Determinants of Democracy

The Harvard community has made this article openly available.

Please share how this access benefits you. Your story matters.

Citation

Barro, Robert J. 1999. Determinants of democracy. Journal of

Political Economy 107(S6): 158-183.

Published Version

doi:10.1086/250107

Accessed

February 25, 2015 6:53:49 PM EST

Citable Link

http://nrs.harvard.edu/urn-3:HUL.InstRepos:3451297

Terms of Use

This article was downloaded from Harvard University's DASH

repository, and is made available under the terms and conditions
applicable to Other Posted Material, as set forth at
http://nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-ofuse#LAA

(Article begins on next page)

Determinants of Democracy

Robert J. Barro
Harvard University

A panel study of over 100 countries from 1960 to 1995 finds that
improvements in the standard of living predict increases in democracy, as measured by a subjective indicator of electoral rights. The
propensity for democracy rises with per capita GDP, primary
schooling, and a smaller gap between male and female primary
attainment. For a given standard of living, democracy tends to fall
with urbanization and with a greater reliance on natural resources.
Democracy has little relation to country size but rises with the
middle-class share of income. The apparently strong relation of
democracy to colonial heritage mostly disappears when the economic variables are held constant. Similarly, the allowance for
these economic variables weakens the interplay between democracy and religious affiliation. However, negative effects from Muslim and nonreligious affiliations remain intact.

An expansion of political freedommore democracyhas opposing effects on economic growth. On the positive side, democratic
institutions provide a check on governmental power and thereby
limit the potential of public officials to amass personal wealth and
to carry out unpopular policies. But on the negative side, more democracy encourages rich-to-poor redistributions of income and may
enhance the power of interest groups. Consequently, the net effect
of democracy on growth is uncertain. (See Sirowy and Inkeles [1990]
and Przeworski and Limongi [1993] for surveys of theories that relate democracy to economic growth.)
My previous cross-country empirical work, as summarized in Barro
(1997), finds a nonlinear effect of democracy on growth. Growth is
initially increasing in an index of electoral rights, but the relation
[ Journal of Political Economy, 1999, vol. 107, no. 6, pt. 2]
1999 by The University of Chicago. All rights reserved. 0022-3808/99/10706S-0010$02.50

S158

determinants of democracy

S159

turns negative once a moderate amount of rights has been attained.

One way to interpret these results is that, in the worst dictatorships,
an increase in democracy tends to stimulate growth because the benefit from limitations on governmental power is the key matter. But
in places that have already achieved a moderate amount of democracy, a further increase impairs growth because the dominant effect
comes from the intensified concern with social programs that redistribute resources.
The present analysis focuses on the reverse channel, that is, the
impact of economic development on a countrys propensity to experience democracy. A common view since Lipsets (1959) research is
that prosperity stimulates democracy; this idea is often called the
Lipset hypothesis. Lipset credits the idea to Aristotle: From Aristotle down to the present, men have argued that only in a wealthy
society in which relatively few citizens lived in real poverty could a
situation exist in which the mass of the population could intelligently participate in politics and could develop the self-restraint necessary to avoid succumbing to the appeals of irresponsible demagogues (p. 75). (For a statement of Aristotles views, see Aristotle
[1932, book 6].)
Theoretical models of the effect of economic conditions on the
extent of democracy are not well developed. Lipset (1959, pp. 83
84) emphasized increased education and an enlarged middle class
as key elements, and he also stressed Tocquevilles (1835) idea that
private organizations and institutions are important as checks on
centralized government power. This point has been extended by
Putnam (1993), who argues that the propensity for civic activity is
the key underpinning of good government in the regions of Italy.
For Huber, Rueschemeyer, and Stephens (1993, pp. 7475), the crucial concept is that capitalist development lowers the power of the
landlord class and raises the power and ability to organize of the
working and middle classes.
In some models, an autocrat would voluntarily relinquish some
authorityfor example, by establishing a constitution, empowering
a legislature, expanding voting rights, and extending civil liberties
in order to deter revolution and to encourage the private sector to
invest (and, thereby, to expand the pie that the government can
tax). Boone (1996) develops a model along these lines and determines the equilibrium amount of freedom by considering the net
benefits of oppression to potential rulers. However, in this type of
setting, most effects turn out to be ambiguous. For example, an
increase in human capital raises the peoples ability to resist oppression but also raises the rulers benefits from subjugating them.

S160

journal of political economy

Similarly, a rise in urbanization makes it easier for people to meet

and communicatewhich presumably makes them harder to suppressbut also makes it easier for an autocrat to monitor and control activities.
Despite the lack of clear predictions from theoretical models, the
cross-country evidence examined in the present study confirms that
the Lipset/Aristotle hypothesis is a strong empirical regularity. In
particular, increases in various measures of the standard of living
forecast a gradual rise in democracy. In contrast, democracies that
arise without prior economic developmentsometimes because
they are imposed by former colonial powers or international organizationstend not to last. Given the strength of this empirical regularity, one would think that clear-cut theoretical analyses ought also
to be attainable. (This seems to be a case in which the analysis works
better in practice than in theory.)
I. The Measure of Democracy
The main definition of democracy in the present study is a narrow
one that focuses on the role of elections. This concept accords with
the one adopted by Huntington (1991, p. 6): The central procedure of democracy is the selection of leaders through competitive
elections by the people they govern. His inspiration for this procedural definition of democracy comes from Schumpeter (1947, p.
269): The democratic method is that institutional arrangement for
arriving at political decisions in which individuals acquire the power
to decide by means of a competitive struggle for the peoples vote.
More specifically, I use as a measure of democracy the indicator
of electoral rights compiled by Gastil (198283 and subsequent issues) and his followers at Freedom House from 1972 to 1995. A related variable from Bollen (1990) is used for 1960 and 1965.1 The
Freedom House concept of electoral rights uses the following basic
definition: Political rights are rights to participate meaningfully in
the political process. In a democracy this means the right of all adults
to vote and compete for public office, and for elected representatives
1
See Gastil (1991) for a discussion of the methods that underlie his data series.
Inkeles (1991) provides an overview of measurement issues on democracy. He finds
a high degree of agreement produced by the classification of nations as democratic
or not, even when democracy is measured in somewhat different ways by different
analysts (p. x). Bollen (1990) suggests that his measures are reasonably comparable
to Gastils. It is difficult to check comparability directly because the two series do
not overlap in time. Moreover, many countriesespecially those in Africaclearly
experienced major declines in the extent of democracy from the 1960s to the 1970s.
Thus no direct inference about comparability can be made from the higher average
of Bollens figures for the 1960s than for Gastils numbers for the 1970s.

determinants of democracy

S161

to have a decisive vote on public policies (Gastil 198687, p. 7).

