Insurance Risk Classification: How much is socially optimal?

Restrictions on insurance risk classification can lead to troublesome adverse selection. A simple version of the usual argument is as follows. If insurers cannot charge risk-differentiated premiums, more insurance is bought by higher risks and less insurance is bought by lower risks. This raises the equilibrium pooled price of insurance above a population-weighted average of true risk premiums. Also, since the number of higher risks is usually smaller than the number of lower risks, the total number of risks insured usually falls. This combination of a rise in price and fall in demand is usually portrayed as a bad outcome, both for insurers and for society.

However, some restrictions on insurance risk classification are common in practice. For example, since 2012 insurers in the European Union has been barred from using gender in underwriting; and many countries have placed some limits on insurers' use of genetic test results. We can observe that policy-makers often appear to perceive some merit in such restrictions. This observation motivates a careful re-examination of the usual adverse selection argument.

In this talk, we study the implications of insurers not being allowed to use risk-differentiated premiums. First, we provide a micro-foundation in variations across individuals' utility of wealth to obtain an aggregate insurance demand function. Then, within this framework, we formulate the concept of loss coverage, defined as expected losses compensated by insurance, as a metric for evaluating different insurance risk classification schemes. Finally, we reconcile loss coverage to a utilitarian concept of social welfare, defined as the sum of individuals' expected utilities over the entire population.

Specifically, we show that if insurance premiums are small relative to wealth, maximising loss coverage maximises social welfare. From a policy perspective, this may be a useful result because maximising loss coverage does not require knowledge of individuals' (unobservable) utility functions; loss coverage is based solely on observable quantities.