Traditional empirical models of retirement which use a self-assessed health measure have often found that the wage rate has a surprisingly small effect on retirement. Using both a self-reported health measure and one not based on self-report (subsequent mortality experience) in a more general reduced-form joint-demand framework, we test the importance of the interaction effects of health and retirement. Our findings are consistent with a joint determination of health and retirement. The wage elasticity using either health measure is about equal to that found in a single-equation model when a mortality measure is used, but all three elasticities are five times greater than one found using self-reported health in a traditional single-equation model.