Asset price data imply a large degree of international risk sharing, while aggregate consumption data do not. We evaluate whether a model with trade in goods and endogenously segmented asset markets accounts for this puzzling discrepancy. Active households pay a Oxed cost to transfer income into or out of assets. These households share risk within and across countries, and their marginal utility growth prices assets, so asset prices imply high international risk sharing. Inactive households consume current income and do not share risk, so aggregate consumption (which averages across all households) reaects lower risk sharing. Trade in goods is essential for generating these di§erences in the asset price-based and the consumption-based measures of risk sharing. Indeed, without trade, consumption is constrained by domestic resources and there is no international risk sharing. The calibrated model predicts risk sharing measures in line with data, and also partly resolves the Backus-Smith-Kollmann puzzle.