Democracy Does Cause Growth Part 2

Table 1: Summary statistics for the main variables used in our analysis. Nondemocracies Democracies Obs. Mean Std. dev. Obs. Mean Std. dev GDP per Capita 3,376 $2,074.46 $3,838.65 3,558 $8,149.97 $9,334.83 Investment Share of GDP 3,225 0.2182 0.1023 3,340 0.2328 0.0741 TFP 1,863 1.0676 0.4056 2,744 0.9345 0.1646 Trade Share of GDP 3,175 0.7162 0.5106 3,485 0.7715 0.4104 Primary Enrollment Rate 2,861 90.29 29.51 2,823 101.60 15.86 Secondary Enrollment Rate 2,424 45.76 31.77 2,538 75.40 29.78 Tax Revenue Share of GDP 3,122 0.1587 0.0948 2,564 0.2075 0.0955 Child Mortality Per 1000 births 4,142 77.29 49.64 3,615 33.26 32.65 Unrest Dummy 3,739 0.2870 0.4524 3,610 0.2191 0.4137 Market Reforms Index (0-100) 3476 21.89 23.26 2,829 52.11 24.75 Notes: See the text for a full description of the variables and their corresponding sources. The table presents the statistics separately for nondemocracies (country/years for which our dichotomous democracy measure is 0) and democracies (country/years for which our dichotomous democracy measure is 1). Table 2: Effect of democracy on (log) GDP per capita. Within estimates Democracy log GDP first lag (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) 0.973 (0.294) 0.973 (0.006) 0.651 (0.248) 1.266 (0.038) -0.300 (0.037) 0.787 (0.226) 1.238 (0.038) -0.207 (0.046) -0.026 (0.028) -0.043 (0.017) 0.959 (0.477) 0.946 (0.009) 0.797 (0.417) 1.216 (0.041) -0.270 (0.038) 0.875 (0.374) 1.204 (0.041) -0.193 (0.045) -0.028 (0.028) -0.036 (0.020) 0.582 (0.387) 1.158 (0.038) -0.217 (0.035) 1.178 (0.370) 1.150 (0.040) -0.127 (0.050) -0.030 (0.026) -0.039 (0.015) 19.599 (8.595) 13.800 (5.550) 0.967 (0.005) 21.240 (7.215) 16.895 (5.297) 0.963 (0.005) 17.608 (10.609) 13.263 (7.281) 0.946 (0.009) [0.01] 14.882 (9.152) 12.721 (7.371) 0.946 (0.009) [0.08] 16.448 (8.436) 14.713 (7.128) 0.947 (0.009) [0.51] 0.659 (0.378) 1.204 (0.038) -0.205 (0.042) -0.020 (0.027) -0.038 (0.033) [ 0.478] 11.810 (7.829) 10.500 (6.653) 0.944 (0.009) [0.95] 0.781 (0.455) 0.938 (0.011) 35.587 (13.998) 17.791 (5.649) 0.973 (0.006) 0.887 (0.245) 1.233 (0.039) -0.214 (0.043) -0.021 (0.028) -0.039 (0.034) [ 0.565] 22.008 (7.740) 17.715 (5.455) 0.960 (0.007) 12.644 (8.282) 10.076 (6.245) 0.938 (0.011) 9.929 (7.258) 8.537 (6.032) 0.941 (0.010) 25.032 (10.581) 20.853 (7.731) 0.953 (0.009) 1.682 (0.352) 1.155 (0.036) -0.122 (0.041) -0.040 (0.024) -0.028 (0.026) [ 0.094] 35.104 (11.140) 29.528 (7.772) 0.952 (0.009) -4.79 [0.00] 6,790 175 -3.89 [0.00] 6,642 175 -4.13 [0.00] 6,336 175 -7.00 [0.00] 5,688 175 6,615 175 6,467 175 6,161 175 5,513 175 6,615 175 6,467 175 6,161 175 5,513 175 log GDP third lag log GDP fourth lag 35 Effect of democracy after 25 years Persistence of GDP process AR2 test p-value Unit root test t−statistics p−value (reject unit root) Observations Countries in sample HHK estimates (2) log GDP second lag p-value lags 5 to 8 Long-run effect of democracy Arellano and Bond estimates (1) Notes: This table presents estimates of the effect of democracy on log GDP per capita. The reported coefficient on democracy is multiplied by 100. Columns 1-4 present results using the within estimator. Columns 5-8 present results using Arellano and Bond’s GMM estimator. The AR2 row reports the p-value for a test of serial correlation in the residuals of the GDP