SOCIAL DETERMINANTS OF HEALTH: DO THE DATA SUPPORT THE RHETORIC?

The WHO Commission on the Social Determinants of Health was set up to collate evidence on the determinants of population health, and in its final report called for action on a long list of determinants as varied as early child development, universal health care, social protection and fair employment, public infrastructure and the natural environment, global market responsibility and governance, fair public financing and development aid, gender equity and political empowerment. While the Commission carefully presented current evidence on the individual importance of all these determinants, it was not able to give overarching guidance on the determinants that have greatest and most direct influence on population health, and differences in the impact of determinants for countries of varying income levels. Policy makers find it difficult to translate the careful work of the Commission into concrete action because it remains unclear what interventions to prioritize.

The underlying problem is the lack of a widely accepted theoretical model of population health. If it existed, such a model would provide guidance on the correct specification of an empirical model, and on the set of explanatory variables that should be chosen from a large pool of potential candidates. The objective of our paper is to obtain the optimal specification with an entirely empirical simulation based approach that bypasses the need for an underlying theory. We model life expectancy across 187 countries, and test the relative importance of a very large number of potential determinants of health -as identified by the WHO Commission- using Extreme Bound Analysis (EBA). The idea of EBA is to estimate a model repeatedly, each time with a different set of control variables. Because each regression yields coefficient estimates for the variables, all the regressions together generate a distribution for the estimated coefficients. A regressor is likely to have a significant impact on life expectancy if a sufficiently large fraction of the density function lies to one side of zero. Thus EBA allows us to narrow down the recommendations of the WHO Commission to a smaller set of social determinants for which there is empirical evidence that they influence population health.

We use the World Bank’s World Development Indicators for the years 1990 to 2012, and analyse separately countries classified into low, lower-middle, upper-middle and high-income categories. We use fixed effects panel data methods, and we address problems of missing data with multiple imputation techniques that utilize the added information provided by variables that are excluded from the overall model specification because of their high degree of correlation with the regressors. Initial results show that employment rates, in particular labor participation of women, gender equality, adolescent fertility rates, and absence of violence and war have significant impact on life expectancy in both low and high income countries. Clean water supply, gender equality, vaccination rates, schooling and overseas development assistance have significant impact in low income countries only. The number of significant determinants for life expectancy is smaller for high-income countries.