915067 .
. scalar define chic = invchi2tail(2,.05) .
. di "Chi-square(2) 95th percentile = " chic Chi-square(2) 95th percentile = 5.9914645 .
. scalar define pvalue = chi2tail(2,jb) .
. di "Jarque-Bera p-value = " pvalue Jarque-Bera p-value = 1.961e-12
APPENDIX II
a. Breusch–Pagan test results for manufacturing firms
Source | SS df MS Number of obs = 152 ---+--- F( 6, 145) = 4.32 Model | 3672.87309 6 612.145514 Prob > F = 0.0005 Residual | 20541.4844 145 141.665409 R-squared = 0.1517 ---+--- Adj R-squared = 0.1166 Total | 24214.3574 151 160.359983 Root MSE = 11.902 rota | Coef. Std. Err. t P>|t| [95% Conf. Interval] cul_dis | -.556176 1.239551 -0.45 0.654 -3.006098 1.893746 eco_dis | -11.44244 4.458519 -2.57 0.011 -20.25452 -2.630354 ins_dis | 4.68552 2.169108 2.16 0.032 .3983657 8.972674 exp_yrs | -.0131333 .077726 -0.17 0.866 -.1667555 .140489 No_sub | .0122145 .0036739 3.32 0.001 .0049531 .0194759 No_employee | -.0741967 .1868884 -0.40 0.692 -.443574 .2951807 _cons | 21.79704 8.022274 2.72 0.007 5.941336 37.65273 ---.
. predict ehat, residual .
. gen ehat2 = ehat*ehat .
. reg ehat2 cul_dis
. scalar LM = e(N)*e(r2) .
. scalar pvalue = chi2tail(1,LM) .
. scalar list LM pvalue LM = 1.337277 pvalue = .24751474 .
. .
. reg ehat2 eco_dis
Source | SS df MS Number of obs = 152 ---+--- F( 1, 150) = 2.76 Model | 203662.802 1 203662.802 Prob > F = 0.0988 Residual | 11074001.4 150 73826.6761 R-squared = 0.0181 ---+--- Adj R-squared = 0.0115 Total | 11277664.2 151 74686.518 Root MSE = 271.71 ehat2 | Coef. Std. Err. t P>|t| [95% Conf. Interval] eco_dis | 136.2225 82.01617 1.66 0.099 -25.83374 298.2787 _cons | -119.4335 154.8495 -0.77 0.442 -425.4015 186.5345 ---. . scalar LM = e(N)*e(r2) .
. scalar pvalue = chi2tail(1,LM) .
. scalar list LM pvalue LM = 2.7449608 pvalue = .09756147 .
. .
Source | SS df MS Number of obs = 152 ---+--- F( 1, 150) = 2.23 Model | 164994.046 1 164994.046 Prob > F = 0.1377 Residual | 11112670.2 150 74084.4678 R-squared = 0.0146 ---+--- Adj R-squared = 0.0081 Total | 11277664.2 151 74686.518 Root MSE = 272.18 ehat2 | Coef. Std. Err. t P>|t| [95% Conf. Interval] ins_dis | 61.89728 41.4764 1.49 0.138 -20.05615 143.8507 _cons | 39.03157 68.08062 0.57 0.567 -95.4893 173.5524 ---. . scalar LM = e(N)*e(r2) .
. scalar pvalue = chi2tail(1,LM) .
. scalar list LM pvalue LM = 2.2237845 pvalue = .13589957 .
. .
. reg ehat2 exp_yrs
---.
. scalar LM = e(N)*e(r2) .
. scalar pvalue = chi2tail(1,LM) .
. scalar list LM pvalue LM = 3.8819202 pvalue = .0488087 .
. .
. reg ehat2 No_sub
Source | SS df MS Number of obs = 152 ---+--- F( 1, 150) = 6.82 Model | 490673.653 1 490673.653 Prob > F = 0.0099 Residual | 10786990.6 150 71913.2704 R-squared = 0.0435 ---+--- Adj R-squared = 0.0371 Total | 11277664.2 151 74686.518 Root MSE = 268.17 ehat2 | Coef. Std. Err. t P>|t| [95% Conf. Interval] No_sub | .198925 .0761548 2.61 0.010 .0484502 .3493997 _cons | 98.6962 25.84148 3.82 0.000 47.63588 149.7565 ---. . scalar LM = e(N)*e(r2) .
