Heterogeneity, Risk Sharing and the Welfare Costs of Risk

How well do people share risk? Do non-market institutions – charity, progressive taxes, transfer payments – make up for the lack of complete insurance markets? Or is risk sharing far worse than what complete markets could achieve? Standard risk-sharing regressions assume that any variation in households’ risk preferences is uncorrelated with variation in income. I combine administrative and survey data with a simple model of imperfect insurance to show that this assumption fails; risk-tolerant workers sort into jobs where earnings carry more aggregate risk. The correlation makes previous risk-sharing regressions too pessimistic. I provide techniques that eliminate the bias, apply them to U.S. data, and find that the welfare losses from uninsured shocks are practically small and statistically difficult to distinguish from zero. In addition, because more risk-tolerant people bear more aggregate risk, the welfare costs of macroeconomic fluctuations are small even for arbitrarily risk-averse households. There is little room to improve households’ welfare by smoothing idiosyncratic or aggregate shocks unless smoothing shocks also raises mean output.