Electronic Supplementary Material:
Table A: State Strength Values for Every Country-Year, by Continent
Country / Strong State (Years) / Weak State (Years) / Country / Strong State (Years) / Weak State (Years)Algeria / 1955-1990 / Suriname / 1964-1989 / 1955-1963, 1990
Angola / 1955-1979 / 1980-1990 / Uruguay / 1963-1972, 1975-1990 / 1955-1962, 1973-1974
Benin / 1955-1978, 1989-1990 / 1979-1988 / Venezuela / 1955-1959, 1961-1990 / 1960
Botswana / 1955, 1964-1979, 1981-1990 / 1956-1963, 1980 / Bahrain / 1955-1990
Burkina Faso / 1975-1990 / 1955-1975 / Bangladesh / 1955-1970, 1974-1990 / 1950, 1971-1973
Burundi / 1978-1988, 1990 / 1955-1978, 1989 / Bhutan / 1969-1978, 1981-1990 / 1955-1968, 1979-1980
Cameroon / 1976-1985, 1987-1990 / 1955-1975, 1986 / China / 1955-1962, 1964-1973, 1982-1990 / 1963, 1974-1981
Cape Verde / 1955-1990 / Hong Kong / 1961-1990 / 1955-1960
Cent Af Rep / 1975-1984, 1988-1990 / 1955-1974, 1985-1987 / India / 1955-1990
Chad / 1956-1975, 1979-1989 / 1955, 1976-1979, 1990 / Indonesia / 1961-1990 / 1955-1960
Comoros / 1956-1975, 1980-1989 / 1955, 1976-1979, 1990 / Iran / 1956-1990 / 1955
Congo-Braz. / 1984-1990 / 1955-1983 / Iraq / 1955-1974, 1977-1990 / 1975-1976
Djibouti / 1967-1976, 1983-1990 / 1955-1966, 1977-1982 / Israel / 1955-1957, 1961-1970, 1972-1981, 1983-1990 / 1958-1960, 1971, 1982
Egypt / 1955-1956, 1960-1969, 1976-1990 / 1957-1959, 1970-1975 / Japan / 1955-1990
Ethiopia / 1984-1990 / 1955-1983 / Jordan / 1955-1970, 1979-1988 / 1971-1978, 1989-1990
Gabon / 1961-1989 / 1955-1960, 1990 / South Korea / 1955-1990
The Gambia / 1963-1990 / 1955-1962 / Kuwait / 1957-1990 / 1955-1956
Ghana / 1955-1957, 1960-1979, 1984-1990 / 1958-1959, 1980-1983 / Laos / 1985-1990 / 1955-1984
Guinea / 1983-1990 / 1955-1982 / Malaysia / 1970-1989 / 1955-1969, 1990
Guinea-Bissau / 1955-1988 / 1989-1990 / Mongolia / 1956-1990 / 1955
Ivory Coast / 1975-1984, 1988-1990 / 1955-1974, 1985-1987 / Myanmar / 1973-1990 / 1955-1972
Kenya / 1955-1957, 1962-1990 / 1958-1961 / Nepal / 1955-1990
Lesotho / 1955-1990 / Oman / 1955-1990
Liberia / 1962-1971, 1974-1990 / 1955-1961, 1972-1973 / Pakistan / 1955-1970, 1972-1990 / 1971
Madagascar / 1975-1984 / 1955-1974, 1985-1990 / Phillipines / 1955-1957, 1960-1990 / 1958-1959
Malawi / 1966-1975, 1977-1990 / 1955-1965, 1976 / Qatar / 1970-1979, 1986-1990 / 1955-1969, 1980-1985
Mali / 1976-1985, 1987-1990 / 1955-1975, 1986 / Saudi Arabia / 1974-1983 / 1955-1973, 1984-1990
Mauritania / 1977-1986, 1988-1990 / 1955-1976, 1987 / Singapore / 1955-1966, 1970-1990 / 1967-1969
Mauritius / 1955-1981, 1983-1990 / 1982 / Sri Lanka / 1955, 1963-1990 / 1956-1962
Morocco / 1960-1969, 1971-1980, 1982-1990 / 1955-1959, 1970, 1981 / Syria / 1960-1979, 1981-1990 / 1955-1959, 1980
Mozambique / 1955-1989 / 1990 / Thailand / 1955-1956, 1960-1990 / 1957-1959
Namibia / 1960-1979, 1981-1990 / 1955-1959, 1980 / U.A.E. / 1968-1990 / 1955-1967
Niger / 1977-1986, 1988-1990 / 1955-1976, 1987 / Yemen / 1955-1990
