Health production models include participation in physical activity as an input. We investigate the relationship between participation in physical activity and health using a bivariate probit model. Participation is identified with an exclusion restriction on a variable reflecting sense of belonging to the community. Estimates based on data from Cycle 3.1 of the Canadian Community Health Survey indicate that participation in physical activity reduces the reported incidence of diabetes, high blood pressure, heart disease, asthma, and arthritis as well as being in fair or poor health. Increasing the intensity above the moderate level and frequency of participation in physical activity appears to have a diminishing marginal impact on adverse health outcomes. Our results provide support for guidelines about engaging in exercise regularly to achieve health benefits. Copyright © 2013 John Wiley & Sons, Ltd.