Health & Place 47 (2017) 90–99

Contents lists available at ScienceDirect

Health & Place

journal homepage: www.elsevier.com/locate/healthplace

Not all built the same? A comparative study of electoral systems and MARK
population health
⁎
Andrew C. Patterson
Prentice Institute for Global Population and Economy, University of Lethbridge, 4401, University Dr W, Lethbridge, AB, Canada T1K3M4

A R T I C L E I N F O A BS T RAC T

Keywords: Much literature depicts a worldwide democratic advantage in population health. However, less research
Autocracy compares health outcomes in the different kinds of democracy or autocracy. In an examination of 179 countries
Democracy as they existed between 1975 and 2012, advantages in life expectancy and infant health appear most reliably for
Elections democracies that include the principle of proportional representation in their electoral rules. Compared to
Infant mortality
closed autocracies, they had up to 12 or more years of life expectancy on average, 75% less infant mortality, and
Life expectancy
double the savings in overall mortality for most other age groups. Majoritarian democracies, in contrast, did not
experience longitudinal improvements in health relative to closed autocracies. Instead their population health
appeared to be on par with or even superseded by competitive autocracies in most models. Findings suggest that
the principle of proportional representation may be good for health at the national level. Implications and
limitations are discussed.

1. Introduction
Much literature asserts that citizens of democratic nations enjoy
better health than those living under autocratic rule. Studies for
example link democratic governance to longer life expectancy (Besley
and Kudamatsu, 2006; Wigley and Akkoyunlu-Wigley, 2011a; Lin
et al., 2012), lower mortality (Navia and Zweifel, 2003; Gerring
et al., 2012; McGuire, 2013; Franco et al., 2004; Álvarez-Dardet and
Franco-Giraldo, 2006), and lower rates of parasitic disease (Thornhill
et al., 2009), patterns which have been remarkably reliable across
studies (Muntaner et al., 2011).
Democracy alone does not guarantee the best health, however.
Meanwhile many autocracies assume some democratic features, such
as the legitimation of rule through elections in the post-communist era
(Levitsky and Way, 2010). Moreover, variation in GDP accounts for a
large share of cross-national differences in life expectancy (Patterson
and Veenstra, 2016), which may suggest democracies are healthier
simply because they are richer or that much of this difference is
spurious. Questions remain as to which kinds of governing structures
in particular support health the best and why that might be so.
To help illuminate potential linkages between political institutions
and population health, this study compares national health statistics
while specifying five different regime types – two autocratic and three
democratic – according to their electoral rules. In so doing, the working
presumption is that the ongoing obligations leaders have after they win
office, rather than just the electoral process itself, may be what matters
most for health.
1.1. How do electoral systems relate to population health?
Some of the most recognized arguments about the potential benefits
of democratic elections for health come from Amartya Sen's paper
(1994) on the global distribution of famines. In it he explains how
elections keep leaders accountable to the decisions they make, in that
they bear the brunt of poor policy choice in democracies especially. By
making foolish decisions (or failing to endorse wise ones) that can
impact many people, elected leaders risk decimating their supportive
votes and consequentially losing their contracts during a forthcoming
election. To avert that risk, the reasoning continues, these leaders will
tend to promote policies that protect the health of citizens.
An alternative view is that some social groups, such as the poor
working class, incur the most risk to health and yet in democracies can
use their votes to compete for policies that will improve their
circumstances. The reasoning follows that if democratic elections
provide a venue for underprivileged groups to push for egalitarian
social policy (Lipset, 1960), and yet if underprivilege (Link and Phelan,
1995, 2010) and inequality (Pickett and Wilkinson, 2015) are also
negative determinants of health, then elections act to lift population
health overall by incorporating feedback from those most at risk. The
emphasis here is thus more equity in policy choice, rather than wise

⁎
Correspondence to: Prentice Institute for Global Population and Economy, 4401, University Dr W, Lethbridge, AB, Canada T1K3M4.
E-mail address: andrew.patterson@uleth.ca.

http://dx.doi.org/10.1016/j.healthplace.2017.07.003
Received 5 January 2017; Received in revised form 21 June 2017; Accepted 6 July 2017
Available online 07 August 2017
1353-8292/ © 2017 Elsevier Ltd. All rights reserved.
A.C. Patterson
policies per se, while the political mechanism linking electoral democ-
racy to population health is more the representation of a diversity of
perspectives.
1.2. Types of electoral regime
Although this literature has offered other insightful explanations for
the correspondence between regime type and health, the above two
may be most useful for the purposes of this paper since they ostensibly
correspond with different types of electoral democracy. Of these, the
majoritarian system (also referred to as ‘first past the post’ and other
names), is perhaps the easiest to grasp: the winner of a contest is the
single candidate who amasses the most supporting votes, while the
losers do not take office and cannot control policy. In this system, only
one representative reports to each district, which makes them maxi-
mally accountable. Other electoral systems support proportional
representation. These are designed so that the distribution of party
affiliations among the winning candidates should mimic the distribu-
tion of preferences among the electorate. To represent a given
constituency, votes are recalibrated and apportioned in such a way
that runners-up still have the chance to win a seat, depending on how
many residual votes supported them. Power therefore does not shift
along partisan lines in quite as volatile manner in this type of system.
For these reasons, proportional electoral rules are argued to prioritize
broader congruence between the policy preferences of elected leaders
and the preferences of the voting public, while majoritarian systems
prioritize accountability with respect to past policy decisions (Powell,
2000). Some democracies adopt mixed systems that incorporate
aspects of both majoritarian and proportional electoral rules.
Many authoritarian governments also hold elections, albeit to serve
purposes other than to designate leadership. Discussed below as
competitive autocracies, Levitsky and Way (2010) explain how in
some countries a change of leadership is unlikely, but elections still
pose a tangible threat to those in power. This is because governments
in this kind of regime must either duly count and report votes from an
election or face substantial pressure to do so. That being the case,
incumbents violate generally accepted protocols to assure that the
majority will vote in their favor, such as by monopolizing the media or
suppressing private-sector support for the opposition, while still
passing as democratic under the watch of other countries. The
remaining countries serve as the reference category in this study.
Closed autocracies do not hold meaningful elections and instead use
physical force against their citizens, royal lineage, or other non-
democratic means to stay in power. Those that declare public elections
in which multiple parties compete do not oblige themselves to honor
the outcome.
1.3. Testing the consequences of electoral regimes for health
Few studies examine the correspondence between electoral systems
and population health while reaching beyond the conceptual dichotomy
between democracy and autocracy. Studies using two different mea-
sures of the proportionality of electoral rules find that this variable
predicts higher life expectancy and lower rates of infant mortality
(Wigley and Akkoyunlu-Wigley, 2011b; Gassner et al., 2006).
Comparing proportional and majoritarian democracies to autocracies,
Justesen (2012) concludes that only the former type of democracy
increases access to treatment for HIV/AIDS. Meanwhile competitive
autocracies appear to have lower rates of infant mortality and higher
rates of school enrolment compared to closed autocracies (Cassani,
2016).
Less clear is how resilient these patterns are to choice of samples
and methods, or when comparing different subtypes of autocracy and
democracy. Also unclear are the particular causal mechanisms linking
electoral institutions to health. To address these gaps, the following
91
Health & Place 47 (2017) 90–99
study examines life expectancy, rates of infant mortality, and age-
adjusted mortality in older age groups for a larger sample of countries
and years while using a more diversified conception of electoral
regimes. Some models additionally consider economic and policy
outcomes that may be germane to health. The view that “wealthier is
healthier” (Pritchett and Summers, 1996; also see Deaton, 2013, p. 34
regarding logged income and life expectancy) at the same time that
some regimes may be more adept at growing income than others
(Gerring et al., 2005; Knutsen, 2011, 2012) would suggest the latter
might be a mediator. Other, much discussed claims are that democ-
racies prevent famine more effectively (Sen, 1994) and that access to
high-quality food might explain linkages between health and higher
standards of living (McKeown et al., 1972), which may suggest a
mediating role for food supply. These are admittedly controversial
claims, but merit some consideration. Perhaps more straightforward
claims come from the view that electoral rules affect a country's
investments in public infrastructures and social safety nets (Persson
and Tabellini, 2004, 2005). If variables like income inequality harm
health (Pickett and Wilkinson, 2015) while investment in services
helps, these could identify linkages between electoral regimes and
population health. After exploring these different pathways through
mediation analysis below, this study compares results from multilevel
models that parse apart cross-sectional and longitudinal effects.
2. Methods
2.1. Population data
Analysis is based on annual data for 179 countries as they existed
from 1975 through 2012. The Database of Political Institutions (Beck
et al., 2001) was used to categorize electoral regimes. World Bank
Indicators were appended to represent national vital statistics, eco-
nomic prosperity, and national health expenditures. As a first step to
address missing values, data from the United Nations National
Accounts Database and the 1997 historical supplement to the United
Nations Demographic Yearbook were assumed where available and
where the World Bank Indicators were absent. The United Nations
Population Division provided quinquennial, age-specific probabilities
of mortality from any cause. Nutritional data were taken from the Food
Balance Sheets of the United Nations Food and Agricultural
Organization. The Standardized World Inequality Database (Solt,
2014) provided figures on income inequality. Latitude, which played
a role in the multiple imputation, was taken from La Porta et al. (1999).
2.2. Variables
2.2.1. Electoral regime type
A five-part categorical variable is used to distinguish electoral
regimes. A country qualifies as democratic only if it scores ‘7’ on both
of the variables EIEC and LIEC from the Database of Political
Institutions, which measure the degree to which free and fair elections
determine choice of leadership in the executive branch and legislature,
respectively. Democracies with majoritarian electoral rules score ‘1’ on
the variable PLURALTY (‘pluralism’ being another name for the
majoritarian system). Democracies with proportional electoral rules
score ‘1’ for the variable PR. Some countries meet both criteria and
they are regarded as having mixed electoral systems. Often these are
the ‘mixed member’ type of system as in Germany, in which one
representative is chosen via the majoritarian method and another is
chosen through a proportional logic. A competitive autocracy is any
country scoring between ‘6’ and ‘7’ for either EIEC or LIEC, but not ‘7’
for both. This is in the spirit of Levitsky and Way's (2010) construct of
competitive authoritarianism, but is not the operational definition they
use, so the current term is used instead. The remaining countries are
regarded as closed autocracies and serve as the referent in statistical
A.C. Patterson Health & Place 47 (2017) 90–99