Nigeria / 1955-1961, 1963-1972 / 1962, 1973-1990 / Austria / 1955-1990
Rwanda / 1978-1987 / 1955-1977, 1988-1990 / Belgium / 1955-1956, 1961-1979, 1981-1990 / 1957-1960, 1980
Senegal / 1976-1985, 1988-1990 / 1955-1975, 1986-1987 / Bulgaria / 1955-1990
Seychelles / 1955-1956, 1960-1969, 1971-1990 / 1957-1959, 1970 / Cyprus / 1955, 1960-1969, 1982-1990 / 1956-1959, 1970-1981
Sierra Leone / 1963-1972, 1974-1983, 1985-1990 / 1955-1962, 1973, 1984 / Czech. / 1955-1959, 1961-1989 / 1960, 1990
Somalia / 1975-1984, 1987-1990 / 1955-1975, 1985-1986 / Denmark / 1955-1990
South Africa / 1955-1990 / Finland / 1955-1990
Sudan / 1955-1964, 1973-1990 / 1965-1972 / France / 1955-1990
Swaziland / 1955-1990 / West Germany / 1955-1959, 1961-1979, 1987-1989 / 1960, 1980-1986
Tanzania / 1967-1976, 1978-1990 / 1955-1966, 1977 / Greece / 1955-1990
Togo / 1970-1979, 1981-1990 / 1955-1969, 1980 / Hungary / 1955-1958, 1960-1990 / 1959
Tunisia / 1955-1984 / 1985-1990 / Iceland / 1955-1990
Uganda / 1955-1957, 1959-1989 / 1958, 1990 / Italy / 1955-1990
Zaire (DRC) / 1974-1990 / 1955-1973 / Luxembourg / 1955-1956, 1960-1979, 1981-1990 / 1957-1959, 1980
Zambia / 1955-1959, 1963-1978, 1980-1990 / 1960-1962, 1979 / Malta / 1955-1976, 1985-1990 / 1977-1984
Zimbabwe / 1962-1978, 1982-1990 / 1955-1961, 1979-1981 / Netherlands / 1955-1956, 1960-1969, 1971-1980 / 1957-1959, 1970, 1981-1990
Bahamas / 1955-1990 / Norway / 1955-1990
Barbados / 1955, 1960-1990 / 1956-1959 / Poland / 1955-1990
Belize / 1960-1989 / 1955-1959, 1990 / Portugal / 1955-1979, 1981-1990 / 1980
Canada / 1955-1990 / Romania / 1955-1975, 1977-1986 / 1976, 1987-1990
Costa Rica / 1955-1960, 1963-1982, 1984-1990 / 1961-1962, 1983 / Spain / 1955-1979, 1981-1990 / 1980
Dominica / 1955, 1960-1979, 1981-1990 / 1956-1959, 1980 / Sweden / 1955-1990
Dom Rep / 1955-1979, 1981-1990 / 1980 / Switzerland / 1955-1990
El Salvador / 1955-1959, 1961-1980 / 1960, 1981-1990 / Turkey / 1955-1990
Grenada / 1955, 1960-1979, 1981-1990 / 1956-1959, 1980 / United Kingdom / 1955-1990
Guatemala / 1955-1959, 1964-1990 / 1960-1963 / Yugoslavia / 1955-1990
Haiti / 1955-1959, 1971-1980, 1982-1990 / 1960-1970, 1981 / Australia / 1955-1990
Honduras / 1955-1959, 1961-1970, 1974-1983, 1988-1990 / 1960, 1971-1973, 1984-1987 / Fiji / 1955-1990
Jamaica / 1955-1979, 1982-1990 / 1980-1981 / New Zealand / 1955-1990
Mexico / 1955-1990 / P.N.G. / 1966-1990 / 1955-1965
Nicaragua / 1955-1959, 1963-1981 / 1960-1962, 1982-1990 / Solomon Isl. / 1970-1990 / 1955-1969
Panama / 1955-1990 / Tonga / 1956-1990 / 1955
St Kitts & Nevis / 1955, 1960-1989 / 1956-1959, 1990 / Vanuatu / 1967-1976 / 1955-1966, 1977-1990
St Lucia / 1955, 1960-1989 / 1956-1959, 1990
St Vincent / 1955, 1960-1989 / 1956-1959, 1990
Trin & Tobago / 1955, 1960-1990 / 1956-1959
United States / 1955-1990
Argentina / 1955-1956, 1960-1989 / 1957-1959, 1990
Bolivia / 1955-1959, 1976-1985 / 1960-1975, 1986-1990
Brazil / 1955-1989 / 1990
Chile / 1955-1979, 1982-1990 / 1980-1981
Colombia / 1955-1959, 1964-1982, 1985-1990 / 1960-1963, 1983-1984
