To what extent does poor mental health affect employment outcomes? Answering this question involves multiple technical difficulties: two-way causality between health and work, unobservable confounding factors and measurement error in survey measures of mental health. We attempt to overcome these difficulties by combining 10 waves of high-quality panel data with an instrumental variable model that allows for individual-level fixed effects. We focus on the extensive margin of employment, and we find evidence that a one-standard-deviation decline in mental health reduces employment by 30 percentage points. Further investigations suggest that this effect is predominantly a supply rather than a demand-side response and is larger for older than young workers.