analysis. (See Appendix A, available online as a supplementary table, Table 1

for detail on how electoral regime type was operationalized for each Descriptive statistics.
country.)
1975 2012 All years

2.2.2. Dependent variables Mean SD Mean SD Mean SD

This study examines predictors of life expectancy, infant mortality
Regime
rates, and age-adjusted mortality for groups aged 5–24, 25–44, 45–
characteristics
64, and over 65. The latter figures comprise aggregate mortality Regime tenure 1.000 0.000 19.497 12.421 11.430 9.159
probabilities that are weighted according to the worldwide distribution Electoral fraud 0.043 0.205 0.127 0.334 0.104 0.306
of age groups as they existed at the start of the study period. Sensitivity Regime types
analysis (not shown) indicated that when using a later standardizing Closed autocracy 0.587 0.494 0.121 0.328 0.324 0.468
(reference)
year model coefficients generally decreased for the group aged 65 or
Competitive 0.188 0.392 0.225 0.419 0.206 0.404
older, but substantive findings as discussed below did not change. autocracy
Similar to findings from Welander et al. (2015), results based on World Majoritarian 0.072 0.260 0.202 0.403 0.141 0.348
Bank figures for the mortality rates of children five years of age and democracy
Proportional 0.116 0.321 0.254 0.437 0.204 0.403
under were similar and substantive conclusions identical to those based
democracy
on infant mortality, so only the latter are presented. Mixed 0.036 0.188 0.197 0.399 0.126 0.332
democratic
2.2.3. Control variables systems
To reduce threats of confounding, some control variables were Economic and policy
performance
included in all models. Year of data collection is an integer minimized
GDP per capita 8.474 1.689 8.399 1.508 8.430 1.709
at zero to represent 1975 that increases by one for each additional year. Food supply 2.452 0.520 2.537 0.341 2.635 0.518
This variable was included to reduce the confounding influence of Public health 4.019 2.039 5.118 1.877 4.471 2.057
global trends in health. Taken in logarithmic form, population size is expenditure
Private health 3.784 1.562 4.760 1.481 4.214 1.585
another potential complication that Justesen (2012) suggests may
expenditure
facilitate economies of scale in service delivery on the one hand while Gini (net) 34.329 10.772 36.248 7.855 36.861 9.735
on the other hand imposing the need to cater to a more diverse set of Population health
preferences among the electorate. Wigley and Akkoyunlu-Wigley Life expectancy 65.266 10.141 69.819 9.471 64.906 10.761
(2011b) also point out that it may increase party fragmentation in Infant mortality 45.625 38.309 27.322 24.623 47.624 41.118
ratea
larger countries and/or reduce electoral proportionality in smaller
Age-adjusted
countries. Regime tenure is the length of time any given electoral mortality rateb
regime has been in power and is measured as the number of years a 05–24 12.170 10.199 6.116 6.362 10.146 11.192
country has remained one of the five types of electoral regime 25–44 21.979 14.212 15.941 16.352 21.561 20.174
45–64 63.624 22.050 45.642 20.861 58.698 25.293
continuously, without changing to another type. Given how many
65+ 332.485 46.642 281.449 49.846 316.220 54.429
countries democratized during the study period, this variable is
included to reduce the confounding influence of a country's long-term a
Infant mortality rate is per 1000 population at birth.
experience in policy implementation irrespective of the particular b
Age-adjusted mortality rates are per 1000 population aggregated in 5-year intervals
regime type it has adopted. Regime tenure is also used as the measure and are standardized to be consistent with the worldwide age distribution as of 1975.
These estimates reflect averages over the years 1975-1980, 2005–2010, and all
of time in models testing the longitudinal impact of electoral regimes
observations combined.
on health. Because countries exist within a global political environment
that presumably benefits some more than others, most models
adding a constant of 1 to the argument. Descriptive statistics are
additionally control for world region. Indicators are included to
reported in Table 1.
confirm whether countries are landlocked as well as location in sub-
Saharan Africa, North Africa and the Middle East, Southeast Asia, or
Latin America, with the remaining countries being the referent. Finally, 2.3. Missing data and multiple imputation
several of the models control for the presence of electoral fraud. This is
in the spirit of hypotheses linking electoral institutions to population Missing values impacted cases that were disproportionately auto-
health but also out of appreciation for its potential to complicate this cratic and poor. Care was therefore taken to address the issue of
relationship by harming the integrity of democratic institutions, missing data in the working dataset. Combining the Database of
confusing the measurement thereof, etc. Political Regimes with the World Bank Indicators excludes subnational
territories and protectorates, countries whose sovereignty is contested,
2.2.4. Potential mediators countries with populations of less than 200,000, and some cases and
Some models consider additional variables that correlate with both occasions in which countries were occupied by a foreign power or were
political institutions and health, which may help identify the mechan- undergoing regime change. Also excluded are the former country of
isms linking the two. Per-capita gross domestic product (GDP) at Serbia and Montenegro and its seceding countries. Since these exclu-
current exchange rates is taken as the measure of economic prosperity, sions were decided arbitrarily by the makers of the databases rather
reflected in 2010 United States dollars and then logged. Food supply than the countries recusing themselves from reporting their data, they
assesses each country's total food availability and is measured as are thought of as ‘missing at random’ and are disregarded in analysis.
calorie consumption in units of a thousand kilocalories per capita per In some cases, data exist for territories and constituent polities but
day. The net Gini coefficient measures country-level income inequality. not for the countries to which these belonged. World Bank Indicators
To calculate public health expenditure, the World Bank Indicator for have largely complete records for each of the former Soviet states, for
per-capita health expenditure is reflected in 2010 United States dollars, example, but do not acknowledge the USSR as a distinct country. In
then multiplied against the Indicator representing the percentage these situations, values for the uniting country were estimated as a
thereof that is public. Private health expenditure is similarly calcu- weighted estimate of the mean based on the respective values and
lated, but by multiplying against the complement of this percentage. population sizes from the constituent countries. In other cases some
Both measures of health expenditure are expressed in logged form after datasets acknowledged only the uniting country. World Bank records

92
A.C. Patterson Health & Place 47 (2017) 90–99

Table 2
Models predicting life expectancy and infant mortality rates according to electoral regime type.a