In addition to the basic definition, the classification scheme rates
countries (somewhat impressionistically) as less democratic if minority parties have little influence on policy.
Freedom House applied the concept of electoral rights on a
subjective basis to classify countries annually into seven categories;
group one is the highest level of rights and group seven is the lowest.
The classification was made by Gastil and his associates and followers
on the basis of an array of published and unpublished information
about each country. The original ranking from one to seven was
converted here to a scale from zero to one, where zero corresponds
to the fewest rights (Gastils rank seven) and one to the most rights
(Gastils rank one). The scale from zero to one corresponds to the
system used by Bollen.
To fix ideas on the meaning of the zero to one scale, note first
that the United States and most other OECD countries in recent
years received the value 1.0, thereby being designated as full representative democracies. Dictatorships that received the value 0.0 in
1995 included Indonesia, Iraq, Syria, Zaire, and several other countries in Africa. Places that were rated at 0.5halfway between dictatorship and democracyincluded Colombia, Dominican Republic,
Ghana, Guatemala, Malaysia, Mexico, Nicaragua, Paraguay, Senegal,
and Sri Lanka.
The solid line in figure 1 shows the time path of the unweighted
average of the electoral rights index for the countries that I am considering for the years 1960, 1965, and 197295. The number of
countries covered rises from 99 in 1960 to 109 in 1965 and 138 from
1972 to 1995.2 The figure shows that the mean of the index peaked
at 0.66 in 1960, fell to a low point of 0.44 in 1975, and rose subsequently to 0.59 in 1995.
Figure 1 also demonstrates that the main source of the decline in
electoral rights after 1960 was the experience in sub-Saharan Africa.
The dotted line shows that the average of the indicator in subSaharan Africa peaked at 0.58 in 1960 (26 countries) and then (for
43 countries) fell to low points of 0.19 in 1977 and 0.18 in 1989
before rising to 0.40 in 1995. This pattern emerges because many of
the African countries began with ostensibly democratic institutions
when they became independent in the early 1960s, but most evolved
into one-party dictatorships by the early 1970s. (See Bollen [1990]
for further discussion.) The democratization in Africa since 1989
has been substantial; whether it will be sustained is not yet known.
2
The Gastil data cover more than 138 countries, but my sample is more limited
to reflect the availability of other data.

S162

journal of political economy

Fig. 1.Democracy in the world, 196095

For countries outside of sub-Saharan Africa, the dashed line in

figure 1 shows that the average of the electoral rights index fell from
0.68 in 1960 (73 countries) to 0.55 in 1975 (95 countries). It then
returned to 0.68 in 1995. Thus, outside of sub-Saharan Africa, the
democratization since the mid 1970s has been sufficient to reattain
the average value for 1960.
Some of the analysis also uses the Freedom House indicator of
civil liberties. The definition here is civil liberties are rights to free
expression, to organize or demonstrate, as well as rights to a degree
of autonomy such as is provided by freedom of religion, education,
travel, and other personal rights (Gastil 198687, p. 7). Otherwise,
the subjective approach is the same as the one used for the electoral
rights indicator. The original scale for the civil liberties index from
one to seven has again been converted to zero to one, where zero
represents the fewest civil liberties and one the most. In practice, as
observed by Inkeles (1991), the indicator for civil liberties is extremely highly correlated with that for electoral rights. This high
degree of correlation does not apply if the indexes of electoral rights
and civil liberties are compared with measures of property rights
and legal structure, such as Knack and Keefers (1995) indicator for
maintenance of the rule of law, which is used in the subsequent analysis.

determinants of democracy

S163

II. Framework for the Determination of

Democracy
Inspection of the cross-country data suggests that countries at low
levels of economic development typically do not sustain democracy.
For example, the political freedoms installed in most of the newly
independent African states in the early 1960s did not tend to last.
Conversely, nondemocratic places that experience substantial economic development tend to become more dramatic. Examples include Chile, South Korea, Taiwan, Spain, and Portugal. Moreover,
the countries of central and eastern Europewhich have been reasonably advanced economically for some time, especially in terms
of educationeventually became more democratic. Thus a casual
view of the data seems to support the Lipset/Aristotle hypothesis.
To assess this hypothesis formally, I consider systems of the form
DEMOC it a 0t a 1 DEMOC i,t T
a 2 DEMOC i ,t 2T a 3 Z i,t T u it ,

(1)

where i is the country; t is the time period; T is a time lag, usually

taken to be five years; DEMOC is the indicator for democracyelectoral rights or civil liberties; Z is a vector of variables, such as per
capita gross domestic product and education, that influence the extent of democracy; and u is an error term. The idea in equation (1)
is that if a 1 0, a 2 0, and 0 a 1 a 2 1, then the extent of
democracy in a country converges gradually over time toward a
(moving) target that is determined by the Z variables.3 In practice,
the Z variables are themselves highly persistent over time.
Operationally, I use a panel setup in which the dependent variable, DEMOC it , is observed at most six times for each country: 1972,
1975, 1980, 1985, 1990, and 1995. (The year 1972 is the initial date
of the Freedom House sample.) The variables Z i ,tT refer to observations roughly five years prior to these dates, and DEMOC i ,tT and
DEMOC i,t2T refer, respectively, to (approximately) five- and 10-year
lags of the dependent variable.4 (The values for DEMOC tT pertain
to 1965, 1972, 1975, and so on.) The system includes a separate con3
The democracy indicators take on only seven discrete values between zero and
one, but the linear specification on the right-hand side does not take this pattern
into account. In practice, the most negative fitted value for electoral rights turned
out be 0.02 for Indonesia in 1972, and the highest fitted value was 1.09 for the
United States in 1995. Some improvement might result from the adoption of a nonlinear specification in which the dependent variable was constrained to the interval
(0, 1).
4
The empirical results turn out to be virtually the same if contemporaneous values
of the Z variables are entered into eq. (1) (i.e., if the lag T is set to zero), but lagged
values of the Z variables and the lags of democracy are used as instruments.

S164

journal of political economy

stant term, a 0t , for each date. The other coefficients (a 1, a 2, and a 3)

are constrained to be the same for each time period.
The panel estimation is carried out by the seemingly unrelated
regression (SUR) method. In this approach, the error term, u it , for
country i at time t is allowed to be correlated with the term u it for
the same country at different dates, t . The variance of u it varies with
t but not with i (so that countries are weighted equally in the estimation). In practice, the estimated correlations of the error terms
across the time periods turn out to be small, and the results are
similar if weighted least squares is used instead of the SUR method.
The results are also similar with ordinary least squares (so that the
equations for all time periods are weighted equally).
III.

Regression Results for Democracy: Standard

of Living and Related Variables

The basic regression results for electoral rights are in column 1 of

table 1. This system contains the five-year and 10-year lags of the
dependent variable as regressors. The explanatory variables also include the log of real per capita GDP5 and measures of educational
attainment. These indicators are observed roughly five years prior
to the dependent variable. The schooling figures that turn out to
have the most explanatory power are the average years of attainment
at the primary level for persons aged 25 and over and the gap in
average years of primary attainment between males and females
aged 25 and over.6
A dummy for oil-exporting countries, as designated by the International Monetary Fund (IMF),7 is also included as a rough adjustment of GDP for the contribution of natural resources. The idea
here is that the income generated from natural resources such as oil
may create less pressure for democratization than income associated
with the accumulation of human and physical capital.
The specification includes some other possible influences on democracy that have been proposed in the political science literature
that began with Lipset (1959); see Lipset, Seong, and Torres (1993)
5
The GDP data are taken from Summers and Heston (1991, 1995). Descriptions
of these and other variables are in the Barro-Lee data set, which is available on
the Internet from the World Bank (worldbank.org/html/prdmg/grthweb/growtht.htm).
6
The data on years of schooling are updated values described in Barro and Lee
(1996).
7
The IMF definition includes countries whose net oil exports represent a minimum of two-thirds of total exports and are at least equivalent to approximately 1
percent of world exports of oil. A definition based on OPEC membership would
add Ecuador and subtract Bahrain and Oman.

determinants of democracy

S165

TABLE 1
Regressions for Electoral Rights and Civil Liberties Indexes
Dependent Variable
Independent Variable
5-year lag of dependent variable
10-year lag of dependent variable
Log(GDP)
Years of primary schooling
Gap between male and female primary
schooling
Urbanization rate
Log(population)
Oil country dummy
R2
Observations

Electoral Rights
(1)

Civil Liberties
(2)

.608
(.041)
.102
(.040)
.058
(.016)
.0134
(.0059)
.047
(.013)
.095
(.048)
.0080
(.0044)
.094
(.031)
.62, .76, .67
.76, .76, .56
76, 88, 102
102, 103, 100

.536
(.041)
.148
(.039)
.054
(.014)
.0143
(.0051)
.043
(.011)
.075
(.041)
.0012
(.0038)
.096
(.027)
.63, .81, .77
.82, .75, .70
76, 88, 102
102, 103, 100

