Loss Coverage: Why Insurance Works Better with Some Adverse Selection

I want to answer that in two parts: first from the viewpoint of insurers, and then from the viewpoint of society as a whole. The book is very much focused on the second, the social viewpoint.

From the viewpoint of insurers, adverse selection means that higher risks are more likely to buy insurance, and lower risks are less likely. To the extent that insurers don’t anticipate this in pricing, they will lose money. Obviously this is a bad thing for the insurers.

From the viewpoint of society as a whole, the argument is slightly different.  Typically, higher risks are fewer in number than lower risks. Think of life insurance: typically perhaps 80% of lives are lower risks which are offered standard prices, and 20% are higher risks which are rated or declined.  So if higher risks are more likely to buy, and lower risks less likely, that implies we end up with a lower number of lives insured than if buying patterns were uniform.  Also, because the average risk of the insured pool is higher, the average price of insurance is higher than the average risk in the whole population. This combination – a higher average price, and a lower number of lives insured – is why adverse selection is traditionally seen as a bad thing, from the viewpoint of society as a whole.

That makes sense, but your book title seems to disagree.  Why?

Step back for a moment and think about what adverse selection – “higher risks more likely to buy, and lower risks less likely” – actually means, from a social viewpoint.  It means that insurance tends to be bought by people who need it most!  From a social viewpoint, why is this bad?

The traditional answer is the one I gave above: adverse selection also means a higher average price for insurance, and a lower total number of lives insured. What this argument overlooks is the shift in coverage towards higher risks – those who need insurance most. If this shift in coverage is large enough, it can more than outweigh the fall in numbers insured, so that loss coverage is increased.

What is loss coverage?

Suppose 100 deaths are expected in a population, of which 50 are expected to be compensated by insurance. This corresponds to loss coverage of 50%.

Technically, we can define loss coverage as the expected losses compensated by insurance for the population as a whole. 

Loss coverage takes account of both the number of risks covered by insurance, and their probabilities of loss.  It captures the trade-off between covering the lives who need insurance most (the higher risks), and covering a larger total number of risks.

Why is loss coverage important?

Loss coverage may be a good metric for assessing how effective insurance is, from the viewpoint of society as a whole.  Compensation of losses is main social purpose of insurance, which public policy generally seeks to promote.  When insurance is compulsory, (eg third-party liability for drivers and employers in the UK), the purpose of compulsion is to ensure 100% loss coverage. 

Although other insurances usually aren’t compulsory, compensation of losses under other insurances still seems to be generally regarded as a good objective, and one which public policy should seek to promote.

Don’t limits on risk classification lead to a ‘death spiral’ ending in very few people being insured?

That depends on the response of higher and lower risks to changes in price – technically, the demand elasticities of higher and lower risks. In theoretical models, with plausible demand elasticities, the market does not spiral away to nothing – the spiral always stabilises.

Apart from the theoretical models, it’s hard to find convincing real world examples of so-called death spirals.  On the one hand, you have the evidence of adverse selection, which is often for fairly limited effects; and on the other hand at the rhetoric of adverse selection, which is often quite exaggerated.  I go into both the evidence and the rhetoric in some detail in the book.

How could insurers using big data affect your argument?

I think large and rapid improvements in predictive power using big data are rather unlikely.   But at the margin, it does seems plausible that advances in the use of big data may enable insurers to classify risk more accurately. Potentially, that could reduce adverse selection too much, and so reduce loss coverage. So advances in the use of big data could become a possible reason for imposing some new regulatory limits on risk classification.  And this loss coverage argument is complementary to other  arguments based on privacy or fairness, which may point in the same direction.

Loss Coverage: Why Insurance Works Better with Some Adverse Selection can be bought direct from the publishers Cambridge University Press at a 20% discount (enter code THOMAS2017 at the checkout).

About the Author: Guy Thomas is an actuary and investor, and an honorary lecturer at the University of Kent, Canterbury. His academic publications have received prizes from the Institute and Faculty of Actuaries and the International Actuarial Association. He is also the author of Free Capital: How Twelve Private Investors Made Millions in the Stock Market.

CPD: Actuaries Institute Members can claim two CPD points for every hour of reading articles on Actuaries Digital.