The effect of social capital on one's health has drawn researchers' attention. In East-Asian countries, however, such an effect has been less studied than in Western countries. Mindful of this background, this study aimed to investigate the linkage between social capital and health at the level of a small area in Japan, and also to examine whether social capital mediates the relation between income inequality and health. The main survey targeted 34,374 people aged 65 years and older from 25 communities who were without the need of nursing care. We collected 17,269 questionnaires (response rate 50.2%), from which we used 15,225 (response rate 44.3%) which had complete information on self-rated health, age, and sex. The main outcome measure was self-rated health status (1=fair/poor; 0=very good/good). The individual-level variables of age, sex, equivalised income, marital status, educational attainment, and type of housing were included to control for compositional effects. Average income, social capital and the Gini coefficient were used as community-level variables. The variable of social capital was calculated as the percentage of individuals in the community samples who reported "yes" or "it depends" to the question "Generally speaking, would you say that most people can be trusted?" When the individual-level variables and the average equivalised income at the community level were controlled for, we found that high social capital and a decreased Gini coefficient were significantly associated with good self-rated health using a multilevel model. The association between social capital and self-rated health was insignificant after adjustment for the Gini coefficient. However, in other series of models, we also detected a statistically significant linkage between the increased Gini coefficient and negative responses to the question of individual-level trust using multilevel models. This suggests that people who live in conditions of high-income inequality tend to exhibit low trust levels, and that social capital mediates the relation between income inequality and health. This study is, to our knowledge, the first investigation of the linkage between social capital and health in Japan using multilevel analysis, and reports the first supported result of the relative income hypothesis in Japan. Additionally, the statistically significant linkage between the increased Gini coefficient and low trust observed in this study provides additional evidence to support the social cohesion and collective social pathway between income inequality and health.