4 August 2020
Pradip Tapadar, Senior Lecturer in Actuarial Science at the University of Kent and Guy Thomas, Honorary lecturer in Actuarial Science at the University of Kent discuss adverse selection in insurance under COVID-19.
The early response of insurers to the COVID-19 pandemic featured a number of changes to underwriting procedures for life, health and travel insurance products. Examples include new questions about possible coronavirus symptoms, withdrawal of income protection products with short waiting periods, exclusions of COVID-related illness, and even complete withdrawal from writing new policies (particularly for travel insurance).
One justification sometimes invoked for these responses is that insurers are responding to the risk of adverse selection (or synonymously, anti-selection). Adverse selection in insurance arises from information asymmetry: it occurs when some customers have knowledge indicating that they have higher risk, but insurers do not have this knowledge, so that these customers are charged less than they otherwise would be. Alternatively, insurers may ‘not know’ in the sense that they are forbidden from using particular information in setting premiums. Examples of the latter include bans on using gender and predictive genetic tests in pricing.
However, our view is that the possibilities for adverse selection arising from COVID-19 are limited. COVID-19 gives insurers many things to worry about, but a material increase in adverse selection is not one of them. This is for two main reasons, one macro and one micro.
First, at a macro level, COVID-19 increases mortality by a common multiplicative factor of the baseline mortality rate at each age. The same is true for all socioeconomic groups, but with some evidence of a slightly lower multiplicative factor for the middle groups. The proportional nature of the risk increment across ages and socioeconomic groups constrains the possibilities for adverse selection by any large group.
Second, at a micro level, in the current state of knowledge, there is no equivalent for COVID-19 of a predictive genetic test. There is no way that an individual can privately ascertain materially higher risk well in advance of any symptoms, and so create an extended window of opportunity for adverse selection. As it happens, because of the small numbers of people with strongly predictive genetic tests, these are also not currently material to insurers; but at least at an individual level, the opportunity for adverse selection may be there. It doesn’t seem to be there for COVID-19.
There remains a possibility that some individuals might anticipate higher COVID-19 risk by reason of their occupational exposures or other lifestyle information. But this type of information will often become known to the insurer through customary underwriting processes. And where it does not, we need to remember that what the insurer is pricing is not COVID-19 risk, but all-causes risk for the particular peril insured (income protection, travel disruption, etc). Any increase in COVID-19 risk arising from adverse selection will represent a smaller increase in all-causes risk.
If, hypothetically, insurers did rate for COVID-19 occupational risk, this would essentially mean charging more to doctors, nurses and care workers. This would be an unpopular policy, and one can envisage social pressure or regulation to prevent it. Such regulation might induce a modicum of what is sometimes called “regulatory adverse selection”. Would it make insurance work less well?
We think the answer is no. Provided insurers are free to adjust the aggregate level of prices to the aggregate level of claims, some adverse selection (but not too much adverse selection) can make insurance work better for society as a whole.
Restrictions on the use of relevant information in premium rating lead to a rise in the average price of insurance, and a fall in demand. (This is regulatory adverse selection.) But they also lead to a shift in coverage towards higher risks – the ‘right’ risks, the people who need insurance most. Provided this shift in coverage is large enough, it can more than out-weigh the fall in demand.
In this scenario, despite the rise in average premiums and the fall in demand, there is a rise in what we call “loss coverage”, the fraction of society’s total losses which is compensated by insurance. From a social viewpoint, this seems a good thing.
The arithmetic of this “loss coverage” argument is sketched in the illustration, and has been expanded at greater length in our academic papers and in Guy’s book Loss Coverage: Why Insurance Works Better with Some Adverse Selection.
In summary, COVID-19 seems unlikely to precipitate a material increase in adverse selection in insurance. But if it did, that might – within limits – be a good thing.