series. Columns 9-12 present results using the HHK estimator. In all specifications we control for a full set of country and year fixed effects. Columns 4, 8 and 12 include 8 lags of GDP per capita as controls, but we only report the p-value of a test for joint significance of lags 5 to 8. Standard errors robust against heteroskedasticity and serial correlation at the country level are reported in parentheses. Table 3: Effect of democracy on the growth rate of GDP per capita. Within estimates Democracy GDP growth first lag (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) 1.028 (0.250) 0.287 (0.037) 1.201 (0.237) 0.274 (0.037) 0.047 (0.024) 1.269 (0.243) 0.263 (0.039) 0.060 (0.025) 0.023 (0.018) -0.033 (0.022) 1.458 (0.385) 0.294 (0.040) 1.715 (0.385) 0.278 (0.038) 0.057 (0.025) 1.545 (0.368) 0.270 (0.041) 0.064 (0.025) 0.031 (0.018) -0.022 (0.022) 1.413 (0.341) 0.230 (0.037) 0.075 (0.023) 1.343 (0.348) 0.211 (0.039) 0.082 (0.025) 0.029 (0.015) -0.007 (0.015) 1.768 (0.372) 43.245 (9.046) 0.321 (0.041) 1.845 (0.370) 45.275 (9.038) 0.312 (0.038) 2.032 (0.511) 49.685 (12.448) 0.305 (0.039) 1.961 (0.527) 47.791 (12.770) 0.315 (0.043) 6,178 175 5,523 175 2.349 (0.602) 57.296 (14.572) 0.342 (0.041) [0.98] 6,003 175 1.867 (0.448) 46.067 (11.017) 0.245 (0.034) 6,490 175 2.579 (0.624) 62.957 (15.105) 0.335 (0.043) [0.81] 6,315 175 1.253 (0.330) 0.211 (0.036) 0.057 (0.020) 0.036 (0.021) -0.004 (0.020) [0.02] 1.790 (0.488) 40.543 (10.837) 0.300 (0.049) 6,642 175 2.065 (0.559) 50.767 (13.718) 0.294 (0.040) [0.10] 6,467 175 1.554 (0.405) 0.254 (0.036) 0.047 (0.024) 0.028 (0.018) -0.013 (0.025) [0.00] 2.273 (0.644) 50.090 (13.322) 0.316 (0.044) [0.93] 5,348 174 1.410 (0.323) 0.245 (0.034) 1.443 (0.360) 35.483 (8.820) 0.287 (0.037) 1.378 (0.284) 0.250 (0.037) 0.042 (0.023) 0.024 (0.017) -0.018 (0.025) [0.00] 1.960 (0.441) 42.715 (9.191) 0.297 (0.043) 6,467 175 6,315 175 6,003 175 5,348 174 GDP growth third lag GDP growth fourth lag 36 Level effect of democracy after 25 years Persistence of growth rate process AR2 test p-value Observations Countries in sample HHK estimates (2) GDP growth second lag p-value lags 5 to 8 Long-run effect of democracy (on growth rate) Arellano and Bond estimates (1) Notes: This table presents estimates of the effect of democracy on the growth rate of GDP per capita. The reported coefficient on democracy is multiplied by 100. Columns 1-4 present results using the within estimator. Columns 5-8 present results using Arellano and Bond’s GMM estimator. The AR2 row reports the p-value for a test of serial correlation in the residuals of the growth rate series. Columns 9-12 present results using the HHK estimator. In all specifications we control for a full set of country and year fixed effects. Columns 4, 8 and 12 include 8 lags of the growth rate of GDP per capita as controls, but we only report the p-value of a test for joint significance of lags 5 to 8. Standard errors robust against heteroskedasticity and serial correlation at the country level are reported in parentheses. Table 4: Effect of democracy on (log) GDP per capita controlling for covariates. (1) GDP in 1960 quintiles × year effects (2) Soviet dummies (3) 0.787 (0.226) 21.240 (7.215) 16.895 (5.297) 0.963 (0.005) 6,336 175 0.718 (0.249) 22.173 (8.702) 16.261 (5.982) 0.968 (0.005) 5,523 149 0.911 (0.251) 