Life expectancy Infant mortality

b
Model 1 Model 2 Model 3 Model 4b

Regime tenure 0.48 (0.39 to 0.58) 0.28 (0.21 to 0.34) −5.4% (−6.3% to −4.5%) −3.9% (−4.7% to −3.2%)
Electoral fraud −5.56 (−8.10 to −3.01) −2.03 (−3.67 to −0.38) 53.9% (29.9 to 82.3%) 24.1% (8.5 to 41.9%)
Closed autocracy (ref.)
Competitive autocracy 7.47 (4.92 to 10.03) 4.34 (2.68 to 6.01) −53.3% (−62.2% to −42.3%) −40.4% (−50.9% to −27.5%)
Majoritarian democracy 9.21 (6.24 to 12.18) 4.45 (2.07– 6.84) −65.0% (−72.9% to −54.9%) −43.7% (−56.8% to −26.6%)
Proportional democracy 12.68 (10.05 to 15.31) 6.90 (4.91 to 8.89) −76.1% (−81.4% to −69.2%) −67.9% (−75.4% to −58.0%)
Mixed democratic systems 14.26 (11.31 to 17.20) 7.86 (5.56 to 10.16) −75.8% (−83.0% to −65.4%) −65.3% (−76.0% to −49.7%)
Countries 179 179 179 179
Country-years 5914 5914 5914 5914

a
All models control for regime tenure, log population size, and year of data collection to rule out worldwide trends in health.
b
Models 2 and 4 additionally control for landlocked countries and world region.

for GDP in Germany during the Cold War era are parsed into records
separately representing East Germany and West Germany by compar-
ing their relative performance in version 5.6 of the Penn World Tables.
Since northern and southern counterparts of Yemen and Vietnam had
identical regime types, analysis only acknowledges the uniting country.
After taking the above steps, relatively few missing values remained
for GDP (n = 10 or 0.2%), infant mortality (n = 58 or 0.9%), and food
supply (n = 497 or 8.3% through 2011). Records were virtually
complete for crude death rates and crude birth rates, which were used
in the procedure for multiple imputation. Listwise deletion was applied
to missing values for net Gini in the models testing its mediating
potential, however, since almost half of the values for this variable were
missing. The same was done with respect to health expenditures, World
Bank Indicators for which were unavailable prior to 1995, although
missing values impacted only a small proportion of observations from
that point forward (n = 56, or 1.8% of the total). Caution is encouraged
when interpreting results from models that test the latter variables
especially.
Regression models were used to impute missing values for life
expectancy based on the annual global trend in this outcome alongside
crude birth rates, crude death rates, and infant mortality rates. Models
for imputing infant mortality rates and per-capita GDP were based on
the annual linear trend in the respective outcome within each country,
taking advantage of the fact that all countries had these data available
in regular intervals if not for each year. Mortality rates for other age
groups were also examined. A similar algorithm was used to impute
missing mortality rates for children under 5 based on the other health
statistics. However, results were similar to those relating to infant
mortality so only the latter are presented. Mortality probabilities for
older age groups were only available in multiyear quantities. These
were appended to rows of data coming from the year prior to the start
of each reporting period and the rest were deleted listwise in the
models of mortality over the life course. Age-specific mortality was
otherwise missing only for East and West Germany in 1979, 1984, and
1989, being available only for the uniting country. Such a small
proportion of missingness (0.5%) was regarded as non-threatening to
the conclusions of the study, however, so listwise deletion was used to
address these cases in lieu of imputation. For those cases in which food
supply was mostly or entirely missing, these were imputed using the
annual global trend along with GDP per capita and latitude as
predictors. For those cases that were missing values for food supply
on singular occasions, these were imputed assuming only the annual
linear trend within each country. Parameters for all imputation models
were estimated based on the known data. To impute each missing
value, the conditional mean was first calculated, then a random
number was drawn from the presumable distribution of residuals
and added to the conditional mean. Results presented below are based
on 30 randomly imputed datasets.
2.4. Statistical analysis
Regression models test the correspondence between electoral
systems and both life expectancy and mortality. Observations were
laid out in country-years and the dependent measures were lagged by
one year to curtail endogeneity. For linear and generalized linear
regression models, Newey-West standard errors were used to estimate
statistical significance in the presence of autocorrelation. Following
these, multilevel models with random intercepts and slopes provided
an alternative solution to autocorrelation.
Statistical tests were performed in three phases. The first phase
examined the contemporaneous relationship between electoral regime
type and health. Linear regression using ordinary least squares was
used to predict life expectancy, while log-linked gamma regression
predicted infant mortality rates and age-adjusted mortality for other
age groups. All models controlled for year of data collection, regime
tenure, and log population size. Most models additionally controlled for
world region and status as landlocked. The second phase considered
socioeconomic and policy correlates for the sample of countries by
additionally controlling for GDP per capita, food supply, net Gini, and
both public and private health expenditures per capita. Supplementary
analysis then tested whether electoral regime type predicts change in
these variables over time. Income inequality and health expenditures
were analyzed separately given the disparate and unresolved patterns
of missingness in these variables. (Appendix B provides a visual
depiction of the tests of mediation.)
The final phase used multilevel models as an alternative method
that allows tests of longitudinal change. Models follow recommenda-
tions from Fairbrother (2014) to parse apart the between-groups
effects from the within-groups effects, essentially comparing the
cross-sectional and longitudinal components of the relationship be-
tween electoral regimes and health. This involved creating two types of
variable, derived from the dummy variables used thus far to represent
each regime type. The first variable is the mean average for each
dummy and is interpretable as the proportion of years that a country
has spent under the respective electoral rules. In models it identifies
the cross-sectional correspondence between electoral systems and
health. For each regime type in each country-year, this mean score
was then subtracted from the respective dummy variable so as to create
a second variable representing short-term changes in electoral rules
within the study period. This second variable is used to test the
quantity of short-term longitudinal change. Based on these two
variables, the first model for each health outcome compares the
cross-sectional and longitudinal influence of the different types of
electoral regime while including country-level random intercepts and
slopes. Country-level change in regime tenure was used to calculate the
latter. The model was then converted to a growth curve model by
additionally interacting the fixed effect for regime tenure with the first

93
A.C. Patterson Health & Place 47 (2017) 90–99

Table 3
Models testing the intervening influence of per-capita GDP, food supply, public/private health expenditures, and net gini on life expectancy.a

Model 1b Model 2 Model 3 Model 4 Model 5

Regime tenure 0.07 (0.01 to 0.13) 0.07 (0.01 to 0.13) 0.06 (0.00 to 0.13) 0.07 (0.01 to 0.14) 0.07 (0.01 to 0.14)
Electoral fraud −0.97 (−2.54 to 0.60) −0.84 (−2.78 to 1.11) −0.78 (−2.78 to 1.22) −0.24 (−1.73 to 1.26) −0.48 (−1.89 to 0.92)
Closed autocracy (ref.)
Competitive autocracy 2.13 (0.74 to 3.51) 1.83 (−0.01 to 3.67) 1.69 (−0.38 to 3.77) 1.21 (−0.60 to 3.01) 1.30 (−0.45 to 3.05)
Majoritarian democracy 1.40 (−0.40 to 3.21) 0.70 (−1.40 to 2.80) 0.57 (−1.60 to 2.74) 0.72 (−1.55 to 3.00) 0.91 (−1.31 to 3.13)
Proportional democracy 1.60 (0.01 to 3.18) 1.78 (−0.49 to 4.05) 1.60 (−0.66 to 3.86) 0.84 (−1.08 to 2.76) 0.58 (−1.27 to 2.43)
Mixed democratic systems 2.71 (0.98 to 4.44) 2.90 (0.24 to 5.55) 2.73 (−0.05 to 5.50) 2.09 (−0.33 to 4.51) 2.07 (−0.23 to 4.36)
GDP per capita 2.32 (1.70 to 2.94) 2.06 (1.24 to 2.89) 1.59 (−0.31 to 3.49) 2.23 (1.49 to 2.96) 2.17 (1.48 to 2.86)
Food supply 1.65 (0.19 to 3.11) 1.86 (−0.11 to 3.83) 1.76 (−0.41 to 3.92) 2.12 (0.27 to 3.96) 1.79 (0.00 to 3.57)
Private health expenditures 0.21 (−1.02 to 1.45)
Public health expenditures 0.30 (−0.96 to 1.55)
Gini (net) −0.11 (−0.22 to 0.00)
Countries 179 171c 171c 159d 159d
Country-years 5914 2835c 2835c 3545d 3545d

a
All models control for year of data collection, log population size, landlocked countries, and world region.
b
The coefficients represented in Model 1 were similar, but slightly larger when removing food supply from the model.
c
Sample size for Models 2 and 3 is reduced after eliminating observations missing health expenditure data, which includes all country-years occurring prior to 1995.
d
Sample size for Models 4 and 5 is reduced after eliminating observations missing net Gini.

of the abovementioned variables for electoral regime type, the country-
averaged score. The interaction effects were then assumed to test the
rate of longitudinal change as it corresponds with electoral regime
type.
3. Results
autocracies have worse infant health compared to the other electoral
regimes. From a hypothesis testing perspective, only purely propor-
tional democracies had smaller infant mortality rates than either type
of autocracy at α = 0.05 in this model. Again affirming the importance
of democratic elections overall, infant mortality is roughly 25–50%
higher in countries characterized by electoral fraud.

