The relationships between democratic experience, adult health, and cause-specific mortality in 170 countries between 1980 and 2016: an observational analysis

Abstract

Background: Previous analyses of democracy and population health have focused on broad measures, such as life expectancy at birth and child and infant mortality, and have shown some contradictory results. We used a panel of data spanning 170 countries to assess the association between democracy and cause-specific mortality and explore the pathways connecting democratic rule to health gains.

Methods: We extracted cause-specific mortality and HIV-free life expectancy estimates from the Global Burden of Diseases, Injuries, and Risk Factors Study 2016 and information on regime type from the Varieties of Democracy project. These data cover 170 countries and 46 years. From the Financing Global Health database, we extracted gross domestic product (GDP) per capita, also covering 46 years, and Development Assistance for Health estimates starting from 1990 and domestic health spending estimates starting from 1995. We used a diverse set of empirical methods-synthetic control, within-country variance decomposition, structural equation models, and fixed-effects regression-which together provide a robust analysis of the association between democratisation and population health.

Findings: HIV-free life expectancy at age 15 years improved significantly during the study period (1970-2015) in countries after they transitioned to democracy, on average by 3% after 10 years. Democratic experience explains 22·27% of the variance in mortality within a country from cardiovascular diseases, 16·53% for tuberculosis, and 17·78% for transport injuries, and a smaller percentage for other diseases included in the study. For cardiovascular diseases, transport injuries, cancers, cirrhosis, and other non-communicable diseases, democratic experience explains more of the variation in mortality than GDP. Over the past 20 years, the average country's increase in democratic experience had direct and indirect effects on reducing mortality from cardiovascular disease (-9·64%, 95% CI -6·38 to -12·90), other non-communicable diseases (-9·14%, -4·26 to -14·02), and tuberculosis (-8·93%, -2·08 to -15·77). Increases in a country's democratic experience were not correlated with GDP per capita between 1995 and 2015 (ρ=-0·1036; p=0·1826), but were correlated with declines in mortality from cardiovascular disease (ρ=-0·3873; p<0·0001) and increases in government health spending (ρ=0·4002; p<0·0001). Removal of free and fair elections from the democratic experience variable resulted in loss of association with age-standardised mortality from non-communicable diseases and injuries.

Interpretation: When enforced by free and fair elections, democracies are more likely than autocracies to lead to health gains for causes of mortality (eg, cardiovascular diseases and transport injuries) that have not been heavily targeted by foreign aid and require health-care delivery infrastructure. International health agencies and donors might increasingly need to consider the implications of regime type in their efforts to maximise health gains, particularly in the context of ageing populations and the growing burden of non-communicable diseases.

Funding: Bloomberg Philanthropies and the Bill & Melinda Gates Foundation.

Figure 1

The average effect of democratic transition on HIV-free life expectancy at age 15 years (A) The average effect on HIV-free life expectancy due to a democratic transition over 15 years. The blue line represents the normalised life expectancy in the 15 countries that underwent a democratic transition. The red line represents the normalised life expectancy in the corresponding synthetic countries that did not undergo a democratic transition. Year 0 is when the transition took place. Divergence after year 0 reflects the estimated average effect of democratic transitions. Life expectancy normalised to 1 in year 0. (B) The results of an empirical permutation test to determine if countries randomly might experience this improvement in HIV-free life expectancy. A low probability indicates that this improvement in life expectancy is unlikely to have happened by chance. (C) Democratising countries used in the synthetic control analysis are marked in blue, with year of transition noted. Continuously autocratic countries are in navy and countries that previously democratised or reverted back to autocracy are in light blue.

Figure 2

Changes in mortality due to democracy and other determinants of health The proportion of variance in Global Burden of Disease Level 2 cause-specific, age-standardised mortality explained by democracy and other determinants of health such as gross domestic product (GDP) per capita, urbanicity, development assistance for health (DAH), and mortality shocks such as war. The sum of the variance explained by each variable is the r², which is the share of the variance of mortality explained by the model. NCDs=non-communicable diseases. LRIs=lower respiratory and other common infectious diseases. CDs=communicable diseases. NTDs=neglected tropical diseases. *Includes self-harm and interpersonal violence. †Includes substance use disorders. ‡Includes diabetes, urogenital, blood, and endocrine diseases. §Includes forces of nature, conflict, terrorism, and state violence.

Figure 3

Changes in cardiovascular disease across time The proportion of explained variance in age-standardised cardiovascular disease death rates across 188 countries from 1995 to 2015. DAH=development assistance for health. GDP=gross domestic product.

Figure 4

Long-term effect of democracy on country disease burden The estimated direct and indirect long-term effects of democracy on health from a structural equation model. The blue bars show the direct effect of democratic change on changes in health. The green, red, and yellow bars show the indirect effects of democratic change on changes in health due to resulting changes in gross domestic product (GDP) per capita, urbanicity, and government health expenditure as source. All changes are from 1995 to 2015. Results that were not statistically significant are shown in grey. HIV/AIDS and forces were omitted because of distortion of the x-axis of the graph, such that the other causes were not visible; all results are available in the appendix. NCDs=non-communicable diseases. LRIs=lower respiratory and other common infectious diseases. CDs=communicable diseases. NTDs=neglected tropical diseases. *Includes self-harm and interpersonal violence. †Includes substance use disorders. ‡Includes diabetes, urogenital, blood, and endocrine diseases.

Figure 5

Cross-country changes in democracy, cardiovascular disease, and health spending The relationships between changes in democracy and cardiovascular age-standardised death rates (A); changes in democracy and growth in gross domestic product (GDP) per person (B); changes in democracy and government health expenditure (GHES) as source (C). All panels show changes from 1995 to 2015. The red line is a linear fit through the bivariate relationship.

Figure 6

Critical component of democracy for four causes The average effect of democracy on non-communicable disease, cardiovascular disease, tuberculosis, and transport injury age-standardised death rates using the leave-one-out strategy. The effect is measured as the percent change in death rates associated with a one-unit increase in democratic experience. The black lines indicate the 99% CI on the effect size. Blue bars indicate significant effects on health, whereas grey bars are insignificant (significance determined at p=0·01).