24.860 (7.783) 19.587 (5.724) 0.963 (0.005) 6,336 175 0.875 (0.374) 16.448 (8.436) 14.713 (7.128) 0.947 (0.009) [0.51] 6,161 175 0.730 (0.387) 14.865 (8.998) 12.759 (7.350) 0.951 (0.008) [0.90] 5,374 149 1.073 (0.403) 20.006 (8.981) 17.874 (7.564) 0.946 (0.009) [0.28] 6,161 175 1.178 (0.370) 25.032 (10.581) 20.853 (7.731) 0.953 (0.009) 6,161 175 0.722 (0.357) 15.731 (8.476) 12.719 (6.503) 0.954 (0.006) 5,374 149 1.059 (0.364) 21.648 (9.431) 18.313 (7.162) 0.951 (0.009) 6,161 175 Covariates included: Lags of unrest (4) Lags of trade (5) Lags of financial flows (6) Lags of demographic structure (7) Region × regime × year effects (8) 0.650 (0.230) 14.153 (5.419) 12.251 (4.552) 0.954 (0.005) 6,262 172 0.834 (0.264) 16.651 (5.546) 14.532 (4.726) 0.950 (0.005) 6,336 175 1.017 (0.373) 18.607 (7.842) 15.903 (6.327) 0.945 (0.007) [0.34] 4,779 171 0.756 (0.370) 12.152 (6.639) 11.334 (6.004) 0.938 (0.010) [0.58] 6,090 172 1.217 (0.420) 18.209 (6.746) 16.861 (6.050) 0.933 (0.010) [0.70] 6,161 175 2.030 (0.359) 32.631 (7.727) 28.896 (6.223) 0.938 (0.008) 4,779 171 1.262 (0.355) 22.161 (6.641) 19.633 (5.647) 0.943 (0.006) 6,090 172 1.482 (0.449) 26.358 (9.178) 22.776 (7.380) 0.944 (0.008) 6,161 175 Panel A: Within estimates. Democracy Long-run effect of democracy Effect of democracy after 25 years Persistence of GDP process Observations Countries in sample Democracy Long-run effect of democracy Effect of democracy after 25 years Persistence of GDP process AR2 test p-value Observations Countries in sample Democracy Long-run effect of democracy Effect of democracy after 25 years Persistence of GDP process Observations Countries in sample 0.705 (0.224) 17.000 (5.980) 13.567 (4.644) 0.959 (0.004) 5,643 171 Panel B: Arellano and 0.595 0.926 (0.264) (0.244) 14.593 23.870 (7.122) (8.211) 11.500 18.149 (5.336) (5.435) 0.959 0.961 (0.006) (0.006) 5,750 4,950 172 171 Bond estimates. 0.693 1.034 (0.396) (0.469) 9.871 17.926 (6.479) (9.021) 9.159 15.659 (5.768) (7.593) 0.930 0.942 (0.012) (0.009) [0.62] [0.72] 5,467 5,570 171 172 Panel C: HHK estimates. 1.203 (0.376) 25.557 (9.842) 20.753 (7.072) 0.953 (0.008) 5,467 171 1.110 (0.332) 24.575 (9.031) 19.407 (6.359) 0.955 (0.008) 5,570 172 Notes: This table presents estimates of the effect of democracy on log GDP per capita. The reported coefficient of democracy is multiplied by 100. Panel A presents results using the within estimator. Panel B presents results using Arellano and Bond’s GMM estimator. The AR2 row reports the p-value for a test of serial correlation in the residuals of the GDP series. Panel C presents results using the HHK estimator. In all specifications we control for a full set of country and year fixed effects and four lags of GDP per capita. Additionally, we control for the covariates specified in each column label and described in the text. Standard errors robust against heteroskedasticity and serial correlation at the country level are reported in parentheses. Table 5: Semi-parametric estimates of the effect of democratizations on (log) GDP per capita. Average effects from: -5 to -1 years (1) 0 to 4 years (2) 5 to 9 years (3) 10 to 14 years (4) 15 to 19 years (5) 20 to 24 years (6) 25 to 29 years (7) 24.075 (8.262) 21.310 (9.643) 24.111 (9.035) 22.184 (11.561) 23.691 (7.674) 21.793 (9.612) Panel A: Linear regression adjustment. Average effect