3.1. Contemporaneous associations 3.2. Economic and policy correlates

Model 1 of Table 2 points to large differences in life expectancy for
the three different types of democratic electoral systems compared to
closed autocracies, with majoritarian systems having 9 additional
years, purely proportional systems having 12 additional years, and
systems that combine majoritarian and proportional electoral rules
having the superior advantage of 14 years. 95% confidence intervals
(indicated in parentheses) comparing the three democratic regimes
overlap, which is to say that differences among these are non-
significant. However, by the same logic only proportional and mixed
democracies have levels of life expectancy that exceed competitive and
closed autocracies. Two other aspects of Model 1 otherwise suggest that
any improvement to the fairness of the electoral system overall
corresponds with health advantages. First, competitive autocracies still
have 7 years more of life expectancy than closed autocracies. Second,
fraudulent manipulation of elections corresponds with 6 fewer years on
average. Robustness checks (not shown) suggested the dummies for
competitive autocracy and electoral fraud measure distinct phenom-
ena.
After controlling for world region and status as landlocked, Model 2
estimates smaller advantages of about four years for majoritarian
democracy and competitive autocracy, but around seven years for
either of the two regimes that include components of proportional
representation. Overlapping confidence intervals indicate that all
electoral systems including competitive autocracy offer advantages
over closed autocracy in terms of life expectancy, but are not distinct
from each other.
The next models depict a similar correspondence with infant
mortality. According to Model 3 competitive autocracies have the
smallest advantage over closed autocracies, proportional and mixed
democracies have the largest, and majoritarian regimes fall somewhere
between these but are not significantly different from competitive
autocracies. Countries with purely proportional and mixed electoral
rules experienced reductions in rates of infant mortality by as much as
75% compared to closed autocracies, advantages which were distinct
from both closed and competitive autocracies at α = 0.05. Controlling
for regional characteristics, Model 4 again indicates that closed
The models shown in Tables 3 and 4 test the impact of electoral
regimes on health while controlling for factors relating to economic
performance and health policy. Samples vary according to the avail-
ability of data for the Gini coefficient and health expenditures, so
caution is urged in the interpretation of these results. After further
controlling for GDP and food supply in Model 1 of Table 3, model
coefficients predicting life expectancy reduce in size by 51% for
competitive autocracies and as much as 76% for the three democracies.
Competitive autocracies and both purely proportional and mixed
democracies retain statistically significant differences in life expectancy
compared to closed autocracies, but majoritarian democracies do not.
These changes are largely attributable to the inclusion of per-capita
GDP, beyond which the inclusion of food supply only modestly reduces
the size of the corresponding model coefficients (not shown). Neither
health expenditures nor Gini appear convincingly to mediate relation-
ships with life expectancy in the remaining models.
Table 4 presents results from models testing the same set of
predictors against infant mortality. These results are similar, but
additionally point to the potential importance of public health expen-
ditures. With each unit increase in the logged per-capita amount
reducing this outcome by 20% this variable emerges as a statistically
significant predictor in Model 3 and its inclusion eliminates the
correspondence with GDP. The same cannot be said for private
investment in health. Results then suggest that public investment in
health infrastructure may explain the correspondence between GDP
and infant mortality.
Additional models tested whether variation in regime type pre-
dicted prospective change in GDP per capita, food supply, public health
expenditure, or net Gini (see Appendix C). After lagging each of these
by ten years to calculate decade-long change, these models controlled
for values in the respective variable as of the start of the decade as well
as the full set of controls used thus far. Overlapping confidence
intervals for model coefficients suggested change in net Gini did not
vary according to electoral regime, nor did growth in health expendi-
tures. However, it is worth noting that the magnitudes for the
coefficients conformed to the patterns observed thus far, with propor-

94
A.C. Patterson Health & Place 47 (2017) 90–99

tional and mixed democracies appearing more egalitarian on both

−27.3% (−39.5% to −12.6%)

−32.8% (−36.7% to −28.6%)
−13.8% (−25.6% to −0.1%)

−23.7% (−36.8% to −7.8%)

−0.9% (−1.5% to −0.3%)
measures. All electoral regime types improved upon closed autocracies

−14.0% (−28.0% to 2.6%)

−10.2% (−22.9% to 4.5%)

3.1% (−7.2% to 14.5%)

1.6% (0.9 to 2.3%)

with respect to the other outcomes, however. Proportional democracies
had the strongest growth in GDP at 51% more than closed autocracies
Model 5

3545d
(exp[β] − 1 = 50.7%; 95% C.I. = 27.7%, 77.9%), followed closely by

159d
mixed systems (exp[β] − 1 = 45.7%; C.I. = 17.9%, 80.1%), while
competitive autocracies (exp[β] − 1 = 30.7%; 95% C.I. = 12.5%, 51.9%)
and majoritarian democracies (exp[β] − 1 = 22.2%; 95% C.I. = 7.4–
53.4%) 1.2%, 47.4%)appeared equivalent to each other with modest
growth. Proportional democracies had the strongest growth in food
supply at 140 calories per capita per day (β =0.14; 95% C.I. = 0.08,
0.20), followed closely by mixed (β = 0.10; C.I. = 0.04, 0.17)
−33.1% (−37.6% to −28.3%)

and majoritarian democracies (β = 0.07; C.I. = 0.01, 0.15), while

−26.6% (−41.0% to −8.5%)
−27.3% (−41.3% to −9.9%)
−0.9% (−1.6% to −0.3%)
−0.6% (−11.3% to 11.3%)

−13.3% (−25.8% to 1.3%)

−11.6% (−27.4% to 7.7%)

−15.0% (−29.2% to 2.1%)

competitive autocracies had the smallest advantage (β =0.05;

C.I. =0.00, 0.10).
Model 4

3545d
159d

3.3. Multilevel models

Results presented in Table 5 used multilevel modeling techniques

to distinguish cross-sectional and longitudinal relationships between
Sample size for Models 2 and 3 is reduced after eliminating observations missing health expenditure data, which includes all country-years occurring prior to 1995.

electoral regime type and health. Model 1 indicates that all democracies
have a cross-sectional advantage in life expectancy compared to closed
autocracies. Purely proportional electoral regimes have the greatest
advantage at 9 years of added life expectancy compared to closed
−31.0% (−48.8% to −7.0%)
−32.0% (−48.8% to −9.6%)

−19.6% (−29.7% to −8.1%)

−0.8% (−1.4% to −0.2%)

−12.2% (−33.1% to 15.2%)

autocracies, while mixed regimes have 6 and majoritarian democracies

−15.6% (−33.1% to 6.6%)

−11.8% (−25.7% to 4.8%)

−12.6% (−27.0% to 4.7%)
−2.6% (−13.7% to 9.9%)
10.0% (−8.3% to 31.9%)
Models testing the intervening influence of per-capita GDP, food supply, public/private health expenditures, and net gini on infant mortality rates.a

again lag behind the other two at 5 years. Of the democracies only those
The coefficients represented in Model 1 were similar, but slightly larger in absolute magnitude when removing food supply from the model.

with mixed electoral rules experienced longitudinal gains in life

Model 3

2835c

expectancy according to this model. Contrasting with the other results,

171c

Model 1 shows no cross-sectional advantage for competitive autocra-

cies, but suggests instead that these countries experienced longitudinal
gains in life expectancy. None of the interaction terms in Model 2 are
statistically significant, however, suggesting that the rate of change in
life expectancy is similar among the different types of electoral regimes.
Model 3 shows again that only the democracies have a cross-
sectional advantage in infant health compared to closed autocracies.
−36.7% (−52.8% to −15.3%)
−37.1% (−52.0% to −17.6%)
−30.3% (−35.6% to −24.6%)
−18.7% (−32.7% to −1.9%)

Proportional and mixed systems have 70% less infant mortality,

−0.9% (−1.5% to −0.3%)

All models control for year of data collection, log population size, landlocked countries, and world region.
−18.0% (−33.3% to 0.9%)
−15.9% (−34.6% to 8.1%)
13.8% (−8.7% to 41.8%)

compared with a smaller advantage of 46% in majoritarian systems.

