Numerous empirical studies, largely conducted in Western societies, have linked social capital to health outcomes. However, few studies thus far have been conducted in northeastern Asian countries, particularly Japan. Therefore, whether social capital is a determinant of health in Japanese society remains unclear. This study attempted to provide new evidence for the impact of social capital on health in Japan by analyzing original survey data with two different statistical models. In total, 1910 subjects were randomly selected from 210 enumeration districts (EDs) considered in the 2000 population census. In the present study, after excluding missing data on the outcome and predictor variables, we conducted an analysis of 1157 individuals nested within 206 EDs. Ecological and multilevel regression analyses were performed to examine the association between social capital and health, measured by the General Health (GH) perception item (a subcategory of Short Form 36 (SF-36)), and to estimate the impact of aggregated indicators of social capital on health. We developed an original Japanese version of three social capital items (perceived helpfulness, kindness, and greeting) based on previous studies and expert opinions and prepared a social cohesion index that integrated these three social capital items. The ecological model, after adjusting for sociodemographic factors, revealed that perceived helpfulness and greeting, along with the social cohesion index, were observed to have statistically significant associations with GH. The multilevel model, after adjusting for individual social capital perceptions, indicated that the two aggregated indicators of community social capital (kindness and greeting), along with the social cohesion index, showed a statistically significant association with GH. These results showed the existence of a contextual effect of social capital on health outcomes, although individual differences in social capital perceptions in Japan were considered.