Planetary-scale events, such as climate change, generate heterogeneous impacts across individuals. If risks are shared completely, aggregate risk is a sufficient statistic for the global welfare effects of such events. However, if risk sharing is incomplete, then the global welfare consequences of such events depend on the distribution of impacts. We show that when transaction costs scale with distance, the degree of risk sharing depends on the spatial correlation of economic shocks. An economy surrounded by similarly affected economies incurs trade costs to share risks with distant, anti-correlated trading partners. We test this idea by examining the risk-sharing capacity of the global food market during exogenous planetary-scale climatic events experienced between 1960-2010. The El Nino Southern Oscillation (ENSO) periodically organizes local temperature and rainfall shocks of opposing sign into two contiguous regions of the planet such that under an El Nino event, cereal output is lower in the tropics and higher in the mid-latitudes while the opposite occurs during a La Ni˜na event. Consistent with incomplete risk sharing, we find that El Ni˜no drives cereal exports from more distant locations towards the tropics but trade is insufficient to prevent price increases there. Using historically estimated ENSO effects, we calibrate a spatial model of global risk sharing and simulate counterfactual scenarios that compare welfare losses due to an increasingly spatially correlated climate with welfare gains from declining trade costs. Joint with Jonathan I. Dingel and Solomon M. Hsiang.