Note.The systems have six equations in which the dependent variables are the values of the electoral
rights or civil liberties indexes for 1972, 1975, 1980, 1985, 1990, and 1995. The five-year lag refers to 1965,
1972, 1975, etc. The 10-year lag refers to 1960, 1965, 1972, etc. The lagged values used for 1960 and 1965
(from Bollen [1990]) are the same in the two equations. The variables GDP (real per capita GDP), primary
schooling (years of attainment for persons aged 25 and over at the primary level), the gap between male and
female primary schooling, urbanization rate, and population refer to 1965, 1970, etc. The oil dummy equals
one for countries designated as oil-exporting by the IMF and zero otherwise. Each system contains a different
constant for each time period. The estimation, by the SUR technique, weights countries equally but allows
for different error variances in each period and for correlation of these errors over the periods. Standard
errors of the estimated coefficients are shown in parentheses. The R 2 values apply to each period individually.

and Lipset (1994) for discussions. The urbanization rate is often

mentioned as a determinant of democracy, although the sign of this
influence is not clear on theoretical grounds. Some observers argue
that the rural population has limited ability to organize and is therefore easy for a dictator to suppress. But one could also argue that a
less dense, rural population is harder for the central authority to
monitor and control. The simple correlation between democracy
and urbanization is strongly positive, but urbanization is also positively related to per capita GDP and the other measures of the standard of living that are included as regressors. In any event, the system
includes as an explanatory variable the rate of urbanization observed
five years prior to the dependent variable.8
8

The figures on urbanization are the standard ones reported by the World Bank.

S166

journal of political economy

The system also contains a measure of country size, the log of the
five-year earlier level of population. It is, however, not apparent a
priori whether a larger place is more or less likely to be democratic.
One problem of interpretation is that country size is endogenous,
as argued by Alesina and Spolaore (1995). Places that are too large
to be manageable are likely to have split apart sometime in the past.
The first observation from column 1 of table 1 is that the estimated
coefficient on the five-year lag of electoral rights is 0.61 (standard
error 0.04) and that on the 10-year lag is 0.10 (0.04). Thus democracy is highly persistent over time, but about 25 percent of the adjustment to a target position (determined by the other variables) occurs
over five years, and nearly 70 percent occurs over 20 years.9
The results are broadly supportive of the idea that more prosperous places are more likely to be democratic. The estimated coefficients on log(per capita GDP) and the level of primary schooling
are each significantly positive, 0.058 (0.016) and 0.013 (0.006), respectively.10 The first coefficient means that a doubling of per capita
GDP (corresponding roughly to a one-standard-deviation change)
would raise the electoral rights indicator by 0.04 in the short run
and by 0.14 after the full lagged adjustment occurs. (Note that a
shift by one Freedom House category corresponds to a change by
0.17 in the electoral rights index.) The second coefficient implies
that an additional year of average school attainment (roughly a onestandard-deviation shift) raises the electoral rights indicator by 0.01
in the short run and 0.04 in the long run.
Democracy is also negatively and significantly related to the gap
between male and female primary attainment, with an estimated
coefficient of 0.047 (0.013). That is, more equal educational opportunity across the sexes raises the target level of democracy. Quantitatively, a rise in the male-female gap by 0.6 year (about a onestandard-deviation change) lowers the electoral rights index by 0.03
in the short run and 0.10 in the long run. One interpretation of this
relation is that the spread between male and female attainment is
a proxy for general inequality of schooling and income. However,
the inclusion of explicit measures of educational and income inequality (discussed below) does not eliminate the explanatory power
9
These results apply when the coefficients of the lagged dependent variables are
interpreted in terms of a partial-adjustment model. Difficulties can arise here if the
lagged dependent variables pick up effects of serially correlated error terms. However, at least when the lagged dependent variables are included as regressors, the
residuals from the equations for electoral rights exhibit negligible correlations over
time.
10
School attainment of persons aged 25 and over has slightly more explanatory
power than attainment of persons aged 15 and over.

determinants of democracy

S167

of the gap between male and female schooling. Perhaps more promising is the idea, reminiscent of Tocqueville (1835), that expanded
educational opportunity for females goes along with a social structure that is generally more participatory and, hence, more receptive
to democracy.
The oil country dummy is significantly negative, 0.094 (0.031),
thereby indicating that the high level of per capita GDP associated
with oil production does not have the usual positive linkage with
democracy. The estimated coefficient implies that, in the long run,
for given values of per capita GDP and the other explanatory variables, an oil country would have an electoral rights indicator that was
lower by 0.32 (nearly two categories on the Freedom House scale).
It seems plausible that this result for oil would extend to natural
resource availability more generally. To test this idea, I introduced
the measure of natural resource intensity suggested by Sachs and
Warner (1995), the ratio of primary-product exports to total exports.11 However, this variable is insignificant if added to the system
shown in column 1 of table 1, 0.022 (0.033), and the oil dummy
remains significant, 0.090 (0.035). One problem with the export
ratio variable is that it reflects a countrys choices on which production activities to focus. Possibly a better measure of exogenous natural resource availability would outperform the oil dummy.
The estimated coefficient of the urbanization rate is negative and
marginally significant, 0.095 (0.048). Thus once indicators of the
standard of living are held constant, the association between urbanization and democracy switches sign and becomes negative. This result means that, for a given standard of living, it is not true that more
rural places are less likely to be democratic.
The estimated coefficient on the log of population is positive and
marginally significant, 0.0080 (0.0044).12 Thus there is some indication that larger places are more likely to be democratic. However,
as mentioned before, this result might reflect the endogeneity of
country size.
The system shown in column 1 of table 1 allows for different intercepts in each of the equations, that is, for 1972, 1975, and so on.
The estimated coefficient for each dateexpressed as a deviation
from the (unweighted) average interceptis shown in column 1 of
table 2. Column 2 of the table shows the unweighted mean of the
electoral rights index at each date for the observations that are included in the regression sample.
11
The value for 1970 enters into the first two equations, that for 1975 in the next
two, and that for 1985 in the last two.
12
A countrys land area is insignificant if it is added to the regressions as another
indicator of country size.

S168

journal of political economy

TABLE 2
Variations in Democracy over Time

Year

Estimated Constant Term

(Relative to Average)
(1)

1972
1975
1980
1985
1990
1995

.092
.047
.043
.035
.023
.038

Sample Mean of
Electoral Rights Index
(2)

(.024)
(.019)
(.020)
(.016)
(.017)
(.022)

.54
.47
.54
.58
.59
.64

Note.Col. 1 shows the estimated constant term for the indicated date from the
system in col. 1 of table 1. The estimate is expressed as a deviation from the average
of the constants over the six periods. The standard error of the coefficient estimate
is shown in parentheses. The p-value for the hypothesis of equal constant terms is
.0000. Col. 2 is the mean of the electoral rights index over the sample included in
the regression system for the indicated date.

The pattern of estimated intercepts suggests, in the language of

Huntington (1991, p. 16), a wave whereby world democracy declined in the early 1970s and then rose in the early 1980s and again
in the early 1990s.13 Note that these results apply for given values of
the explanatory variables, which can also generate patterns in world
democracy. For example, positive trends in per capita GDP and
schooling would imply an upward trend in democracy. This result
seems reasonable in a long-run global context becausewhen evaluated in terms of the Freedom House concept of electoral rights
there have not been many democracies in the world until the twentieth century. Huntington (1991, p. 16) dates the first long wave of
democratization as 18281926.
IV. Additional Influences on Democracy
Table 3 considers other possible determinants of democracy, many
of which have been proposed in the political science literature.
These additional variables are entered one set at a time into the sixperiod regression system described in column 1 of table 1.
A.