However, proportional and mixed democracies were the only type of
Model 2

electoral regime to improve infant health longitudinally. Especially

2835c
171c

striking are the disparate confidence intervals for the longitudinal

Sample size for Models 4 and 5 is reduced after eliminating observations missing net Gini.

components in Model 3, which indicate that only purely proportional

and mixed democracies, and especially the latter, experienced short-
term longitudinal improvements in infant health relative to closed
autocracies. On the other hand, the interaction terms in Model 4
suggest the rate of decline in infant mortality was faster only for purely
proportional democracies and competitive autocracies. This decrease
was about 1% faster than closed autocracies for each year of rule. A
−35.1% (−47.1% to −20.3%)
−33.9% (−45.9% to −19.3%)
−28.5% (−32.8% to −23.9%)
−16.7% (−25.05% to −7.5%)
−17.6% (−26.8% to −7.4%)
−1.4% (−2.0% to −0.8%)

−14.7% (−28.6% to 1.9%)

similar (albeit non-significant) rate of decline appears for mixed

12.9% (−4.4% to 33.3%)

electoral systems.
In summary, most electoral regimes demonstrated some kind of
Model 1b

5914
179

longitudinal improvement in population health, but majoritarian

systems did not. This finding was robust to the removal of random
slopes and controlling for GDP and food supply (see Appendix D).

3.4. Electoral regimes across the life course

Taking all age groups together, results from models testing age-
adjusted mortality as a single dependent variable (not shown) essen-
Private health expenditures
Public health expenditures
Mixed democratic systems

tially replicated those pertaining to life expectancy. Some results were

Proportional democracy
Majoritarian democracy
Closed autocracy (ref.)
Competitive autocracy

distinctive to particular age groups, however, when these were com-

pared in separate models (Appendix E1). First, the protective influence
GDP per capita
Electoral fraud
Regime tenure

of electoral regimes appeared to diminish monotonically across the life

Country-years
Food Supply

course but was still present even for those aged 65 or older. Initial
Gini (net)
Countries

models were similar to those presented in Table 2, in that all four

Table 4

electoral regime types tended to have lower mortality rates than the
d
b
a

reference group of closed autocracies. Age-adjusted mortality in

95
A.C. Patterson Health & Place 47 (2017) 90–99

Table 5
Multilevel models testing the longitudinal impact of political regime type on population health.*

Life expectancy Infant mortality

Model 1 Model 2 Model 3 Model 4

Regime tenure 0.04 (0.01 to 0.07) 0.05 (−0.02 to 0.12) −0.4% (−0.8% to 0.1%) −0.3% (−0.8% to 0.2%)
Closed autocracy (reference)
Competitive autocracy (country mean) 1.44 (−1.89 to 4.77) 1.13 (−2.68 to 4.94) 17.9% (−13.4% to 60.7%) 28.1% (−12.3% to 87.0%)
Competitive autocracy (short-term change) 0.44 (0.17 to 0.71) 0.44 (0.17 to 0.71) 0.0% (−1.5% to 1.6%) 1.2% (−0.4% to 2.9%)
Majoritarian democracy (country mean) 5.22 (1.41 to 9.03) 5.93 (1.64 to 10.21) −46.2% (−65.2% to −16.7%) −47.0% (−65.7% to −18.2%)
Majoritarian democracy (short-term change) 0.19 (−0.15 to 0.53) 0.20 (−0.14 to 0.54) 0.6% (−1.5% to 2.7%) 0.5% (−1.6% to 2.6%)
Proportional democracy (country mean) 9.23 (6.08 to 12.38) 10.07 (6.56 to 13.58) −70.0% (−79.0% to −57.0%) −69.5% (−78.7% to −56.4%)
Proportional democracy (short-term change) 0.31 (−0.08 to 0.69) 0.31 (−0.08 to 0.70) −5.1% (−7.3% to −3.0%) −5.1% (−7.2% to −2.9%)
Mixed democracy (country mean) 6.16 (2.49 to 9.83) 6.08 (1.99 to 10.18) −70.4% (−80.5% to −55.1%) −69.8% (−80.1% to −54.3%)
Mixed democracy (short-term change) 0.87 (0.45 to 1.30) 0.87 (0.44 to 1.30) −10.0% (−12.3% to −7.7%) −10.0% (−12.2% to −7.7%)
Competitive (country mean) X regime tenure 0.02 (−0.11 to 0.16) −1.0% (−1.9% to −0.1%)
Majoritarian (country mean) X regime tenure −0.05 (−0.18 to 0.09) 0.8% (−0.2% to 1.8%)
Proportional (country mean) X regime tenure −0.06 (−0.16–0.05) −0.9% (−1.7% to −0.2%)
Mixed (country mean) X regime tenure 0.01 (−0.12 to 0.13) −0.8% (−1.7% to 0.1%)
% variation explained (intercept) 34.1% 34.3% 31.4% 31.7%
% variation explained (slope) 39.9% 40.4% 61.0% 62.9%
Countries 179 179 179 179
Country-years 5914 5914 5914 5914

*
All models control for year of data collection, log population size, landlocked countries, and world region.

majoritarian democracies appeared similar, if not somewhat smaller in
younger age groups, compared to competitive authoritarian regimes.
Countries with mixed electoral regimes by contrast had the largest
protections against mortality across all age groups and these were
followed closely by purely proportional electoral regimes. A more
discernible gradient appeared across regime types for those aged 65
or older and controlling for regional characteristics, with majoritarian
systems (β = −9.2%; C.I. = −14.0%, −4.0%) having a modest advantage
over competitive autocracies (β = −7.6%; C.I. = −10.6%, −4.4%) while
purely proportional systems had somewhat less of an advantage (β =
−13.9%; C.I. = −16.9%, −10.7%) than mixed systems (β = −17.4%; C.I.
= −21.2%, −13.4%), although these together again had the largest
advantages. Comparing to closed autocracies, mortality for the younger
age groups otherwise conformed to the patterns observed thus far – the
clearest distinction still being between electoral regimes based (partly
or completely) on the principle of proportional representation and the
remaining countries, which either use majoritarian rules or violate
electoral protocols in ways that do not qualify them as ‘democratic’.
Patterns similar to the above appeared when adding the proposed
mediators to models predicting age-adjusted mortality (Appendix E2).
Advantages in survivability were smaller after including food supply
and GDP in the models, which was mostly attributable to the latter
variable (not shown). Nevertheless electoral regimes tended to retain
statistically significant advantages relative to closed autocracies. This
was not true for majoritarian democracy in any age group, however,
while statistically significant advantages appeared across all age groups
only for mixed systems. Other models again pointed to a potential
mediating role for public health expenditures. Among those aged 65
years or older in a model equivalent to Table 3, Model 3, each logged
dollar in per-capita public health expenditure corresponded with a
reduction in mortality by 3% (β = −3.0%; −5.6%, −0.4%). This
compares with a smaller and statistically non-significant impact of
private health expenditures (β = −1.8%; C.I. = −3.9%, 0.4%). Yet after
including these in the model, the coefficient representing per-capita
GDP dropped three quarters in size to β = −1.6% (C.I. = −5.2%, 2.1%).
The coefficient representing per-capita GDP was halved in a similar
model for those aged 5–24, although in this case neither variable for
health expenditure was statistically significant. Food supply and net
Gini did not clearly exert any mediating effects independent of GDP.
Other distinctive results come from the multilevel models and these
are presented in Table 6. The cross-sectional component of these
models showed protection against mortality to be much smaller for
majoritarian regimes than for the other two types of democracy and
only modestly larger than for competitive authoritarian regimes. Tests
of short-term longitudinal change showed statistically significant
decline only for mixed electoral systems, a pattern which persisted
across all age groups, although purely proportional systems came close
for those aged 5–24 (β = −5.9%; C.I. = −11.9%, 0.5%). Electoral regime
type also had bearing on the rate of decline in mortality, but only for
those aged 45 and older. For these older age groups a clear gradient
was apparent. Rates of mortality decline were similar comparing
competitive authoritarian regimes and closed autocracies, while major-
itarian electoral regimes enjoyed modestly faster declines. These fell
short of those experienced by purely proportional and (especially)
mixed electoral regimes, however - again a pattern consistent with the
other results.
Taken together, these results suggest any system that provides even
a remote chance of conferring power to the opposition in an electoral
contest will have advantages in population health compared to the
world's strictest autocracies. However, the largest and/or most con-
sistent advantages again appeared for electoral regimes based (partly
or completely) on the principle of proportional representation.
4. Discussion
Findings support some of the long-standing conclusions of this
literature, but also offer potential embellishments. On the one hand,
model results agree with much research that demonstrates a global
democratic advantage in population health. Compared with autocratic
regime types that do not use elections to designate leadership,
democracies tended to have better population health over the past
half-century. This advantage persists in models based on disparate
statistical approaches, with respect to two different health outcomes,
and controlling for regional characteristics and other likely confoun-
ders.
Upon more careful inspection, though, results do not altogether
support the conceptual dichotomy between autocracy and democracy
that this literature has often presumed. First, competitive autocracies
are estimated to have stronger health relative to closed autocracies,
which supports the view that even modest gains in implementing
elections to choose a country's leadership, though they may not
necessarily be fair, correspond with better health outcomes. Yet
democracies too cannot necessarily be summarized under a singular
category vis-à-vis autocracies. Other results suggest that democracies