of democracy on log GDP 0.060 (0.156) 2.454 (1.382) 3.621 (2.792) 7.806 (4.416) 14.037 (5.384) 38 Panel B: Inverse propensity score reweighting. Average effect of democracy on log GDP -1.586 (1.478) 3.724 (1.789) 3.214 (3.327) 6.818 (4.848) 13.542 (5.892) Panel C: Doubly-robust estimator. Average effect of democracy on log GDP 0.051 (0.150) 2.795 (1.478) 2.969 (3.070) 6.966 (4.354) 12.947 (4.886) Notes: This table presents semi-parametric estimates of the effect of a democratization on log GDP per capita over different time horizons, indicated in the column labels. We report estimates of the average effect on the treated. Panel A presents estimates using regression adjustment to compute counterfactual outcomes for treated countries. Panel B presents estimates obtained via inverse propensity score reweighting. Panel C presents estimates obtained using a doubly-robust estimator, combining the regression adjustment and the inverse propensity score reweighting. Below each estimate we report robust standard errors obtained via bootstrapping. Table 6: Instrumental-variables estimates of the effect of democracy on (log) GDP per capita. Covariates included: (1) (2) GDP in 1960 quintiles× year effects (3) Soviet dummies (4) Regional trends (5) Regional GDP & trade (6) Regional unrest GDP & trade (7) Spatial lag of GDP (8) Spatial lags of GDP and democracy (9) Panel A: 2SLS estimates with fixed effects. Democracy Long-run effect of democracy Effect of democracy after 25 years Persistence of GDP process Hansen p-value Observations Countries in sample Exc. Instruments F-stat. Democracy wave t-1 0.966 (0.558) 26.315 (17.075) 20.836 (12.862) 0.963 (0.005) 39 1.149 (0.554) 31.521 (17.425) 24.866 (12.978) 0.964 (0.005) [0.21] 6,309 174 33.2 1.125 (0.689) 35.226 (23.846) 25.618 (16.538) 0.968 (0.005) [0.18] 5,496 148 16.8 1.292 1.697 1.817 (0.651) (0.885) (0.663) 35.723 36.788 41.544 (19.997) (20.657) (17.157) 27.929 32.051 35.350 (14.944) (17.703) (14.017) 0.964 0.954 0.956 (0.005) (0.006) (0.006) [0.32] [0.28] [0.25] 6,309 6,309 6,309 174 174 174 26.7 23.7 13.6 Panel B: First-stage estimates. 1.107 (0.656) 25.016 (16.002) 21.386 (13.342) 0.956 (0.006) [0.09] 6,309 174 16.7 1.335 (0.536) 37.482 (17.836) 29.217 (12.894) 0.964 (0.005) [0.04] 6,181 173 17.5 1.361 (0.895) 38.439 (27.883) 29.011 (19.692) 0.965 (0.006) [0.19] 6,009 173 4.6 0.800 (0.073) 0.547 (0.101) 0.133 (0.081) 0.227 (0.067) -0.087 (0.110) 0.503 (0.130) 0.109 (0.094) 0.270 (0.077) -0.119 (0.126) 0.480 0.498 0.522 (0.099) (0.092) (0.104) 0.133 0.129 0.117 (0.080) (0.081) (0.079) 0.223 0.228 0.221 (0.065) (0.070) (0.069) -0.075 -0.123 -0.083 (0.110) (0.106) (0.113) Panel C: HHK estimates. 0.508 (0.102) 0.115 (0.078) 0.223 (0.070) -0.064 (0.113) 0.540 (0.103) 0.136 (0.078) 0.224 (0.070) -0.072 (0.113) 0.586 (0.101) 0.128 (0.088) 0.282 (0.077) -0.107 (0.116) 0.690 (0.642) 14.512 (14.703) 11.768 (11.445) 0.952 (0.011) 6,161 174 0.944 (0.479) 24.766 (14.083) 18.670 (9.799) 0.962 (0.008) 6,161 174 1.435 (0.599) 46.767 (22.556) 31.039 (13.113) 0.969 (0.008) 5,374 148 0.719 (0.503) 18.337 (13.688) 13.969 (9.935) 0.961 (0.009) 6,161 174 0.897 (0.371) 17.290 (8.556) 14.668 (6.734) 0.948 (0.010) 6,161 174 1.021 (0.549) 29.286 (18.354) 21.133 (11.942) 0.965 (0.009) 6,132 173 1.206 (0.485) 31.111 (15.167) 23.702 (10.243) 0.961 (0.008) 5,960 