This paper examines whether mental health has a causal impact on physical health. Allowing for individual-level fixed effects, our study aims to overcome some of the main technical difficulties encountered in the literature: two-way causality, unobservable confounders, and measurement error. For identification we implement an instrumental variable approach that exploits the recent death of a close friend as a source of exogenous variation in mental health. Using large-scale panel data from Australia, we find robust evidence that mental health has a positive, causal effect on physical health.