. scalar pvalue = chi2tail(1,LM) .
. scalar list LM pvalue LM = 6.6132839 pvalue = .01012208 .
.
. reg ehat2 No_employee
Source | SS df MS Number of obs = 152 ---+--- F( 1, 150) = 1.04 Model | 77373.1288 1 77373.1288 Prob > F = 0.3103 Residual | 11200291.1 150 74668.6073 R-squared = 0.0069 ---+--- Adj R-squared = 0.0002 Total | 11277664.2 151 74686.518 Root MSE = 273.26 ehat2 | Coef. Std. Err. t P>|t| [95% Conf. Interval] ---+---No_employee | 4.193197 4.11926 1.02 0.310 -3.946071 12.33247 _cons | 120.3856 26.48323 4.55 0.000 68.05727 172.714 ---. . scalar LM = e(N)*e(r2) .
. scalar pvalue = chi2tail(1,LM) .
. scalar list LM pvalue LM = 1.0428326 pvalue = .3071636
b. Breusch–Pagan test results for non-manufacturing firms
. reg rota cul_dis eco_dis ins_dis exp_yrs No_sub No_employee
cul_dis | .7165015 .9903376 0.72 0.470 -1.237887 2.67089 eco_dis | 2.620169 2.164398 1.21 0.228 -1.651178 6.891517 ins_dis | -1.171134 2.449743 -0.48 0.633 -6.005596 3.663329 exp_yrs | .297779 .1539948 1.93 0.055 -.0061231 .6016811 No_sub | -.000915 .0021623 -0.42 0.673 -.0051821 .0033521 No_employee | -.1142697 .1727482 -0.66 0.509 -.4551809 .2266414 _cons | -.2274298 3.399625 -0.07 0.947 -6.936444 6.481585 ---.
. predict ehat, residual .
. gen ehat2 = ehat*ehat .
. reg ehat2 cul_dis
Source | SS df MS Number of obs = 184 ---+--- F( 1, 182) = 0.63 Model | 25548.0482 1 25548.0482 Prob > F = 0.4302 Residual | 7438841.59 182 40872.756 R-squared = 0.0034 ---+--- Adj R-squared = -0.0021 Total | 7464389.64 183 40789.0144 Root MSE = 202.17 ehat2 | Coef. Std. Err. t P>|t| [95% Conf. Interval] cul_dis | 14.63195 18.50719 0.79 0.430 -21.88428 51.14819 _cons | 69.58794 37.92531 1.83 0.068 -5.241888 144.4178 ---. . scalar LM = e(N)*e(r2) .
. scalar pvalue = chi2tail(1,LM) .
. . .
. reg ehat2 eco_dis
Source | SS df MS Number of obs = 184 ---+--- F( 1, 182) = 0.14 Model | 5552.23129 1 5552.23129 Prob > F = 0.7132 Residual | 7458837.41 182 40982.6231 R-squared = 0.0007 ---+--- Adj R-squared = -0.0047 Total | 7464389.64 183 40789.0144 Root MSE = 202.44 ehat2 | Coef. Std. Err. t P>|t| [95% Conf. Interval] eco_dis | -10.4057 28.27076 -0.37 0.713 -66.1863 45.37489 _cons | 114.9639 50.62151 2.27 0.024 15.08339 214.8444 ---. . scalar LM = e(N)*e(r2) .
. scalar pvalue = chi2tail(1,LM) .
. scalar list LM pvalue LM = .13686458 pvalue = .71141816 .
. .
. reg ehat2 ins_dis
ins_dis | 5.685286 32.91434 0.17 0.863 -59.25748 70.62805 _cons | 87.96521 55.28381 1.59 0.113 -21.1144 197.0448 ---. . scalar LM = e(N)*e(r2) .
. scalar pvalue = chi2tail(1,LM) .
. scalar list LM pvalue LM = .03015848 pvalue = .86213112 .
. .