96
A.C. Patterson Health & Place 47 (2017) 90–99

Table 6
Multilevel models testing the longitudinal impact of political regime type on mortality across the life course.a

Model 1 Model 2 Model 3 Model 4

Age range 5–24 25–44 45–64 65+

Regime tenure −0.4% (−1.0% to 0.1%) −0.7% (−1.33% to −0.03%)b −0.2% (−0.5% to 0.1%) 0.0% (−0.1% to 0.1%)
Closed autoc. (reference)
Competitive autoc. (country mean) −19.2% (−43.1% to 14.9%) 1.0% (−25.0% to 36.0%) 5.2% (−10.0% to 23.1%) 0.7% (−5.1% to 6.9%)
Competitive autoc. (short-term change) −4.0% (−8.3% to 0.5%) −1.3% (−7.3% to 5.0%) −1.0% (−4.1% to 2.1%) −0.2% (−1.2% to 0.8%)
Majoritarian democ. (country mean) −27.8% (−51.2% to 6.9%) −11.5% (−36.4% to 23.0%) −9.9% (−24.5% to 7.5%) −7.0% (−13.1% to −0.5%)
Majoritarian democ. (short-term −0.8% (−6.4% to 5.1%) 1.4% (−6.3% to 9.7%) −0.9% (−4.7% to 3.1%) −0.6% (−1.9% to 0.7%)
change)
prop. democ. (country mean) −63.1% (−73.4% to −48.6%) −51.0% (−63.3% to −34.6%) −30.6% (−40.5% to −19.1%) −14.4% (−19.2% to −9.3%)
prop. democ. (short-term change) −5.9% (−11.9% to 0.5%) −2.7% (−11.1% to 6.4%) 0.4% (−4.0% to 4.9%) −0.8% (−2.2% to 0.7%)
Mixed democ. (country mean) −51.5% (−67.1% to −28.7%) −35.0% (−53.5% to −9.3%) −21.3% (−34.2% to −6.0%) −11.1% (−16.9% to −4.9%)
Mixed democ. (short-term change) −8.3% (−14.7% to −1.3%) −10.6% (−19.1% to −1.3%) −7.1% (−11.6% to −2.4%) −3.2% (−4.8% to −1.6%)
Competitive (country mean) X regime −0.1% (−1.2% to 1.1%) 0.1% (−1.2% to 1.4%) −0.1% (−0.7% to 0.5%) −0.1% (−0.3% to 0.2%)
tenure
Majoritarian (country mean) X regime −0.1% (−1.2% to 1.0%) −0.2% (−1.4% to 1.1%) −0.5% (−1.1% to 0.1%) −0.3% (−0.5% to −0.1%)
tenure
prop. (country mean) X regime tenure 0.2% (−0.6% to 1.0%) 0.0% (−0.9% to 1.0%) −0.7% (−1.1% to −0.2%) −0.4% (−0.5% to −0.2%)
Mixed (country mean) X regime tenure −0.4% (−1.5% to 0.6%) −0.5% (−1.7% to 0.7%) −0.9% (−1.5% to −0.3%) −0.5% (−0.6% to −0.3%)
% variance explained (intercept) 35.7% 35.6% 22.4% 21.4%
% variance explained (slope) 45.7% 32.0% 55.1% 65.1%
Countriesb 176 176 176 176
Country-yearsb 1125 1125 1125 1125

a
All models control for year of data collection, log population size, landlocked countries, and world region.
b
Sample size is reduced after restricting observations to the years 1979, 1984, 1989, 1994, 1999, 2004, and 2009 (the last year prior to the respective quinquennial reporting period
for mortality) and removing occasions in which mortality statistics are otherwise missing (East and West Germany prior to 1990).

with proportional electoral rules (either purely proportional or in
combination with majoritarian rules) have superior health outcomes
compared to strictly majoritarian democracies. Health advantages in
majoritarian democracies were not robust to the inclusion of GDP
(Tables 3 and 4) or to multilevel tests of longitudinal change in
population health (Tables 5 and 6). The interpretation then that free
and fair democratic elections have something to do with stronger
population health cannot be applied as easily to those that operate via
majoritarian electoral rules.
4.1. Do competitive autocracies improve upon the strictest type of
autocratic regime?
Levitsky and Way (2010) characterize ‘competitive authoritarianism’
as a type of political regime that has curiously strong electoral infrastruc-
tures, but where incumbents (ab)use their powers in various ways to
guarantee their victory. Those strategies violate every common sense of
what democracy signifies, for example, monopolizing news outlets or
harassing political opponents. Yet this kind of regime still stands in contrast
with autocracies that rely principally on military force to stay in power and
have little substantive interest in the outcomes of elections. Because
opposition parties have at least a chance of winning an election, the
characteristics of ‘free’ and ‘fair’ elections, albeit minimally so, are preserved
in competitive autocracies: ex ante uncertainty (ultimate uncertainty as to
who will win office after votes are counted), ex post irreversibility (the
likelihood that the incumbent party will honor the election results if it
loses), and repeatability (the likelihood that elections will recur before too
long) (Alvarez et al., 1996). The implication is that these types of regimes
provide at least a modicum of accountability for the ruling elite as well as a
forum through which policy issues can be discussed openly. Indeed,
competitive autocracies emulated the population health of democracies in
many of the results presented in this study. The same was true according to
supplementary analysis with respect to food provision and economic
growth.
4.2. Do democracies with proportional electoral rules have the best
population health?
A large literature has debated the implications of proportional and
majoritarian electoral rules. Even recent studies suggest variously that
policies in majoritarian systems have a conservative bias (Döring and
Manow, 2017), that they are more reliably congruent with the preferences
of the median voter than in proportional systems (Stadelmann et al., 2016),
and that this congruence changes between the two types of regimes
according to timing within the electoral cycle (Wlezien and Soroka,
2015). Proportional democracies invest more in social welfare (Persson
and Tabellini, 2004, 2005), but if this were a mechanism linking electoral
rules to better health then it is unclear why net Gini should fail to be a
mediator. Meanwhile the combination of majoritarian and proportional
electoral rules in mixed systems may help to balance the different
democratic priorities of accountability and representation (Shugart and
Wattenberg, 2001), although results comparing these with purely propor-
tional systems do not make this clear.
Nevertheless, findings support the view that not all democracies are
built the same. In many of the tests the democratic advantage in health
appeared smaller for majoritarian systems, which assume a ‘winner-takes-
all’ model for designating leadership and include countries like the United
Kingdom, the United States, and Canada, than for mixed and proportional
democracies. Differences between the three different types of democracies
were non-significant in many of the models, but it is worth noting that in
most tests majoritarian democracies trailed the other two types.
All of this being said, perhaps the most trustworthy results are those
coming from the multilevel models during the last phase of analysis (Tables
5 and 6). These uniquely distinguish longitudinal from cross-sectional
effects, thus removing the confounding influence of starting advantages that
the different countries have at the time that a new regime comes to power
while still addressing challenges like autocorrelation. These last results
indicate that advantages in population health for majoritarian democracies
are exclusively cross-sectional. Unlike for proportional and mixed systems,
comparative longitudinal gains in health eluded majoritarian democracies