Health Indicators

Regression 1 of table 3 adds a measure of health statusthe log of

life expectancy at birthas another indicator of the standard of
13
The pattern for the estimated intercepts in col. 1 of table 2 differs from that
for the mean of the electoral rights index in col. 2 because of the dynamic relation
between the dependent and independent variables. In particular, a permanent
downward shift in the intercept would reduce democracy contemporaneously and
would imply further declines in democracy in future periods.

TABLE 3
Additional Determinants of Democracy
Independent Variable
1. Log(life expectancy at birth)
2. Infant mortality rate
3. Years of upper schooling
Gap between male and female schooling
p-value
4. Income inequality (Gini coefficient)
5. Share of middle class in income
6. Educational inequality
7. Ethnolinguistic fractionalization
8. Rule-of-law index
9. Dummy for former colony
10. Dummy for British colony
Dummy for French colony
Dummy for Spanish colony
Dummy for Portuguese colony
Dummy for other colony
p-value
11. Muslim religion fraction
Protestant religion fraction
Hindu religion fraction
Buddhist religion fraction
Miscellaneous eastern religion fraction
Jewish religion fraction
Nonreligion fraction
Other religion fraction
p-value

Regression Coefficient
.129
.44
.008
.010
.76
.15
.26
.008
.055
.031
.018
.031
.020
.007
.022
.032
.33
.093
.006
.049
.003
.081
.035
.244
.103
.002

(.079)
(.30)
(.011)
(.024)
(.09)
(.15)
(.046)
(.028)
(.057)
(.018)
(.018)
(.026)
(.022)
(.048)
(.034)
(.026)
(.031)
(.050)
(.047)
(.072)
(.072)
(.089)
(.050)

Note.The indicated groups of explanatory variables are added, one at a time, to the system for the
electoral rights index shown in col. 1 of table 1. (Regression 8 applies only to the three periods that start
with the value of the electoral rights index for 1985.)
Life expectancy at birth applies to 196569, 197074, etc. The infant mortality rate applies to 1965, 1970,
etc. Upper schooling (from Barro and Lee [1996]) is the years of secondary and higher schooling for persons
aged 25 and over in 1965, 1970, etc. The Gini coefficient for income inequality and the income share of the
middle class (the three middle quintiles of income) are taken from Deininger and Squire (1996) and apply
around 1970 in the first two equations, around 1980 in the next two equations, and around 1990 in the last
two equations. A higher number for the Gini coefficient signifies more inequality. Educational inequality
(from Barro and Lee [1996]) is the standard deviation of log(1 years of schooling) for the population
aged 15 and over in 1965, 1970, etc. The ethnolinguistic fractionalization variable, which runs between zero
and one, is a measure of heterogeneity of language and ethnicity. The number, observed once for each
country, represents the probability that two randomly selected persons come from different groups; hence,
a higher value signifies more heterogeneity. See Taylor and Hudson (1972, table 4.15) and Mauro (1995)
for a discussion of these data. The rule-of-law index, discussed in Knack and Keefer (1995) and available for
198297 from Political Risk Services, is a subjective indicator of the extent of maintenance of the rule of law.
The variable runs from zero to one, with a higher value indicating a more favorable environment.
Colony is a dummy for countries that are former or present colonies; any country that was independent
before 1776 is designated as a noncolony. In regression 10, dummies for former British, French, Spanish,
Portuguese, and other colonies are added together to the system from col. 1 of table 1. Colonial status is
based on the most recent ruler; e.g., the Philippines is attributed to the United States rather than to Spain.
In regression 11, the fractions of the population affiliated with eight major religious groups are entered
together into the system from col. 1 of table 1. The left-out religion category is Catholic (including Eastern
Orthodox). The religion data pertain to 1970 (in the first three equations) and 1980 (in the last three equations) and come from Barrett (1982). The Protestant group includes Anglicans, marginal Protestants ( Jehovahs Witnesses, Mormons, and new-age cults), and cryto-Christians (secret believers in Christ not professing
publicly). Eastern religions include Chinese folk religions, Shinto, Confucianism, and new-religions. The
nonreligion category comprises those professing no religion and atheism. Other religions include Parsis,
Spiritists, tribal religions, indigenous third-world Christians not of western importation, and Bahais. Jains and
Sikhs are classed with Hindus.

S170

journal of political economy

living. The estimated coefficient is positive but only marginally significant, 0.13 (0.08). With life expectancy included, the estimated
coefficients of log(per capita GDP) and average years of primary
schooling become less significant; the estimated coefficients are now
0.042 (0.018) and 0.011 (0.006), respectively. Thus it is difficult to
sort out precisely the measures of standard of living that matter for
democracy.
Similar results apply if the infant mortality rate is used as a measure of health status instead of life expectancy. Regression 2 of table
3 shows that the estimated coefficient of the infant mortality rate is
0.44 (0.30).
B. Upper-Level Schooling
Regression 3 in table 3 includes two additional schooling variables:
the average years of schooling for persons aged 25 and over at the
secondary and higher levels and the gap between male and female
schooling for persons aged 25 and over at these levels. These variables are individually and jointly insignificant, as shown, whereas
the estimated coefficients on primary schooling remain significant
(0.014 [0.006] for years of schooling and 0.051 [0.016] for the gap
between males and females). Hence, it appears to be early education
that matters for democratization. Similar results apply to the determination of fertility rates and health status. However, as discussed
in Barro (1997), rates of economic growth and investment relate far
more to secondary and higher schooling than to primary education.
C. Inequality of Income and Schooling
Regression 4 of the table includes measures of income inequality,
as gauged by Gini coefficients for the distribution of income. (A
higher Gini coefficient signifies more inequality.) The data come
from the careful compilation of information assembled by Deininger and Squire (1996).14 The underlying values were categorized
as applying around 1970, 1980, and 1990 and were entered accordingly into the regression system (see the note to table 3). Because
of the limited availability of data on income distribution, the inclusion of the Gini coefficient substantially reduces the number of observations: to 51 for the 1972 equation, 56 for 1975, 60 for 1980 and
14
The regressions use their high-quality observations, which exclude observations based on incomplete geographical coverage or incomplete measures of income. I have added to their high-quality set some observations that Deininger and
Squire excluded because of incomplete references to primary sources.

determinants of democracy

S171

1985, and 68 for 1990 and 1995. Within this truncated sample, the
estimated coefficient on the Gini coefficient is negative but only marginally significant, 0.15 (0.09). However, in this reduced sample,
some of the variables from the basic specification are also statistically
insignificant: average years of primary schooling, the urbanization
rate, the log of population, and the oil dummy. In any event, there is
some indication thatfor given measures of the standard of living
greater income inequality predicts less democracy.
Many researchers stress a positive interaction between the size of
the middle class and the extent of democracy. To check this relation,
regression 5 of table 3 includes in the regression system the share of
income accruing to the middle three quintiles (also compiled from
Deininger and Squire [1996]). Because of the more severe limitation on data by quintile shares, this sample was even smaller than
that for the Gini coefficient. (With the middle-class variable included, the numbers of observations for the various dates are 38,
40, 51, 51, 65, and 65.) The estimated coefficient of the middleclass share is positive, 0.26 (0.15), but only marginally significant.
However, in this sample, the estimated coefficients of average years
of primary schooling, urbanization rate, the log of population, and
the oil dummy are even less significant.
If the Gini coefficient is entered along with the middle-class share,
then the estimated coefficients are 0.05 (0.24) for the Gini (the
wrong sign) and 0.32 (0.38) for the middle-class share. The
middle-class variable also appears to be the more important indicator of inequality when it is entered jointly with the share of the lowest
or highest quintile. Thus there is some evidence that the weight of
the middle class is the aspect of inequality that matters most for democracy.
Another possibility is to use recently assembled data on educational attainment at seven levels to construct measures of schooling
inequality (see n. 6 above). Regression 6 of table 3 uses as an independent variable the standard deviation of log(1 years of schooling)15 for the population of both sexes aged 15 and over. This variable is observed for 1965, 1970, and so on. The estimated coefficient
is close to zero, 0.008 (0.046). The estimated coefficients of primary schooling remain significant here: 0.013 (0.006) for average
years of schooling and 0.046 (0.016) for the male-female gap. If
the Gini coefficient for years of schooling is used as an alternative measure of educational inequality, then the findings are similar. Hence, these results indicate that the primary enrollment vari15
The value one can be thought of as the effective years of educational human
capital possessed by a person with no formal schooling.

