Should sellers of health insurance be allowed to use the buyer's age when setting prices? This paper considers competitive markets with adverse selection where firms compete in prices as in Akerlof (1970). I describe the optimal extent to which a signal of buyer type should be contractible by the sellers. A signal (e.g., age) partitions consumers into subsets (e.g., young and old). Community rating (i.e., restricting the contractibility of this signal) increases welfare if the consumer subsets where the level of cost is higher are also the subsets where there is greater deadweight loss due to adverse selection. Such signals are empirically common. There is typically an interior optimal constraint on contractibility that maximizes welfare.
To illustrate empirically the potential benefit of community rating policies, I estimate the distribution of valuations and cost for UK annuities for several groups of consumer ages and gender. The model is estimated using proprietary data that include the annuity seller's estimate of each individual's longevity.