97
A.C. Patterson
in every test. Such results are consistent with the view that electoral rules
supporting broad representation of multiple perspectives in the policy-
making process may protect the health of a country's citizens most
effectively. Meanwhile electoral rules that penalize leaders and parties for
taking disagreeable policy positions could still allow optimal health, but
only in combination with proportional electoral rules. More work is needed,
however, to corroborate such claims.
4.3. What explains the correspondence between electoral regimes and
health?
Consistent with findings from a previous study (Patterson and Veenstra,
2016), among the proposed mediators GDP coincides with the largest
shares of variance that electoral regime type has common with cross-
national differences in population health. Results are consistent with the
claims that democratic institutions (Gerring et al., 2005; Knutsen, 2012)
and proportional democracies especially (Knutsen, 2011) promote income
growth the best. Taken together with the argument that ‘wealthier is
healthier’ (Pritchett and Summers, 1996), there may be a case to consider
GDP a mediator. Large literatures have debated these claims, however. For
instance growth spurts often correspond, curiously, with worse health
(Ruhm, 2000, 2003). Lipset's (1959) seminal study of the cultural origins of
democracy argues that countries have more difficulty democratizing if they
are not already wealthy. Except by controlling for GDP in some models,
economic stability is a potential confounder not considered in this study.
Another recent study (Sen et al., 2016) observes an underwhelming
correspondence between democratic institutions and medium-term income
growth, finding that democracies perform better on average only by way of
avoiding economic collapse. Although there may be a prima facie case for
seeing GDP as a potential mediator when taking all the countries together
cross-sectionally, in light of these divergent literatures such a claim can only
be speculative within the scope of this study.
Although food availability is a statistically significant predictor of
both life expectancy and infant mortality, it did not emerge as a
mediator in supplementary tests when already accounting for GDP (not
shown). Results therefore do not support the supposition that food
availability is one of the mechanisms linking a higher standard of living
to lower mortality (McKeown et al., 1972). Neither were there
convincing results to suggest that electoral regimes promote health
by way of reducing income inequality.
On the other hand, investment in public health could be a mediator
for some age groups. When measured similarly in logged dollars per
capita, the inclusion of public health expenditures in models eliminated
the independent association between GDP and infant mortality for
country-years from 1995 forward. The same was true with respect to
the mortality of adults aged 65 and over. These findings support the
view that economic prosperity promoted the health of the most
vulnerable age groups – the youngest and the oldest – only by way of
providing the means to improve public health infrastructure.
4.4. Implications
More than any other policy area, this study has implications for the
rules that guide electoral institutions. Findings suggest that of all
democratic electoral systems the majoritarian one as used in United
Kingdom and North America may be the worst for health. Surprisingly,
many tests indicated that this type of electoral system performs on par
with competitive authoritarian regimes. The best system for health
might instead be one that somehow incorporates the principle of
proportional representation.
Evidently these variations play an important role in shaping the
environment in which policymakers act. A common claim is that the ability
of winning parties to consolidate control over policy in the majoritarian
system becomes a source of punctuated frustration for losing parties and
their supporting voters, which undermines the representativeness of
policies adopted by the ruling party. Such a situation arguably poses
98
Health & Place 47 (2017) 90–99
obstacles to shrewd policy choice. For example David Orentlicher (2017)
illustrates this point well in the aftermath of the United States’ 2016 federal
election. Noting how members of Congress often start with the best of
intentions to cooperate “across the aisle” but quickly become mired in
partisan conflict, he joins other scholars (e.g., Norris, 2017; Ceaser, 2017)
in making fresh recommendations for reform. Meanwhile in early 2017
Canada's Liberal Party under Prime Minister Justin Trudeau mysteriously
dropped the subject of electoral reform from their policy agenda, despite
that it played a key role in the platform they used to win office in 2015.
Given possible connections to health, both countries should (re)consider
prospects for electoral reform.
4.5. Limitations
Some important limitations must be noted. First, this study cannot
claim to offer the only or the best means for characterizing electoral
regimes. Divergent validity is especially unclear with respect to competitive
autocracies. Although preliminary results were similar using other plausible
specifications for competitive autocracies (not shown), these were not
restricted to country-years after the closure of the Cold War as Levitsky and
Way (2010) claim them to be and no easy solution could be found. Findings
may confuse different conceptions of electoral authoritarianism. The
typology used in this study is also not specific to the manifold variations
in electoral rules. Runoff elections in majoritarian democracies, for
example, or rules that allow voters to choose particular candidates (‘open
lists’) in proportional democracies are not considered in this study. Neither
are other potentially relevant dimensions of political institutions, such as
presidentialism versus parliamentarianism or relationships between legis-
lative and judicial systems.
Included are attempts to explain why exactly electoral institutions might
have bearing on health. However, only a few intervening variables were
tested, some of which touch upon long-standing controversies in the social
sciences. A literal interpretation of these results may point to GDP, but any
confident assertion that this variable is a mediator must address discre-
pancies coming from other theoretical and empirical literatures, as noted
above. In the meantime, comparison of countries according to electoral
rules was used to explore the implications of accountability and/or
representation of diverse perspectives in policy choice, the working
assumption being that different electoral regimes prioritize different values.
Other measures are available that touch upon issues of accountability and
representation, but it was not obvious which one(s) were most appropriate
within the scope of this study.
Further consideration is also needed with respect to choice of control
variables. Regime tenure offers a rough measure of political stability, while
some models control for GDP. However, a more concerted investigation of
the confounding influences of political and economic stability is needed in
light of comments from Lipset (1959) as noted above. Further to this point,
social groups and cleavages too have their own implications for both
political processes and health. Examples are studies finding correspon-
dences between mortality differentials and party of choice in the United
Kingdom (Smith and Dorling, 1996) and voter participation among Black
Americans (Rodriguez et al., 2015), the latter of which is further
complicated by combination with disenfranchisement as a consequence
of racialized criminalization. Through detailed analysis of a small sample of
European countries Mackenbach (2013) links gains in life expectancy to
democratization throughout Europe, but another study (Mackenbach et al.,
2013) finds that patterns were inconsistent across time and health outcome,
partly due to complications coming from the fall of communism in the
Eastern Bloc.
This study investigates health outcomes only at the national level.
Excluded are analyses at the local level, the global level, and across multiple
levels of governance. As do national governments, international trade pacts
arguably constrain sovereignty at subordinate levels to varying degrees.
Subnational units too respond to the environments in which they exist, with
large cities like London, Tokyo, and New York now having a global
presence.
A.C. Patterson
Through the use of multiple imputation this study addresses the
concerns of Ross (2006) about how missing data has influenced this
literature. No imputation strategy was used to address missing values
for income inequality or public health expenditures, however, and the
use of listwise deletion to address these removed 40% and 52% of the
respective observations. More work is needed to investigate multi-
variate solutions for missing data.