S172

journal of political economy

ablesand specifically the male-female gapdo not enter the

regressions merely as proxies for educational inequality.
D. Ethnolinguistic Fractionalization
The populations degree of heterogeneity with respect to ethnicity,
language, and culture may also matter for democracy. The usual idea
is that more heterogeneity makes it more difficult to sustain democracy. A standard measure of a populations heterogeneity is its ethnolinguistic fractionalization, a measure of disparity of languages
and ethnicity within a country.16 The variable runs between zero and
one and is intended to measure the probability that two randomly
chosen persons in a country come from different groups. Hence,
zero is the most homogeneous, and one is the most heterogeneous.
Regression 7 of table 3 shows that the estimated coefficient of the
fractionalization variable (observed once per country) is negative
and marginally significant, 0.055 (0.028). Thus there is some indication that more ethnically diverse countries are less likely to sustain
democracy.
One possibility is that ethnic diversity reduces democratic tendency by contributing to inequality. To test for this possibility, I included the middle-class share of income along with the fractionalization variable. The result was that the estimated coefficient of
fractionalization was essentially zero, 0.002 (0.036), whereas that on
the middle-class share was 0.24 (0.15), similar to that found in regression 5 of table 3. It turns out, however, that the result does not
reflect a high correlation between the fractionalization variable and
the middle-class share (a correlation that is negative but small in
magnitude) but rather the truncation of the sample. Within the limited sample for which data on the middle-class share are available,
the estimated coefficient of the fractionalization variable is essentially zero even if the middle-class variable is omitted from the
system.
E.

The Rule of Law

Knack and Keefer (1995) discuss a variety of subjective country indexes of property rights and legal/political structure that have been
prepared for fee-paying international investors by International Country Risk Guide. The concepts covered include quality of the bureaucracy, political corruption, likelihood of government repudiation of
contracts, risk of government expropriation, and overall mainte16
Most of the data come from Miklukho-Maklaya Institute (1964), as reported in
Taylor and Hudson (1972, table 4.15). See Mauro (1995) for a discussion.

determinants of democracy

S173

nance of the rule of law (also referred to as law and order tradition). The various time series cover 198297 and are available for
a fee from Political Risk Services of Syracuse, New York. The general
idea of these indexes is to gauge the attractiveness of a countrys
investment climate by considering the effectiveness of law enforcement, the sanctity of contracts, and the state of other influences on
the security of property rights. Although these data are subjective,
they have the virtue of being prepared contemporaneously by local
experts. Moreover, the willingness of customers to pay substantial
amounts for this information is perhaps some testament to their validity.
Among the various series available, the indicator for overall maintenance of the rule of law17 seemed a priori to be the most relevant
for investment and growth. My previous empirical work (Barro 1997)
confirms that this indicator is an important predictor of economic
growth.
The connection between democracy and property rights is unclear, as stressed by Sirowy and Inkeles (1990) and Przeworski and
Limongi (1993). The simple correlations between the electoral
rights and rule-of-law indicators are .64 in 1982, .55 in 1985, .60 in
1990, and .49 in 1995. These correlations are much smaller than
those between the electoral rights and civil liberties indexes (see
below).
Regression 8 of table 3 checks out the relationship between electoral rights and the rule of law by entering lagged values of the ruleof-law index into the equations for democracy. Since the data on
the rule of law begin in 1982, this system includes the equations for
electoral rights for only 1985, 1990, and 1995. (The second lag of
the dependent variable is excluded here.) The values for the rule
of law in this system apply to 1982, 1985, and 1990. The result is that
the estimated coefficient on the rule-of-law variable is positive but
insignificant, 0.031 (0.057). Thus, when the measures of standard
of living are held fixed, there is not much timing evidence that better
maintenance of the rule of law promotes electoral rights. (However,
the rule of law can stimulate electoral rights indirectly by promoting
economic growth.)
The rule-of-law measure can also be viewed as the dependent variable in a system in which the independent variables are its own lags
and the lags of the other variables, including the electoral rights
index. (Three equationsfor 1985, 1990, and 1995are used
here.) In this setting, electoral rights turn out to enter with a positive
17 The indicator was initially measured in seven categories on a zero to six scale,
with six the most favorable. The scale has been revised here to zero to one, with
zero indicating the worst maintenance of the rule of law and one the best.

S174

journal of political economy

TABLE 4
Democracy in Relation to Colonial Status and Religion
A. Colonial Status

Colonial Status
Noncolony
Colony
British colony
French colony
Spanish colony
Portuguese colony
Other colony
All countries

Number of
Countries

Electoral
Rights Index*

32
106
53
23
16
5
9
138

.69
.46
.54
.26
.60
.30
.36
.52

B. Religious Affiliation
Primary Religious Affiliation
in 1980
Catholic
Muslim
Protestant
Hindu
Buddhist
Miscellaneous eastern religions
Jewish
Nonreligion
Other religion
All countries with data on religion

Number of
Countries

Electoral
Rights Index*

53
32
21
5
4
3
1
1
16
136

.60
.26
.80
.66
.56
.46
.86
.10
.30
.51

Note.See the discussion in the text and table 3 for definitions of colonial status and religious
affiliation. Panel B shows averages for 197595 of the electoral rights index for groups of countries
in which the most common religious affiliation in 1980 is of the indicated type.
* Average for 197595.

coefficient, 0.023 (0.026), which is not statistically significant. Thus

there is also not strong timing evidence that electoral rights stimulate the maintenance of the rule of law. Overall, the results indicate
that, for given measures of the standard of living, there is a lot of
independence in the ways electoral rights and the rule of law evolve.
F.

Colonial History

Colonial heritage would be important for democracy if countries

inherited tendencies for more or less political freedom from their
previous rulers. For example, Lipset et al. (1993, p. 168) argue that
British rule provided a crucial learning experience for subsequent
democracy. In table 4, a noncolony is defined to be a country that
was independent since 1775 (so that the United States is treated as
a former possession of Britain). Each former colony is attributed to
its most recent occupier; for example, the Philippines is associated

determinants of democracy

S175

with the United States rather than with Spain, Rwanda and Burundi
are attached to Belgium rather than to Germany, and several Caribbean countries are related to Britain rather than to Spain. The classification treats as noncolonies places such as South Korea, Taiwan,
Hungary, and Poland, which were occupied by a foreign power for
some periods.
Panel A of table 4 shows that the 32 noncolonies are more likely
to be democratic (average value for the electoral rights indicator
from 1975 to 1995 of 0.69) than the colonies (average value of 0.46).
Within the colonies, the former possessions of Britain and Spain are
substantially more democratic than those of France, Portugal, and
other countries. (The former Spanish colonies in Latin America
would, however, look less democratic at some earlier dates.)
In the statistical analysis, with the measures of standard of living
held constant, regression 9 of table 3 shows that a dummy variable
for colonial status (one for former colony, zero for noncolony) is
negative but statistically insignificant, 0.018 (0.018). Moreover, regression 10 shows that a breakdown among British, French, Spanish,
Portuguese, and other colonies fails to generate any significant coefficients. The p-value for joint significance of the five colony dummies
is .33. These results, in conjunction with table 4, suggest that the
influence of former colonial status on democratic tendency mostly
works indirectly through effects on the standard of living, as measured here particularly by per capita GDP and primary schooling.
These indirect links with colonial history are worth further study.
G. Religion
Religious affiliation has also been stressed as an important determinant of democracy (see Huntington 1991, pp. 7185; Lipset 1994,
p. 5; Boone 1996, pp. 2528). Unfortunately, however, the theory
of the interplay between religion and political structure is even less
developed than other aspects of the theory of democracy.
To check for a connection between religion and political freedom, I use data compiled by Jong Wha Lee on the fractions of a
countrys population in 1970 and 1980 affiliated with nine major
groups:18 Catholic (including Eastern Orthodox), Muslim, Protes18
The underlying data, from the World Christian Encyclopedia (Barrett 1982), are
estimates of professed affiliation in 1970 and 1980. (Figures are also available for
1900.) This information takes no account of regularity of church attendance or
amounts spent on religious activities. The data provided in the encyclopedia are
compiled from the most authoritative local sources, published or unpublished, including government censuses of religion. A large part of the data was collected directly by the editors, who visited virtually all the countries over the years 196575.
The basic concept of a religious adherent is that the person professes to believe in
the religion when government censuses or public opinion polls ask the question