173 6,312 174 119.1 Democracy wave t-2 Democracy wave t-3 Democracy wave t-4 Democracy Long-run effect of democracy Effect of democracy after 25 years Persistence of GDP process Observations Countries in sample 0.822 (0.480) 16.413 (10.700) 13.778 (8.523) 0.950 (0.010) 6,161 174 1.311 (0.435) 24.040 (9.989) 21.100 (8.038) 0.945 (0.010) 6,161 174 Notes: This table presents IV estimates of the effect of democracy on log GDP per capita. The reported coefficient of democracy is multiplied by 100. Panel A presents 2SLS estimates instrumenting democracy with up to four lags of regional democracy waves and the p-value of a Hansen overidentification test. Panel B presents the corresponding first stage estimates and the excluded instruments F statistic. Panel C presents results using the HHK estimator instrumenting democracy with up to four lags of regional democracy waves (except for column 1, where we only use one lag). In all specifications we control for a full set of country and year fixed effects and four lags of GDP per capita. Additionally, we control for the covariates specified in each column label and described in the text. Standard errors robust against heteroskedasticity and serial correlation at the country level are in parentheses. Table 7: Effects of democracy on potential mechanisms. Dependent variable: Log of investment share in GDP (1) Log of TFP (2) Index of economic reforms (3) Log of trade share in GDP log of tax share in GDP Log of primary enrollment (4) (5) (6) Log of secondary enrollment (7) log of child mortality Dummy for unrest (8) (9) Panel A: Within estimates. Democracy Long-run effect of democracy Effect of democracy after 25 years Persistence of outcome process Observations Countries in sample Democracy Long-run effect of democracy 40 Effect of democracy after 25 years Persistence of outcome process Exc. instruments F-stat. Hansen p-value Observations Countries in sample Democracy 2.391 (1.114) 9.112 (4.255) 9.089 (4.245) 0.738 (0.020) 5,665 169 -0.205 (0.276) -2.883 (3.858) -2.738 (3.648) 0.929 (0.012) 3,879 107 0.687 (0.348) 5.580 (2.883) 5.359 (2.753) 0.877 (0.012) 4,692 150 0.689 (0.676) 5.445 (5.253) 5.303 (5.126) 0.873 (0.011) 5,738 172 3.311 (1.409) 16.062 (6.650) 15.864 (6.574) 0.794 (0.016) 4,511 131 Panel B: 2SLS estimates. 1.042 (0.338) 21.908 (7.624) 18.892 (6.321) 0.952 (0.008) 3,714 166 1.345 (0.610) 18.960 (8.622) 18.057 (8.146) 0.929 (0.013) 2,883 158 -0.253 (0.063) -34.264 (10.747) -21.400 (5.124) 0.993 (0.001) 6,084 173 -7.832 (2.185) -11.944 (3.329) -11.944 (3.329) 0.344 (0.030) 5,646 171 2.211 (2.852) 8.440 (10.705) 8.419 (10.681) 0.738 (0.020) 21.7 [0.29] 5,640 168 -0.941 (0.667) -12.738 (8.854) -12.167 (8.380) 0.926 (0.012) 27.7 [0.06] 3,871 107 3.224 (0.863) 23.775 (6.215) 23.156 (6.039) 0.864 (0.012) 43.7 [0.22] 4,670 149 5.512 (2.005) 40.589 (13.580) 39.817 (13.375) 0.864 (0.012) 21.5 [0.09] 5,714 171 8.088 (3.021) 38.609 (14.330) 38.159 (14.121) 0.791 (0.017) 31.8 [0.69] 4,489 130 Panel C: HHK estimates. 