Ecuador / 1955-1959, 1962-1971, 1974-1990 / 1960-1961, 1972-1973
Guyana / 1955, 1960-1989 / 1956-1959, 1990
Paraguay / 1955-1959, 1962-1990 / 1960-1961
Peru / 1961-1970, 1972-1990 / 1955-1960, 1971
These codings are based on the methods described in the paper; they indicate, based on data drawn from the United States Census Bureau, for each country-year whether a full national census or a component of an annual rolling census has been conducted in the previous ten years. Although census data in this table covers country-years of colonial rule, those cases are not included in the statistical analyses in the paper.
Table B: Binary Time Series Cross-Sectional Analysis:
Using the BTSCS method described in Beck, Katz, & Tucker (1998) to analyze time-dependent cross-sectional data with a binary dependent variable produces results very similar to those found in the binary probit analysis used in Boix (2003) and in my replications. A variable for years since last regime change (Regime Duration) is added, and splines for regime duration are added to the analysis to check for non-linear relationships. The key variable of interest is the interaction between gini and census.
Coefficient / Standard ErrorConstant / -0.4371 / 3.1731
Census / 1.7215 / 3.1600
Lagged Gini / 0.0011 / 0.0664
Gini * Census / -0.0337 / 0.0688
Regime Duration / -0.6011*** / 0.0747
Regime Spline 1 / -0.0005*** / 0.0002
Regime Spline 2 / -0.0036*** / 0.0006
Regime Spline 3 / 0.0035*** / 0.0006
Catholic % / 0.0009 / 0.0027
Protestant % / 0.0448** / 0.0206
Muslim % / 0.0080* / 0.0043
Eth Div / 0.0623 / 0.0458
Rel Fract / 1.4259** / 0.6849
Agri share of GDP / -0.0136 / 0.0121
Lagged Growth / 0.3198 / 1.7202
Log-pseudolikelihood / -114.3225
0.8414
0.0000
1071
P>chi square
Pseudo R sq.
# obs
*: Significant at 90% threshold
**: Significant at 95% threshold
***: Significant at 99% threshold
Since the effect on regime type of the interaction between gini and census cannot be readily interpreted from the regression table, I follow Braumoeller (2004) and graph it for each measure of state strength. This figure (below) shows that here, as in the analyses presented in the paper, GINI has the expected negative effect on transitions when C10=1 (strong state cases), but no effect in weak state cases (when C10=zero).
Coefficients and 95% confidence intervals on lagged GINI by CENSUS10, BTSCS analysis
Table C: Split-Sample Analysis:
An alternative method for assessing the effect of state strength on the relationship between inequality and democracy is a split-sample analysis. I use the SIDD specification of inequality since it generates many more cases of weak states, and replicate the analysis in Table Three. Results in Table C, below, can be interpreted directly since no interaction effects are included. They show, as in the other analyses, that inequality has no effect on regime outcomes in the weak state sample, but effects consistent with Boix (2003) in the strong state sample.