S176

journal of political economy

tant (including Anglican and some other Christian groups), Hindu

(including Jains and Sikhs), Buddhist, miscellaneous eastern religions (Chinese folk religions, Shinto, Confucianism, and newreligionists), Jewish, no professed religion (including atheists), and
other religious groups (such as Parsis, Bahais, Spiritists, tribal religions, and indigenous third-world Christians).
Panel B of table 4 verifies that differences in a countrys primary
religious affiliation relate strongly to democracy. When countries are
sorted in accordance with their most popular religion in 1980, the
average of the electoral rights indicator from 1975 to 1995 is 0.86
for Jewish (1 country), 0.80 for Protestant (21 countries), 0.66 for
Hindu (5 countries), 0.60 for Catholic (53 countries), 0.56 for Buddhist (4 countries), 0.46 for miscellaneous eastern relgions (3 countries), 0.30 for other religions (16 countries), and 0.26 for Muslim
(32 countries). China is the only place in which nonreligion is the
most common affiliation, and the average of the electoral rights index in this case is 0.10. The mean value of electoral rights for all
136 countries with data on religion is 0.51.
A prominent aspect of this breakdown is that Protestant countries
are nearly always highly democratic, whereas Muslim countries are
usually not democratic. Only four of the 32 Muslim countries have
electoral rights indicators that averaged at least 0.5 for 197595:
Gambia (0.70), Senegal (0.50), Malaysia (0.57), and Turkey (0.62).
Regression 11 of table 3 shows the results when eight religious
variables are entered into the equations for electoral rights.19 (The
omitted characteristic is chosen arbitrarily to be Catholic, the most
prevalent religion when countries are weighted equally.) The regressions indicate that the religion variables with the most significant
coefficients are Muslim, 0.093 (0.026); nonreligion, 0.24 (0.09);
and other religion, 0.103 (0.050).
The p-value of .002 indicates that the eight religion coefficients
are significant overall. However, some of this significance hinges on
the presence of a few outlier observations. For example, the positive
coefficient on Hindu mainly indicates that India and Mauritius are
surprisingly democratic, given their indicators of the standard of living. If these two places are omitted from the sample, then the estiWhat is your religion? Each person is considered to have at most one religious
affiliation. Further work on cross-country religion data is ongoing.
19
The system allows for variation over time in religious affiliation in that the 1970
religion figures appear in the first three equations for democracy, and the 1980
figures enter into the last three equations. However, in most cases, the variations
in religious affiliation between 1970 and 1980 are minor. If the 1970 values are
included in all six equations, then the results are virtually indistinguishable from
those shown in regression 11 of table 3.

determinants of democracy

S177

mated coefficient on Hindu falls to 0.003 (0.062). The significance

of the estimated coefficient on nonreligionthough not the point
estimatedepends on the inclusion of China, the one country for
which this affiliation exceeds 0.5.20 If China is omitted, then this coefficient becomes 0.24 (0.13).
Probably the most interesting finding is the significantly negative
coefficient on the Muslim variable, even when the measures of standard of living and the other explanatory variables are held constant.
The estimated coefficient of 0.093 means that, for given values of
the other explanatory variables, the long-run level of electoral rights
is lower by 0.32about two Freedom House categoriesin a Muslim country (as compared to the benchmark of a Catholic country).
Possibly this result reflects the strong linkage between church and
state in many Muslim countries.
The estimated coefficients understate the potential consequence
of religion for democracy because of the potential effects of religion
on some of the explanatory variables, especially the gap between
male and female education and the indicators of the standard of
living. In fact, the main effects of religion on democracy are likely
to work through these indirect channels. Given the striking patterns
that emerge in table 4, these linkages are worth further investigation.
V. Civil Liberties
The electoral rights indicator is a narrow procedural measure that
focuses on the role of elections. In contrast, the Freedom House
index of civil liberties is a broader concept that covers freedoms of
speech, press, and religion and also considers a variety of legal protections. In practice, however, the civil liberties variable is highly correlated with the electoral rights index: .86 in 1972, .93 in 1980, .94
in 1990, and .92 in 1995.
Given this high degree of correlation, it is not surprising that results with the civil liberties index as the dependent variableshown
in column 2 of table 1look similar to those found for the electoral
rights index.21 This result suggests that the economic and social
forces that promote electoral rights are similar to those that stimulate civil liberties.
20
The 1980 value of nonreligion for China is 0.71. The other values that exceed
0.1 are 0.35 for Uruguay; 0.29 for Sweden; 0.17 for Yugoslavia; 0.16 for Italy, Hungary, and France; 0.15 for Australia; 0.14 for Hong Kong; and 0.12 for the Netherlands and Japan.
21
A formal test rejects equality of the coefficients in the systems for electoral rights
and civil liberties. Viewed individually, however, the only estimated coefficients that
are found to differ significantly at the 5 percent critical level are those for the lagged
dependent variables, log of population, and the constant terms.

S178

journal of political economy

VI. Long-Run Forecasts of Democracy

The estimated relation from column 1 of table 1 implies a gradual
adjustment of democracyas measured by electoral rightstoward
the values determined by the explanatory variables aside from
lagged democracy.22 In a full system, the dynamics of these explanatory variables would also be determined. In practice, the level of per
capita GDP and the other variables are highly persistent over time,
although they evolve gradually in line with the process of economic
development, some of which has been studied in previous research
in terms of rates of economic growth.
One simple way to relate the current level of democracy to its
long-run target is to compute at each date the estimated level of
democracy that would arise asymptotically if all the right-hand-side
variables (aside from the lagged dependent variables) were held
fixed at their current values. For example, in 1975, the long-run
level of the electoral rights index is calculated from the 1970 values
of the regressors included in column 1 of table 1.23 The resulting
projected values for 1975 and 1995 are shown along with the actual
values in table 5. The gap is the difference between the current level
of electoral rights and its long-run target. Values of the gap that
exceed 0.33 in magnitude are shown in boldface.
In 1975, out of 101 countries with the necessary data, five were
below the long-run target for electoral rights by at least 0.33 and 21
were above by at least 0.33. In 1995, out of 100 places with the required data, 16 were below target by at least 0.33, and five were above
target by at least 0.33.
One striking observation is that no country appears with the same
sign on both lists (although Yugoslavia comes close). With an estimated rate of convergence for electoral rights of about one-quarter
per five-year period (from the coefficients on the lags of the dependent variable in col. 1 of table 1), a lot of reversion to the mean
occurs over 20 years.
Among the sub-Saharan African countries, Botswana, Gambia,
and Mauritius (if Mauritius is classed with Africa) looked too democratic in 1975, but the situation for Gambia changed with a coup
in 1994. Botswana and Mauritius were still above target for democracy in 1995, but by much smaller amounts than in 1975.
22 More precisely, the model shows how initial electoral rights and the values of
the other explanatory variables influence the probabilities of transition over time
among the seven discrete rankings of electoral rights.
23
The projected value equals [1/(1 coefficient of first lag of electoral rights
coefficient of second lag of electoral rights)] (estimated value based on explanatory variables other than the lagged dependent variables).

