We develop a method to estimate domain-specific risk. We apply the method to sickness insurance by fitting a utility function at the individual level, using European survey data on life satisfaction. Three results stand out. First, relative risk aversion increases with income. Second, marginal utility is higher in the sick state conditional on income, due to an observed fixed cost of sickness. Third, the domain-specificity of risk shifts the focus on the smoothing of utility, not consumption. The optimal policy rule implies that the replacement rates should be non-linear and decrease with income.