Weak States / Strong StatesConstant / -6.238
(5.174) / 13.612
(8.424) / -1.105
(0.692) / 3.679**
(1.218)
SIDD / 0.016
(0.052) / -0.099
(0.116) / -0.045**
(0.011) / 0.042*
(0.023)
Agri as % GDP / -0.011
(0.017) / -0.104
(0.068) / -0.004
(0.007) / -0.005
(0.015)
Catholic % / 0.090
(0.063) / -0.575**
(0.271) / 0.005
(0.003) / -0.006
(0.006)
Protestant % / 0.044
(0.047) / -0.075
(0.082) / -0.0009
(0.016) / 0.010
(0.017)
Muslim % / 0.079
(0.062) / -0.370**
(0.188) / -0.003
(0.004) / -0.005
(0.007)
Rel. Fract. / -4.629*
(2.730) / 57.596**
(25.181) / 1.530**
(0.631) / -1.336
(0.960)
Ethnic Div. / -0.316
(0.860) / 3.105
(2.187) / 0.418
(0.418) / -1.502**
(0.741)
Growthrate / 0.643
(1.262) / -1.490
(5.669) / -2.578*
(1.410) / 8.411**
(1.973)
Log-likelihood
P>chi square
Pseudo R sq.
# obs / -20.4204
0.0000
0.7847
189 / -145.1634
0.0000
0.8638
1559
Alternative model:
The paper is centered on a replication of model 3A from Boix (2003, Chapter Two). We might be concerned that the model excludes a key determinant of regime dynamics: per capita wealth. To assuage this concern, I replicate model 1A below, showing that state capacity mediates the effect of inequality on regime outcomes in this model as well. The table below shows the results from Boix (2003) in the leftmost columns, the replication in the center columns, and includes the census variable (along with its interaction with the inequality measure) in the rightmost columns.
VARIABLE / Boix Results / Boix Replicated / Census AddedBeta / Alpha / Beta / Alpha / Beta / Alpha
Constant / -3.350**
(1.040) / -17.533*
(9.004) / -3.118***
(0.913) / -15.861**
(6.333) / -4.724
(3.681) / -15.616**
(6.053)
Gini / -0.019
(0.013) / 0.040
(0.055) / -0.033**
(0.016) / 0.057
(0.055) / 0.001
(0.073) / 0.489
(0.053)
GDP p/c / 0.082
(0.074) / 1.953**
(0.816) / 0.079
(0.063) / 1.681**
(0.635) / 0.012
(0.048) / 1.672**
(0.623)
Catholic % of Population / 0.002
(0.004) / 0.047**
(0.021) / 0.004
(0.004) / 0.040**
(0.015) / 0.004
(0.004) / 0.040**
(0.014)
Protestant % of Population / 0.055**
(0.022) / 0.377
(0.350) / 0.053**
(0.021) / 0.235
(0.259) / 0.054**
(0.021) / 0.259
(0.261)
Muslim % of Population / 0.002
(0.005) / 1.407***
(0.542) / 0.003
(0.005) / 1.227***
(0.349) / 0.003
(0.005) / 1.218***
(0.384)
Religious Fractionalization / 2.220**
(1.046) / 10.937*
(5.751) / 2.741**
(1.002) / 9.674**
(4.518) / 2.782**
(1.032) / 9.521**
(4.495)
Ethnic Division / 0.380
(0.594) / -4.341**
(2.183) / 0.464
(0.670) / -3.561**
(1.508) / 0.540
(0.687) / -3.555*
(2.074)
Growth rate (lag) / -0.022
(0.022) / 0.092
(0.052) / -3.506*
(2.050) / 13.082***
(3.876) / -3.552*
(2.074) / 13.098***
(3.811)
Census / 1.630
(3.583) / 0.255
(0.733)
Census*Gini / -0.036
(0.076)
Log-likelihood
P>Chi Sq
Pseudo R sq
# obs / -76.50
0.0000
0.8887
1042 / -72.41
0.0000
0.9002
1080 / -71.87
0.0000
0.8998
1067
Once again, the use of interaction effects means that the coefficients above cannot be directly interpreted. Using the code in Braumoeller (2004), I find that when the state is weak, the effect of inequality on regime type is not significantly different from zero (mean=0.006, 90% CI is {-0.402, 0.515}. But when the state is strong, inequality has a significant and negative effect on regime type (mean=-0.043, 90% CI is {-0.059, -0.264}. Thus, here too, inequality only has a significant effect on regime type where sufficient state capacity is present.