determinants of democracy

S179

A surge of democratization since the late 1980s meant that many

of the countries in sub-Saharan Africa became more democratic
than predicted by 1995. This group includes Benin, Central African
Republic, Guinea-Bissau, Malawi, Mali, Niger, and Zambia. In some
of these cases, the democratization may be explicable from the pressures and rewards exerted by international organizations, such as
the IMF and the World Bank. (The recent U.S. efforts in Haiti are
analogous.) In any case, the regression analysis predicts that, as with
the African experience of the 1960s, democracy that gets well ahead
of economic development will not last. As a possible indicator of this
process, Niger had a military coup in January 1996 and then became
nondemocratic.
Some sub-Saharan African countries still had below-target levels
of democracy in 1995. Prominent here are Cameroon, Kenya,
Rwanda, Sierra Leone, Sudan, and Swaziland. (Nigeria and Somalia
would likely be included in this category but have missing data and
are therefore excluded from table 5.)
For Latin America in 1975, several countries were surprisingly
nondemocratic, including Panama, Chile, and Ecuador. All these
places subsequently experienced sharp increases in electoral rights.
In 1995, Bolivia had more democracy than would be predicted from
its economic situation. Mexico and Peru were below target in 1995
and were predicted to become more democratic.
Among Asian countries, surprisingly low democracy prevailed in
1995 in China, Hong Kong, Indonesia, Singapore, and Syria. The
model predicts increases in democracy in these cases, but the model
has not been informed of Hong Kongs postJuly 1997 relationship
with China. It will be especially interesting to see whether prosperous
Singapore joins South Korea and Taiwan in their marked democratizations.
On the other side, democracy was higher than predicted in 1975
for India, Indonesia, Malaysia, Sri Lanka, and Thailand. However,
with the decline in the Freedom House measure of electoral rights
in 1991 and 1993, India no longer looked like an outlier in 1995.
Indonesia, Malaysia, Sri Lanka, and Thailand were all below target
in 1995.
Because of lack of data, only three countries from central or eastern Europe are represented in the sample: Hungary, Poland, and
Yugoslavia. In each of these cases, democracy was strikingly below
its target level in 1975. Two countries in western EuropePortugal
and Spainwere also below target. By 1995, all these countries except Yugoslavia hadas predictedbecome far more democratic.
The model forecasts a large increase of democracy in Yugoslavia,
which should perhaps now be identified with Serbia. The model also

TABLE 5
Actual and Long-Run Values of Democracy
Democracy 1975
Country
Algeria
Benin
Botswana
Cameroon
Central African Republic
Congo
Egypt
Gambia
Ghana
Guinea-Bissau
Kenya
Lesotho
Liberia
Malawi
Mali
Mauritius
Mozambique
Niger
Rwanda
Senegal
Sierra Leone
South Africa
Sudan
Swaziland
Tanzania
Togo
Tunisia
Uganda
Zaire
Zambia
Zimbabwe
Barbados
Canada
Costa Rica
Dominican Republic
El Salvador
Guatemala
Haiti
Honduras
Jamaica
Mexico
Nicaragua
Panama
Trinidad and Tobago
United States
Argentina
Bolivia
Brazil
Chile
Colombia
Ecuador
Guyana
Paraguay
Peru

Democracy 1995

Actual

Projected

Gap

Actual

Projected

Gap

.17
.00
.83
.17
.00
.33
.17
.83
.00
.17
.33
.33
.17
.00
.00
.83
.17
.00
.00
.17
.17
.50
.17
.17
.17
.00
.17
.00
.00
.33
.17
1.00
1.00
1.00
.50
.83
.50
.17
.17
1.00
.50
.33
.00
.83
1.00
.50
.17
.50
.00
.83
.00
.50
.33
.17

(.14)
.23
.26
.08
.03

.15*
.06*
.09

.04
.32
.08
(.03)
.03
.20
.33
.19
.19
.14
.20
.55
.22
.37
.08**
.06
.14
.10
(.01)
.01
.18
.65
.82
.49
.31
.21
.32
.16
.27
.49
.32
.35
.44
.64
.95
.57
.15
.43
.49
.36
.36
.40
.30
.29

(.31)
.23
.57
.08
.03

.01
.78
.09

.30
.01
.09
(.03)
.03
.63
.17
.19
.19
.03
.03
.05
.06
.20
.09
.06
.03
.10
(.01)
.32
.02
.35
.18
.51
.19
.62
.18
.01
.11
.51
.18
.02
.44
.20
.05
.07
.02
.07
.49
.48
.36
.10
.04
.12

.17
.83
.83
.00
.67
.50
.17
.00
.50
.67
.00
.50
.00
.83
.83
1.00
.67
.67
.00
.50
.00
1.00
.00
.17
.33
.17
.17
.33
.00
.67
.33
1.00
1.00
1.00
.50
.67
.50
.33
.67
.83
.50
.50
.83
1.00
1.00
.83
.83
.83
.83
.50
.83
.83
.50
.33

.26
.36
.75
.44
.21
.56
.48
.33
.33
.33**
.40
.81

.28
.34
.79
.40
.36
.39
.44
.38
.78
.47
.65
.33**
.22
.53
.34
.16
.16
.49
1.00
(1.23)
.85
.62
.62
.66
.28
.62
.75
.89
.49
.82
.96
(1.30)
.87
.51
.82
.79
.82
.76
.66
.71
.61

.09
.47
.09
.44
.45
.06
.31
.33
.17
.33
.40
.31

.55
.49
.21
.27
.31
.39
.06
.38
.22
.47
.48
.00
.05
.37
.01
.16
.51
.15
.00
(.23)
.15
.12
.05
.16
.06
.04
.08
.39
.01
.01
.04
(.30)
.04
.32
.01
.04
.32
.08
.17
.21
.28

S180

TABLE 5 (Continued )
Democracy 1975
Country
Uruguay
Venezuela
Bahrain
Bangladesh
China
Hong Kong
India
Indonesia
Iran
Iraq
Israel
Japan
Jordan
South Korea
Malaysia
Nepal
Pakistan
Philippines
Singapore
Sri Lanka
Syria
Taiwan
Thailand
Austria
Belgium
Cyprus
Denmark
Finland
France
West Germany
Greece
Hungary
Iceland
Ireland
Italy
Malta
Netherlands
Norway
Poland
Portugal
Spain
Sweden
Switzerland
Turkey
United Kingdom
Yugoslavia
Australia
Fiji
New Zealand
Papua New Guinea

Actual
.33
.83
.17
.00
.00
.67
.83
.33
.17
.00
.83
.83
.17
.33
.67
.17
.33
.33
.33
.83
.17
.17
.83
1.00
1.00
.50
1.00
.83
1.00
1.00
.83
.17
1.00
1.00
1.00
1.00
1.00
1.00
.17
.33
.33
.83
1.00
.67
1.00
.17
1.00
.83
1.00
.67

Projected
.50
.21

.25
.12*
.12
.10
(.19)
.11
.05
.50
.76
.02
.25
.13
.22
.16
.44
.01
.30
.13
.25
.35
.68
.83
.36
.83
.93
.76
.91
.45
.80
.56
.68
.70
.30
.75
.80
.62
.44
.61
.70
.86
.28
.77
.44
.83
.33
.81
.34

Democracy 1995

Gap

Actual

.17
.62

.25
.12
.54
.74
(.53)
.06
.05
.33
.07
.14
.08
.53
.05
.17
.11
.32
.54
.04
.08
.48
.32
.17
.14
.17
.10
.24
.09
.38
.64
.44
.32
.30
.70
.25
.20
.46
.10
.27
.13
.14
.39
.23
.27
.17
.50
.19
.33

