There is a persistent association between health and portfolio choice, but hardly anything is known about the underlying sources of heterogeneity: what makes healthier individuals hold more risky assets? This paper uses rich Dutch longitudinal data to take into account and explain unobserved heterogeneity in theassociation between health and portfolio choice. We show that the association largely reflects unobserved heterogeneity, which is driven partly by behavioural variables. Yet even when adding an extensive set of behavioural variables including risk aversion, stock aversion, loss aversion, time preferences, and mental accounting, the association between health and portfolio choice does not completely vanish.