1.757 (0.721) 36.693 (15.505) 31.611 (12.863) 0.952 (0.008) 12.1 [0.09] 3,710 164 4.116 (1.626) 57.072 (21.698) 54.252 (20.267) 0.928 (0.013) 10.4 [0.12] 2,879 156 -0.715 (0.164) -95.728 (26.347) -58.625 (13.123) 0.993 (0.001) 26.3 [0.02] 6,057 172 -5.569 (5.682) -8.471 (8.577) -8.471 (8.577) 0.343 (0.030) 28.6 [0.84] 5,619 170 6.603 0.388 1.121 1.255 4.277 1.384 2.144 -0.306 -3.638 (1.336) (0.294) (0.371) (0.790) (2.044) (0.366) (0.644) (0.068) (2.931) Long-run effect of democracy 25.495 7.518 22.655 10.182 24.622 41.349 43.070 -54.798 -5.742 (5.313) (6.011) (11.199) (6.584) (11.858) (14.855) (15.445) (15.745) (4.630) Effect of democracy after 25 years 25.432 6.748 15.698 9.807 23.966 29.049 36.865 -29.139 -5.742 (5.294) (5.366) (5.953) (6.307) (11.461) (8.614) (11.888) (6.131) (4.630) Persistence of outcome process 0.741 0.948 0.951 0.877 0.826 0.967 0.950 0.994 0.366 (0.018) (0.009) (0.018) (0.014) (0.031) (0.007) (0.012) (0.001) (0.037) Observations 5,125 3,557 4,236 4,866 4,045 3,579 2,683 5,454 5,233 Countries in sample 168 107 149 171 130 164 156 172 170 Notes: This table presents estimates of the effect of democracy on the different channels specified in the columns labels. The reported coefficient of democracy is multiplied by 100 (except for columns 3 and 9). Panel A presents within estimates. Panel B presents 2SLS estimates instrumenting democracy with four lags of regional democracy waves, the F statistic for the excluded instruments and the p-value of Hansen’s overidentification test. Panel C presents results using the HHK estimator instrumenting democracy with four lags of regional democracy. In all specifications we control for a full set of country and year fixed effects, four lags of GDP per capita and four lags of the dependent variable. Standard errors robust against heteroskedasticity and serial correlation at the country level are in parentheses. Table 8: Heterogeneous effects of democracy on (log) GDP per capita. Interaction with: Measured at: Log GDP per capita: Share with secondary: 1960 (1) 1970 (2) 1980 (3) Lagged (4) 0.432 (0.275) 0.001 (0.002) 16.231 (11.160) 10.013 (6.565) 0.973 (0.005) 4,281 93 0.572 (0.248) 0.001 (0.001) 18.631 (9.073) 12.916 (5.960) 0.969 (0.005) 4,909 109 0.687 (0.248) 0.002 (0.002) 20.489 (8.608) 14.985 (5.848) 0.966 (0.005) 5,525 131 0.744 (0.246) 0.001 (0.002) 19.843 (8.255) 15.877 (5.943) 0.963 (0.006) 6,336 175 Panel B: 2SLS 0.500 (1.088) -0.002 (0.005) 18.838 (43.554) 11.592 (25.784) 0.973 (0.006) 6.6 [0.81] 4,273 93 0.155 (0.961) 0.000 (0.004) 4.978 (31.473) 3.486 (21.795) 0.969 (0.006) 6.1 [0.73] 4,901 109 0.645 (0.929) -0.000 (0.004) 19.275 (30.208) 14.078 (21.085) 0.967 (0.006) 7.0 [0.54] 5,517 131 1.326 (0.887) -0.003 (0.004) 36.116 (29.900) 28.377 (21.317) 0.963 (0.008) 14.0 [0.33] 6,153 174 Panel C: HHK 0.222 (0.379) 0.004 (0.003) 7.692 (13.442) 4.869 (8.286) 0.971 (0.009) 4,180 93 0.234 (0.401) -0.000 (0.003) 7.453 (13.213) 5.084 (8.850) 0.969 (0.008) 4,792 109 0.144 (0.445) 0.001 (0.004) 4.480 (14.002) 3.054 (9.435) 0.968 (0.008) 5,386 131 1960 (5) 1970 (6) 1980 (7) Lagged (8) 0.446 (0.254) 0.046 (0.028) 13.785 (8.550) 10.081 (5.964) 0.968 (0.005) 5,300 138 estimates. 