There is a substantial debate over the direction of the causal relation between income and health. This is important for our understanding of the health production process and for the policy debate over improving healthcare. We instrument income with rainfall measurements by matching satellite information on timing and positioning of 21 rainfall stations to longitudinal data (1991–94) of over 4,000 individuals in 51 villages in Tanzania. A 10% increase in income reduces the number of illnesses by 0.02. We also find that a 10% increase in income implies an increase of about 0.1 vaccinations of children under six.