Levels of social capital can change after a natural disaster; thus far, no study has examined how changes in social capital affect the mental health of disaster victims. In this study, we examined how predisaster social capital and its changes after a disaster were associated with the onset of mental disorders. In October 2013, we mailed a questionnaire to participants in the Japan Gerontological Evaluation Study living in Mifune, a town in Kumamoto Prefecture, Japan, and measured predisaster social capital. In April 2016, the Kumamoto earthquake struck the region. Three years after the baseline survey, postdisaster social capital and symptoms of mental disorders were measured using the Screening Questionnaire for Disaster Mental Health (n = 828). Multiple Poisson regression indicated that a 1-standard-deviation change in predisaster social cohesion at the community level reduced the risk of depression among women (relative risk = 0.44, 95% confidence interval: 0.24, 0.78); a decline in social capital after the disaster elevated the risk (relative risk = 2.44, 95% confidence interval: 1.33, 4.47). In contrast to social cohesion, high levels of social participation at the community level were positively associated with the risk of depression among women. Policy-makers should pay attention to sex differences and types of social capital when leveraging social capital for recovery from disasters. © The Author(s) 2020. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health.