0.340 (0.253) 0.049 (0.020) 10.480 (8.275) 7.679 (5.872) 0.968 (0.005) 5,300 138 0.385 (0.246) 0.038 (0.014) 11.841 (8.118) 8.687 (5.728) 0.967 (0.005) 5,300 138 0.495 (0.241) 0.020 (0.013) 14.597 (8.432) 10.953 (5.821) 0.966 (0.006) 5,300 138 -0.119 (0.662) 0.174 (0.060) -3.649 (19.968) -2.692 (14.837) 0.967 (0.006) 18.5 [0.44] 5,292 138 estimates. -0.484 (0.665) 0.156 (0.047) -14.586 (19.023) -10.843 (14.524) 0.967 (0.006) 17.6 [0.41] 5,292 138 -0.474 (0.639) 0.116 (0.033) -14.135 (18.114) -10.574 (13.901) 0.966 (0.006) 16.0 [0.25] 5,292 138 0.600 (0.576) 0.049 (0.023) 17.373 (18.629) 13.133 (13.312) 0.965 (0.006) 12.4 [0.50] 5,218 138 1.101 (0.686) 0.093 (0.046) 31.605 (21.502) 23.787 (15.084) 0.965 (0.008) 5,154 138 0.887 (0.679) 0.089 (0.037) 25.022 (20.748) 19.159 (14.981) 0.965 (0.008) 5,154 138 0.790 (0.638) 0.058 (0.028) 22.375 (19.522) 17.091 (14.107) 0.965 (0.008) 5,154 138 1.713 (0.584) 0.016 (0.013) 49.338 (23.950) 36.069 (14.116) 0.965 (0.009) 5,154 138 Panel A: Within estimates. Democracy Interaction Long-run effect of democracy Effect of democracy after 25 years Persistence of GDP process Observations Countries in sample Democracy Interaction Long-run effect of democracy Effect of democracy after 25 years Persistence of GDP process Exc. instruments F-stat. Hansen p-value Observations Countries in sample Democracy Interaction Long-run effect of democracy Effect of democracy after 25 years Persistence of GDP process Observations Countries in sample 1.619 (0.477) 0.002 (0.004) 48.375 (21.975) 34.304 (11.965) 0.967 (0.009) 6,110 174 Notes: This table presents estimates of the effect of democracy on log GDP per capita and its interaction with other country characteristics indicated in the columns’ headers. The reported coefficients of democracy and the interaction are multiplied by 100. We report main effects and long-run effects evaluated at the 25th percentile of the interacted variable. Panel A presents within estimates. Panel B presents 2SLS estimates instrumenting democracy (and the interaction term) with four lags of regional democracy waves. It also reports the F statistic for the excluded instruments and the p-value of Hansen’s overidentification test. Panel C presents results using the HHK estimator instrumenting democracy (and the interaction term) with four lags of regional democracy waves. In all specifications we control for a full set of country and year fixed effects and four lags of GDP per capita. Standard errors robust against heteroskedasticity and serial correlation at the country level are in parentheses. −5 Change in GDP per capita log points 0 5 10 15 20 25 Figure 1: GDP per capita before and after a democratization. −15 −10 −5 0 5 10 15 Years around democratization 20 25 30 Notes: This figure plots GDP per capita in log points around a democratic transition. We normalize log GDP per capita to zero in the year preceding the democratization. Time (in years) relative to the year of democratization runs on the horizontal axis. 0 Change in GDP per capita log points 10 20 30 Figure 2: Dynamic panel model estimates of the over-time effects of democracy on the log of GDP per capita. 0 5 10 15 20 Years around democratization 25 30 Notes: This figure plots the estimated change in the log of GDP per capita caused by a permanent transition to democracy. The effects are obtained by forward iteration of the estimated process for GDP modeled in equation (1). A 95% confidence interval obtained using the delta method is presented in dotted lines. Time (in years) relative to the year of democratization runs on the horizontal axis. −20 Change in GDP per capita log points −10 0 10 20 30 40 Figure 3: Semi-parametric estimates of the over-time effects of democracy on the log of GDP. Estimates obtained using a regression model to estimate