Hurricanes, Hoarding and Replacement: Estimating the Causal Effect of Mortality on Fertility

Claus C. Pörtner, University of Washington

Despite a long-standing interest in the causal effect of child mortality on fertility there is relatively little empirical work because of two main problems. First, child mortality and fertility are affected by the same (unobservable) household variables. Second, having many children in response to mortality can itself lead to an increase in the risk of child mortality. A possible solution is to use instrumental variables estimation, but there has so far been little agreement on valid instruments. This paper uses data on the risk and occurrence of hurricanes in Guatemala to overcome this problem. While fertility is affected by the risk of hurricanes the actual shock of an hurricane is shown not to change fertility. Both hurricane shock and risk affect child mortality; the first directly and the second through changes in desired fertility. Hence, as long as information on risk is included the actual shocks are valid instruments.