. reg ehat2 exp_yrs
Source | SS df MS Number of obs = 184 ---+--- F( 1, 182) = 0.91 Model | 37005.4157 1 37005.4157 Prob > F = 0.3422 Residual | 7427384.23 182 40809.8034 R-squared = 0.0050 ---+--- Adj R-squared = -0.0005 Total | 7464389.64 183 40789.0144 Root MSE = 202.01 ehat2 | Coef. Std. Err. t P>|t| [95% Conf. Interval] exp_yrs | -2.817808 2.959109 -0.95 0.342 -8.656378 3.020763 _cons | 124.1738 32.04035 3.88 0.000 60.95553 187.3921 ---. . scalar LM = e(N)*e(r2) .
. scalar pvalue = chi2tail(1,LM) .
pvalue = .33953207 .
. .
. reg ehat2 No_sub
Source | SS df MS Number of obs = 184 ---+--- F( 1, 182) = 1.47 Model | 59751.6314 1 59751.6314 Prob > F = 0.2271 Residual | 7404638.01 182 40684.8242 R-squared = 0.0080 ---+--- Adj R-squared = 0.0026 Total | 7464389.64 183 40789.0144 Root MSE = 201.7 ehat2 | Coef. Std. Err. t P>|t| [95% Conf. Interval] No_sub | -.0511241 .0421858 -1.21 0.227 -.1343603 .0321121 _cons | 105.1705 16.27298 6.46 0.000 73.0626 137.2785 ---. . scalar LM = e(N)*e(r2) .
. scalar pvalue = chi2tail(1,LM) .
. scalar list LM pvalue LM = 1.4729001 pvalue = .22488868 .
. .
. reg ehat2 No_employee
ehat2 | Coef. Std. Err. t P>|t| [95% Conf. Interval] ---+---No_employee | -1.547995 3.079017 -0.50 0.616 -7.623155 4.527165 _cons | 100.8166 16.59811 6.07 0.000 68.06713 133.5661 ---. . scalar LM = e(N)*e(r2) .
. scalar pvalue = chi2tail(1,LM) .
. scalar list LM pvalue LM = .255187 pvalue = .6134462
APPENDIX III
a. Results of autocorrelation for the sample of manufacturing firms
. xtset company year
panel variable: company (strongly balanced) time variable: year, 2000 to 2007
delta: 1 unit
. xtserial rota cul_dis eco_dis ins_dis exp_yrs No_sub No_employee Wooldridge test for autocorrelation in panel data
b. Results of autocorrelation for the sample of non-manufacturing firms
. xtset company year
panel variable: company (strongly balanced) time variable: year, 2000 to 2007
delta: 1 unit .
. xtserial rota cul_dis eco_dis ins_dis exp_yrs No_sub No_employee Wooldridge test for autocorrelation in panel data
H0: no first-order autocorrelation F( 1, 22) = 8.714 Prob > F = 0.0074
APPENDIX IV
a. Results of the Hausman-test for the sample of manufacturing firms
. sort year company .
. xtset year company
panel variable: year (strongly balanced) time variable: company, 1 to 19
delta: 1 unit .
. xtreg rota cul_dis eco_dis ins_dis exp_yrs No_sub No_employee, fe
between = 0.8440 avg = 19.0 overall = 0.1481 max = 19 F(6,138) = 5.12 corr(u_i, Xb) = -0.0657 Prob > F = 0.0001 rota | Coef. Std. Err. t P>|t| [95% Conf. Interval] cul_dis | -.6799319 1.186866 -0.57 0.568 -3.026726 1.666862 eco_dis | -13.19269 4.287271 -3.08 0.003 -21.66993 -4.715453 ins_dis | 5.71207 2.091644 2.73 0.007 1.576255 9.847885 exp_yrs | -.0731125 .0758847 -0.96 0.337 -.2231596 .0769347 No_sub | .0114042 .0035222 3.24 0.002 .0044397 .0183687 No_employee | -.0947777 .1790502 -0.53 0.597 -.4488142 .2592588 _cons | 24.75811 7.711009 3.21 0.002 9.511102 40.00511 sigma_u | 4.5320803 sigma_e | 11.391798
rho | .13664671 (fraction of variance due to u_i)
---F test that all u_i=0: ---F(7, 138) = 2.90 Prob > ---F = 0.0074 .
. estimates store fixed .