5. Conclusions
Although a large literature now connects political regimes to
population health, this study suggests that not all democracies (or
autocracies, for that matter) are built the same. If elections are only a
blunt instrument for holding leaders accountable to their decisions,
then electoral rules that in some way support the broad representation
of diverse interests may be more effective in promoting health. Other
findings relating to competitive autocracy suggest that even modest
gains in creating electoral infrastructures for choosing a country's
leaders correspond with better population health. More work is needed
to substantiate these findings, however. Moreover, now that some
scholars (Muntaner, 2013) are encouraging researchers to focus on
questions of why political institutions might have bearing on health,
rather than just whether they do or not, the investigation not only of
institutional arrangements but also of specific explanations for their
influence continues to be a priority.
Acknowledgments
For their feedback on this paper, the author gives special thanks to
Susan McDaniel, Gerry Veenstra, Karen Stanbridge, Thomas Brambor,
Patrick Egan, and anonymous reviewers. This project was not sup-
ported by funding or sponsorship of any kind.
Appendix A. Supporting information
Supplementary data associated with this article can be found in the
online version at doi:10.1016/j.healthplace.2017.07.003.
References
Alvarez, M., Cheibub, J.A., Limongi, F., Przeworski, A., 1996. Classifying political
regimes. Stud. Comp. Int. Dev. 31 (2), 3–36.
Álvarez-Dardet, C., Franco-Giraldo, Á., 2006. Democratisation and health after the fall of
the wall. J. Epidemiol. Commun. Health 60, 669–671.
Beck, T., Clarke, G., Groff, A., Keefer, P., Walsh, P., 2001. New tools in comparative
political economy: the database of political institutions. World Bank Econ. Rev. 15
(1), 165–176.
Besley, T., Kudamatsu, M., 2006. Health and democracy. Am. Econ. Rev. 96 (2),
313–318.
Cassani, A., 2016. Do all bad things go together? Electoral authoritarianism and the
consequences of political change short of democratisation. Politikon. http://
dx.doi.org/10.1080/02589346.2016.1248605.
Ceaser, J.W., 2017. The nomination game. J. Democracy 28 (2), 45–49.
Deaton, A., 2013. The Great Escape. Health, Wealth, and the Origins of Inequality.
Princeton University Press, Princeton, NJ.
Döring, H., Manow, P., 2017. Is proportional representation more favourable to the left?
Electoral rules and their impact on elections, parliaments and the formation of
cabinets. Br. J. Political Sci. 47 (1), 149–164.
Fairbrother, M., 2014. Two multilevel modeling techniques for analyzing comparative
longitudinal survey datasets. Political Sci. Res. Methods 2 (1), 119–140.
Franco, Á., Álvarez-Dardet, C., Ruiz, M.T., 2004. Effect of democracy on health:
ecological study. BMJ 329 (7480), 1421–1423.
Gassner, M., Ontiveros, D.U., Vincenzo, V., 2006. Human development and electoral
systems. J. Hum. Dev. 7 (1), 43–57.
Gerring, J., Bond, P.J., Barndt, W.T., Moreno, C., 2005. Democracy and economic
growth: a historical perspective. World Polit. 57 (3), 323–364.
Gerring, J., Thacker, S.C., Alfaro, R., 2012. Democracy and human development. J. Polit.
74 (1), 1–17.
Justesen, M.K., 2012. Democracy, dictatorship, and disease: political regimes and HIV/
AIDS. Eur. J. Political Econ. 28 (3), 373–389.
Knutsen, C.H., 2011. Which democracies prosper? Electoral rules, form of government
99
Health & Place 47 (2017) 90–99
and economic growth. Elect. Stud. 30 (1), 83–90.
Knutsen, C.H., 2012. Democracy and economic growth: a survey of arguments and
results. Int. Area Stud. Rev. 15 (4), 393–415.
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1999. J. Law Econ. Organ. 15
(1), 222–279.
Levitsky, S., Way, L.A., 2010. Competitive Authoritarianism: Hybrid Regimes After the
Cold War. Cambridge University Press, New York.
Lin, R.-T., Chen, Y.-M., Chien, L.-C., Chan, C.-C., 2012. Political and social determinants
of life expectancy in less developed countries: a longitudinal study. BMC Public
Health 12, 85. http://dx.doi.org/10.1186/1471-2458-12-85.
Link, B.G., Phelan, J., 1995. Social conditions as fundamental causes of disease. J. Health
Soc. Behav. 36, 80–94.
Link, B.G., Phelan, J., 2010. Social conditions as fundamental causes of health
inequalities. In: Bird, C., Conrad, P., Fremont, A., Timmermans, S. (Eds.), Handbook
of Medical Sociology6th ed.. Vanderbilt University Press, Nashville, TN, 3–17.
Lipset, S.M., 1959. Some social requisites of democracy: economic development and
political legitimacy. Am. Political Sci. Rev. 53 (1), 69–105.
Lipset, S.M., 1960. Political Man: The Social Bases of Politics. Doubleday, Garden City,
NY.
Mackenbach, J.P., 2013. Political conditions and life expectancy in Europe, 1900–2008.
Soc. Sci. Med. 82, 134–146.
Mackenbach, J.P., Hu, Y., Looman, C.W.N., 2013. Democratization and life expectancy in
Europe, 1960–2008. Soc. Sci. Med. 93, 166–175.
McGuire, J.W., 2013. Political regime and social performance. Contemp. Polit. 19 (1),
55–75.
McKeown, T., Brown, R.G., Record, R.G., 1972. An interpretation of the modern rise of
population in Europe. Popul. Stud. 26 (3), 345–382.
Muntaner, C., 2013. Democracy, authoritarianism, political conflict, and population
health: a global, comparative, and historical approach. Soc. Sci. Med. 86, 107–112.
Muntaner, C., Borrell, C., Ng, E., Chung, H., Espelt, A., Rodriguez-Sanz, M., Benach, J.,
O'Campo, P., 2011. Politics, welfare regimes, and population health: controversies
and evidence. Sociol. Health Illn. 33 (6), 946–964.
Navia, P., Zweifel, T.D., 2003. Democracy, dictatorship, and infant mortality revisited. J.
Democr. 14 (3), 90–103.
Norris, P., 2017. Why American Elections Are Flawed (And How to Fix Them). Cornell
University Press, Ithaca, NY.
Orentlicher, D., 2017. Political dysfunction and the election of Donald Trump: problems
of the U.S. constitution's presidency. Indiana Law Review 50, 247–264.
Patterson, A.C., Veenstra, G., 2016. Politics and population health: testing the impact of
electoral democracy. Health Place 40, 66–75.
Persson, T., Tabellini, G., 2004. Constitutional rules and fiscal policy outcomes. Am.
Econ. Rev. 94 (1), 25–45.
Persson, T., Tabellini, G., 2005. The Economic Effects of Constitutions. MIT Press,
Cambridge, MA.
Pickett, K.E., Wilkinson, R.G., 2015. Income inequality and health: a causal review. Soc.
Sci. Med. 128, 316–326.
Powell, G.B., 2000. Elections as Instruments of Democracy: Majoritarian and
Proportional Visions. Yale University Press, New Haven, CT.
Pritchett, L., Summers, L.H., 1996. Wealthier is healthier. J. Hum. Resour. 31 (4),
841–868.
Rodriguez, J.M., Geronimus, A.T., Bound, J., Dorling, D., 2015. Black lives matter:
Differential mortality and the racial composition of the US electorate, 1970–2004.
Soc. Sci. Med. 136, 193–199.
Ross, M., 2006. Is democracy good for the poor? Am. J. Political Sci. 50 (4), 860–874.
Ruhm, C.J., 2000. Are recessions good for your health? Q. J. Econ. 115 (2), 617–650.
Ruhm, C.J., 2003. Good times make you sick. J. Health Econ. 22 (4), 637–658.
Sen, A., 1994. Freedom and needs: an argument for the primacy of political rights. New
Repub. 210 (2/3), 31–38.
Sen, K., Pritchett, L., Kar, S., Raihan, S., 2016. Democracy versus dictatorship? The
political determinants of growth episodes. ESID Working Paper No. 70. University of
Manchester, Manchester, UK. Available at www.effective-states.org.
Shugart, M.S., Wattenberg, M.P. (Eds.), 2001. Mixed-Member Electoral Systems: The
Best of Both Worlds?. Oxford University Press, Toronto.
Smith, G.D., Dorling, D., 1996. “I'm all right, John”: voting patterns and mortality in
England and Wales, 1981–92. BMJ 313 (7072), 1573–1577.
Solt, F., 2014. The standardized world income inequality database. Working paper.
SWIID Version 5.0, October 2014.
Stadelmann, D., Portmann, M., Eichenberger, R., 2016. The influence of party affiliations
onrepresentation of voter preferences in majoritarian vs. proportional systems,
Beiträge zur Jahrestagung des Vereins für Socialpolitik 2016: Demographischer
Wandel - Session: Political Processes:Theory and Evidence, No. G22-V3. Available at
https://www.econstror.eu/handle/10419/145705.
Thornhill, R., Fincher, C.L., Aran, D., 2009. Parasites, democratization, and the
liberalization of values across contemporary countries. Biol. Rev. 84, 113–131.
Welander, A., Hampus Lyttkens, C., Nilsson, T., 2015. Globalization, democracy, and
child health in developing countries. Soc. Sci. Med. 136–137, 52–63.
Wigley, S., Akkoyunlu-Wigley, A., 2011a. The impact of regime type on health: does
redistribution explain everything? World Polit. 63 (4), 647–677.
Wigley, S., Akkoyunlu-Wigley, A., 2011b. Do electoral institutions have an impact on
population health? Public Choice 148 (3/4), 595–610.
Wlezien, C., Soroka, S.N., 2015. Electoral systems and opinion representation.
Representation 51 (3), 273–285. http://dx.doi.org/10.1080/
00344893.2015.1111928.