Robert Putnam showed that a social capital index, created as a weighted sum of 14 variables chosen to describe the civic degree of sociability and community mindedness, is correlated with many community outcomes, such as education, child well-being, crime, and the total mortality rate. Although correlation does not establish causation, we can find that in a large number of studies this index, a selection of its elements, or similar measures register as significantly correlated with health variables, virtually always in a direction consistent with the hypothesis that social capital improves health. The potential benefit of this relationship is substantial, especially if it proves to be robust to differences in time and place, statistical contexts, and ultimately if the relation can be supported to be causal. This paper subjects the social capital and health hypothesis to an expanded set of rigorous tests, which, by surviving, it becomes stronger or, by failing, its weaknesses are better revealed. The paper seeks to extend this body of research by a combination of study characteristics that are each relatively unusual in social capital and health research. Though causality cannot be established by these tests, the work shows that the association of social capital with health is quite robust when challenged in the following ways: (1) seven different health measures are studied, including five mortality rates; (2) the 48 contiguous states are observed at six points in time covering the years from 1978 to 1998 over four year intervals, thus forming a panel; (3) the multivariate tests feature economic variables from the production of health literature; and (4) a statistical method (instrumental variables) is applied to account for the possibility that omitted variables are confounding the social capital estimates. The results and the discussion find cases for which the social capital and health hypothesis performs only weakly, but, on the whole, the hypothesis is remarkably robust to these variations.