. xtreg rota cul_dis eco_dis ins_dis exp_yrs No_sub No_employee, re
ins_dis | 4.68552 2.169108 2.16 0.031 .4341464 8.936893 exp_yrs | -.0131333 .077726 -0.17 0.866 -.1654733 .1392068 No_sub | .0122145 .0036739 3.32 0.001 .0050137 .0194153 No_employee | -.0741967 .1868884 -0.40 0.691 -.4404911 .2920978 _cons | 21.79704 8.022274 2.72 0.007 6.073668 37.5204 sigma_u | 0 sigma_e | 11.391798
rho | 0 (fraction of variance due to u_i)
---.
. estimates store random .
. hausman fixed random, sigmamore
Note: the rank of the differenced variance matrix (4) does not equal the number of coefficients being tested (6); be sure this is what you expect, or there may be problems computing the test. Examine the output of your estimators for anything unexpected and possibly consider scaling your variables so that the coefficients are on a similar scale.
Coefficients
| (b) (B) (b-B) sqrt(diag(V_b-V_B)) | fixed random Difference S.E.
cul_dis | -.6799319 -.556176 -.1237559 .0353938 eco_dis | -13.19269 -11.44244 -1.750253 .4320906 ins_dis | 5.71207 4.68552 1.026551 .2662069 exp_yrs | -.0731125 -.0131333 -.0599792 .0156484 No_sub | .0114042 .0122145 -.0008103 .0002125 No_employee | -.0947777 -.0741967 -.020581 .0083404 b = consistent under Ho and Ha; obtained from xtreg B = inconsistent under Ha, efficient under Ho; obtained from xtreg Test: Ho: difference in coefficients not systematic
chi2(4) = (b-B)'[(V_b-V_B)^(-1)](b-B) = 17.32
Prob>chi2 = 0.0017
b. Results of the Hausman-test for the sample of non-manufacturing firms
. sort year company .
. xtset year company
panel variable: year (strongly balanced) time variable: company, 1 to 23
delta: 1 unit .
. xtreg rota cul_dis eco_dis ins_dis exp_yrs No_sub No_employee, fe
Fixed-effects (within) regression Number of obs = 184 Group variable: year Number of groups = 8 R-sq: within = 0.0399 Obs per group: min = 23 between = 0.7327 avg = 23.0 overall = 0.0297 max = 23 F(6,170) = 1.18 corr(u_i, Xb) = -0.0567 Prob > F = 0.3209 rota | Coef. Std. Err. t P>|t| [95% Conf. Interval] cul_dis | .5349412 .9709872 0.55 0.582 -1.381804 2.451686 eco_dis | 1.330578 2.152024 0.62 0.537 -2.917553 5.578709 ins_dis | 1.307044 2.499348 0.52 0.602 -3.626711 6.240798 exp_yrs | .0416682 .1670301 0.25 0.803 -.288052 .3713885 No_sub | -.0012347 .0021187 -0.58 0.561 -.005417 .0029475 No_employee | -.2274478 .1721504 -1.32 0.188 -.5672755 .1123799 _cons | 1.086207 3.349619 0.32 0.746 -5.525996 7.69841 sigma_u | 3.4520184 sigma_e | 9.8375718
rho | .10963258 (fraction of variance due to u_i)
. estimates store fixed .
. xtreg rota cul_dis eco_dis ins_dis exp_yrs No_sub No_employee, re
Random-effects GLS regression Number of obs = 184 Group variable: year Number of groups = 8 R-sq: within = 0.0251 Obs per group: min = 23 between = 0.9100 avg = 23.0 overall = 0.0555 max = 23 Random effects u_i ~ Gaussian Wald chi2(6) = 10.39 corr(u_i, X) = 0 (assumed) Prob > chi2 = 0.1091 rota | Coef. Std. Err. z P>|z| [95% Conf. Interval] cul_dis | .7165015 .9903376 0.72 0.469 -1.224525 2.657527 eco_dis | 2.620169 2.164398 1.21 0.226 -1.621973 6.862312 ins_dis | -1.171134 2.449743 -0.48 0.633 -5.972541 3.630274 exp_yrs | .297779 .1539948 1.93 0.053 -.0040452 .5996032 No_sub | -.000915 .0021623 -0.42 0.672 -.0051529 .0033229 No_employee | -.1142697 .1727482 -0.66 0.508 -.4528499 .2243105 _cons | -.2274298 3.399625 -0.07 0.947 -6.890573 6.435713 sigma_u | 0 sigma_e | 9.8375718
rho | 0 (fraction of variance due to u_i)
---.