.83
.67
.17
.67
.00
.50
.50
.00
.17
.00
1.00
1.00
.50
.83
.50
.67
.67
.83
.33
.50
.00
.67
.67
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
.33
1.00
.17
1.00
.50
1.00
.83

Projected
.82
.51
.47
.55
.54
.92
.48
.38
.31
.04
.96
(1.21)
.45
.94
.77
.45
.50
.86
.79
.81
.57
.96
.95
(1.07)
(1.16)
.93
(1.20)
(1.30)
(1.12)
(1.26)
.79
(1.01)
.97
(1.10)
(1.05)

(1.09)
(1.12)
(1.04)
.94
(1.06)
(1.11)
(1.20)
.75
(1.19)
.81
(1.18)
.83
(1.13)
.54

Gap
.02
.15
.30
.11
.54
.42
.02
.38
.14
.04
.04
(.21)
.05
.10
.27
.22
.17
.03
.45
.31
.57
.30
.28
(.07)
(.16)
.07
(.20)
(.30)
(.12)
(.26)
.21
(.01)
.03
(.10)
(.05)

(.09)
(.12)
(.04)
.06
(.06)
(.11)
(.20)
.42
(.19)
.64
(.18)
.33
(.13)
.29

Note.Actual values pertain to the electoral rights index. Projected values are based on the estimated
system shown in table 1, col. 1. The 1975 projection is [1/(1 coefficient of first lag of electoral rights
coefficient of second lag of electoral rights)] (estimated value based on explanatory variables other than
the lagged dependent variables included in the 1975 equation). The 1995 projection is formed analogously.
Values in parentheses are linearly fitted values that lie outside the range (0, 1). Values shown in bold have
a magnitude of at least .33.
* Projected value for 1975 uses 1975 schooling values because of missing data for 1970.
** Projected value based on estimated schooling value for persons aged 25 and over based on available
data for persons aged 15 and over.

S181

S182

journal of political economy

predicts substantial democratization for Turkey, which was surprisingly nondemocratic in 1995.
VII. Concluding Observations
The data for a large panel of countries confirm the Lipset/Aristotle
hypothesis, which says that a higher standard of living promotes democracy. This relation shows up when democracy is represented by
electoral rights or civil liberties and when the standard of living is
measured by per capita GDP, primary school attainment, the gap
between male and female primary schooling (which enters negatively), and the importance of the middle class. Democracy does not
relate significantly to school attainment at the secondary and higher
levels. For a given standard of living, democracy tends to fall with
urbanization and a greater reliance on natural resources but has
little relation to country size.
The apparently strong relation of democracy to colonial heritage
mostly disappears when the measures of standard of living are held
constant. Similarly, the allowance for standard of living weakens the
interplay between democracy and religious affiliation. However, negative effects from Muslim and nonreligious affiliations remain intact.
Given the strength of the Lipset/Aristotle hypothesis as an empirical regularity, it is surprising that convincing theoretical models of
the relation do not exist. Thus development of such a theory is a
priority for future research.
At an empirical level, it would be especially interesting to investigate further the relation of democracy to inequality, colonial status,
and religion. Hopefully, the development of satisfactory theories of
the determination of democracy will suggest additional empirical
linkages that ought to be explored.
References
Alesina, Alberto, and Spolaore, Enrico. On the Number and Size of Nations. Manuscript. Cambridge, Mass.: Harvard Univ., October 1995.
Aristotle. Politics. Translated by H. Rackham. Cambridge, Mass.: Harvard
Univ. Press, 1932.
Barrett, David B., ed. World Christian Encyclopedia: A Comparative Study of
Churches and Religions in the Modern World, ad 19002000. Oxford: Oxford
Univ. Press, 1982.
Barro, Robert J. Determinants of Economic Growth: A Cross-Country Empirical
Study. Cambridge, Mass.: MIT Press, 1997.
Barro, Robert J., and Lee, Jong Wha. International Measures of Schooling
Years and Schooling Quality. A.E.R. Papers and Proc. 86 (May 1996): 218
23.
Bollen, Kenneth A. Political Democracy: Conceptual and Measurement
Traps. Studies Comparative Internat. Development 25 (Spring 1990): 724.

determinants of democracy

S183

Boone, Peter. Political and Gender Oppression as a Cause of Poverty.

Manuscript. London: London School Econ., April 1996.
Deininger, Klaus, and Squire, Lyn. Measuring Income Inequality: A New
Data-Base. Manuscript. Washington: World Bank, March 1996.
Gastil, Raymond D. Freedom in the World. Westport, Conn.: Greenwood, various years.
. The Comparative Survey of Freedom: Experiences and Suggestions. On Measuring Democracy: Its Consequences and Concomitants, edited
by Alex Inkeles. New Brunswick, N.J.: Transaction, 1991.
Huber, Evelyne; Rueschemeyer, Dietrich; and Stephens, John D. The Impact of Economic Development on Democracy. J. Econ. Perspectives 7
(Summer 1993): 7185.
Huntington, Samuel P. The Third Wave: Democratization in the Late Twentieth
Century. Norman: Univ. Oklahoma Press, 1991.
Inkeles, Alex, ed. On Measuring Democracy: Its Consequences and Concomitants.
New Brunswick, N.J.: Transaction, 1991.
Knack, Stephen, and Keefer, Philip. Institutions and Economic Performance: Cross-Country Tests Using Alternative Institutional Measures.
Econ. and Politics 7 (November 1995): 20727.
Lipset, Seymour Martin. Some Social Requisites of Democracy: Economic Development and Political Legitimacy. American Polit. Sci. Rev.
53 (March 1959): 69105.
. The Social Requisites of Democracy Revisited: 1993 Presidential
Address. American Sociological Rev. 59 (February 1994): 122.
Lipset, Seymour Martin; Seong, Kyoung-Ryung; and Torres, John Charles.
A Comparative Analysis of the Social Requisites of Democracy. Internat.
Soc. Sci. J. 45 (May 1993): 15575.
Mauro, Paolo. Corruption and Growth. Q. J.E. 110 (August 1995): 681712.
Miklukho-Maklaya Institute of Ethnography of the Academy of Sciences.
Atlas Narodov Mira. Moscow: Dept. Geodesy and Cartography, State Geological Comm. USSR, 1964.
Przeworski, Adam, and Limongi, Fernando. Political Regimes and Economic Growth. J. Econ. Perspectives 7 (Summer 1993): 5169.
Putnam, Robert D. Making Democracy Work: Civic Traditions in Modern Italy.
Princeton, N.J.: Princeton Univ. Press, 1993.
Sachs, Jeffrey D., and Warner, Andrew M. Natural Resource Abundance
and Economic Growth. Manuscript. Cambridge, Mass.: Harvard Inst.
Internat. Development, December 1995.
Schumpeter, Joseph A. Capitalism, Socialism, and Democracy. 2d ed. New York:
Harper, 1947.
Sirowy, Larry, and Inkeles, Alex. The Effects of Democracy on Economic
Growth and Inequality: A Review. Studies Comparative Internat. Development 25 (Spring 1990): 12657.
Summers, Robert, and Heston, Alan. The Penn World Table (Mark 5):
An Expanded Set of International Comparisons, 19501988. Q. J.E. 106
(May 1991): 32768.
. Penn World Tables, Version 5.6. Cambridge, Mass.: NBER, 1995.
www.nber.org.
Taylor, Charles Lewis, and Hudson, Michael C. World Handbook of Political
and Social Indicators. 2d ed. New Haven, Conn.: Yale Univ. Press, 1972.
Tocqueville, Alexis de. Democracy in America. Translated by Henry Reeve.
London: Saunders & Otley, 1835.