The contextual effects of social capital on health: a cross-national instrumental variable analysis

Past observational studies of the associations of area-level/contextual social capital with health have revealed conflicting findings. However, interpreting this rapidly growing literature is difficult because estimates using conventional regression are prone to major sources of bias including residual confounding and reverse causation. Instrumental variable (IV) analysis can reduce such bias. Using data on up to 167,344 adults in 64 nations in the European and World Values Surveys and applying IV and ordinary least squares (OLS) regression, we estimated the contextual effects of country-level social trust on individual self-rated health. We further explored whether these associations varied by gender and individual levels of trust. Using OLS regression, we found higher average country-level trust to be associated with better self-rated health in both women (beta=0.051, 95% confidence interval 0.011 to 0.091, P=0.01) and men (beta=0.038, 0.0002 to 0.077, P=0.049). IV analysis yielded qualitatively similar results, although the estimates were more than double in size (in women, using country population density and corruption as instruments: beta=0.119, 0.028 to 0.209, P=0.005; in men: beta=0.115, 0.025 to 0.204, P=0.01). The estimated health effects of raising the percentage of a country's population that trusts others by 10 percentage points were at least as large as the estimated health effects of an individual developing trust in others. These findings were robust to alternative model specifications and instruments. Conventional regression and to a lesser extent IV analysis suggested that these associations are more salient in women and in women reporting social trust. In a large cross-national study, our findings, including those using instrumental variables, support the presence of beneficial effects of higher country-level trust on self-rated health. Past findings for contextual social capital using traditional regression may have underestimated the true associations. Given the close linkages between self-rated health and all-cause mortality, the public health gains from raising social capital within countries may be large.