. estimates store random .
. hausman fixed random, sigmamore
Note: the rank of the differenced variance matrix (4) does not equal the number of coefficients being tested (6); be sure this is what you expect, or there may be problems computing the test. Examine the output of your estimators for anything unexpected and possibly consider scaling your variables so that the coefficients are on a similar scale.
| (b) (B) (b-B) sqrt(diag(V_b-V_B)) | fixed random Difference S.E.
cul_dis | .5349412 .7165015 -.1815603 .056577 eco_dis | 1.330578 2.620169 -1.289591 .3856632 ins_dis | 1.307044 -1.171134 2.478177 .7198377 exp_yrs | .0416682 .297779 -.2561108 .0735013 No_sub | -.0012347 -.000915 -.0003197 .0000963 No_employee | -.2274478 -.1142697 -.1131781 .0329756 b = consistent under Ho and Ha; obtained from xtreg B = inconsistent under Ha, efficient under Ho; obtained from xtreg Test: Ho: difference in coefficients not systematic
chi2(4) = (b-B)'[(V_b-V_B)^(-1)](b-B) = 12.72
Prob>chi2 = 0.0127
(V_b-V_B is not positive definite)
APPENDIX V
a. Robust results of the FE estimator to the sample of non-manufacturing firms
. sort year company . xtset year company
panel variable: year (strongly balanced) time variable: company, 1 to 19
delta: 1 unit
. xtreg rota cul_dis eco_dis ins_dis exp_yrs No_sub No_employee, fe vce(cluste > r year)
R-sq: within = 0.1821 Obs per group: min = 19 between = 0.8440 avg = 19.0 overall = 0.1481 max = 19 F(6,7) = 25.33 corr(u_i, Xb) = -0.0657 Prob > F = 0.0002 (Std. Err. adjusted for 8 clusters in year) | Robust
rota | Coef. Std. Err. t P>|t| [95% Conf. Interval] cul_dis | -.6799319 .2652045 -2.56 0.037 -1.307041 -.0528228 eco_dis | -13.19269 1.607742 -8.21 0.000 -16.9944 -9.390983 ins_dis | 5.71207 .7250806 7.88 0.000 3.997527 7.426613 exp_yrs | -.0731125 .0398403 -1.84 0.109 -.1673199 .021095 No_sub | .0114042 .0053059 2.15 0.069 -.0011421 .0239506 No_employee | -.0947777 .1539375 -0.62 0.558 -.458782 .2692266 _cons | 24.75811 2.68914 9.21 0.000 18.3993 31.11691 sigma_u | 4.5320803 sigma_e | 11.391798
rho | .13664671 (fraction of variance due to u_i)
---b. Robust results of the FE estimator to the sample of non-manufacturing firms
. sort year company .
. xtset year company
panel variable: year (strongly balanced) time variable: company, 1 to 23
delta: 1 unit .
. xtreg rota cul_dis eco_dis ins_dis exp_yrs No_sub No_employee, fe vce(cluste > r year)
R-sq: within = 0.0399 Obs per group: min = 23 between = 0.7327 avg = 23.0 overall = 0.0297 max = 23 F(6,7) = 27.17 corr(u_i, Xb) = -0.0567 Prob > F = 0.0002 (Std. Err. adjusted for 8 clusters in year) | Robust
rota | Coef. Std. Err. t P>|t| [95% Conf. Interval] cul_dis | .5349412 1.141827 0.47 0.654 -2.165051 3.234933 eco_dis | 1.330578 1.695353 0.78 0.458 -2.678295 5.339451 ins_dis | 1.307044 1.002312 1.30 0.233 -1.063048 3.677136 exp_yrs | .0416682 .0587354 0.71 0.501 -.097219 .1805554 No_sub | -.0012347 .0010001 -1.23 0.257 -.0035995 .0011301 No_employee | -.2274478 .0461266 -4.93 0.002 -.3365197 -.1183758 _cons | 1.086207 3.87177 0.28 0.787 -8.069073 10.24149 sigma_u | 3.4520184 sigma_e | 9.8375718
rho | .10963258 (fraction of variance due to u_i)