We measure income uncertainty at the level of U.S. states, and the extent to which it has been reduced through risksharing, using a method recently developed by Athanasoulis and van Wincoop (2000). Risk is measured as the standard deviation of state-specific income growth uncertainty, measured by using the error term of a regression of income growth on variables in the information set. Risk sharing is measured by the extent to which this standard deviation has been reduced through financial markets and federal fiscal policy. The advantage of this measure over the existing risk sharing literature is that the interpretation does not depend on many auxiliary assumptions. Our findings on the extent of risk sharing are insensitive to the only assumption we need to make, the variables that are in the information set. We find that the standard deviation of state-specific income growth uncertainty is reduced by less than half through financial markets and federal fiscal policy. We show that the extent of risk sharing would be much higher if agents held better diversified portfolios across the states.

A method is constructed for decomposing the variance of changes in incomes in the world into components, to indicate the most important risk-sharing opportunities among people of the world. A constant absolute risk premium (CARP) model, an intertemporal general-equilibrium model of the world, is presented to permit optimal contract design. For a contract designer maximizing a social welfare function, the optimal contracts maximize the equilibrium world real interest rate. Securities are defined in terms of eigenvectors of a transformed variance matrix. The method is applied using Penn World Table data on the G-7 countries, 1950–1992.