counterfactuals. −15 −10 −5 0 5 10 15 Years around democratization 20 25 30 Notes: This figure plots semi-parametric estimates of the effect of democratizations on GDP per capita in log points. The solid line plots the estimated average effect on GDP per capita on countries that democratized (in log points), together with a 95% confidence interval in dashed lines. Time (in years) relative to the year of democratization runs on the horizontal axis. The estimates are obtained by assuming and estimating a linear model for counterfactual outcomes, which we use to control for the influence of GDP dynamics. Section 4 explains our approach in full detail. −20 Change in GDP per capita log points −10 0 10 20 30 40 Figure 4: Semi-parametric estimates of the over-time effects of democracy on the log of GDP. Estimates obtained using inverse-propensity score reweighting. −15 −10 −5 0 5 10 15 Years around democratization 20 25 30 Notes: This figure plots semi-parametric estimates of the effect of democratizations on GDP per capita in log points. The solid line plots the estimated average effect on GDP per capita on countries that democratized (in log points), together with a 95% confidence interval in dashed lines. Time (in years) relative to the year of democratization runs on the horizontal axis. The estimates are obtained by assuming and estimating a probit model for democratizations based on GDP lags, which we use to estimate the propensity score and reweight the data. Section 4 explains our approach in full detail. −20 Change in GDP per capita log points −10 0 10 20 30 40 Figure 5: Semi-parametric estimates of the over-time effects of democracy on the log of GDP. Doubly-robust estimates. −15 −10 −5 0 5 10 15 Years around democratization 20 25 30 Notes: This figure plots semi-parametric estimates of the effect of democratizations on GDP per capita in log points. The solid line plots the estimated average effect on GDP per capita on countries that democratized (in log points), together with a 95% confidence interval in dashed lines. Time (in years) relative to the year of democratization runs on the horizontal axis. The estimates are obtained by assuming and estimating a probit model for democratizations based on GDP lags, which we use to estimate the propensity score and reweight the data. In addition, we partial out lags of GDP linearly, making our approach doubly robust. Section 4 explains our approach in full detail. 0 .1 Mean democracy .2 .3 .4 .5 Figure 6: Regional democratizations and reversal waves. −10 −5 0 5 10 15 Years around first democratization in the region 20 .8 .85 Mean democracy .9 .95 1 Initial nondemocracies in the region Initial nondemocracies in other regions −10 −5 0 5 10 Years around first reversal in the region 15 20 Initial democracies in the region Initial democracies in other regions Notes: These figures illustrate the existence of regional democracy waves. The top figure plots average democracy among initial nondemocracies around the first democratization in the region. For comparison it also plots average democracy among other initial nondemocracies in other regions. The bottom figure plots average democracy among initial democracies around the first reversal in the region. For comparison it also plots average democracy among other initial democracies in other regions.