With so much variety of risk and variety of policyholder, the way in which underwriters approach their setting of premiums has huge ethical implications. It’s often summarised as the ethics of categorisation, but I prefer to look at it along its two principal dimensions: pooling and differentiation.
Insurance has always involved the pooling of risk, be it by quantum (big or small risks) or by quality (low or high risks). The rationale behind pooling is that enables losses to be shared amongst broadly similar policyholders. It’s so fundamental to insurance that without it, insurance could be said to no longer exist. Yet pooling can face challenges, in particular from another fundamental aspect of insurance, that everyone participating in an insurance pool should pay a premium that reflects the likely losses from the risk they present. Too big a pool increases the spread of risk quality within it, which in turn results in some lower risk policyholders paying more than a proportionate share because of the higher risk policyholders in the same pool. That seems unfair.
On the other hand, too small a pool means fewer policyholders and less sharing. Reduce the sharing and risks are priced closer and closer to the losses calculated for each individual policyholder, whereupon it takes on the form not so much of insurance, but of a savings plan. This is fine so long as you present only a small risk or have no claims. Everyone else’s premium will rocket and some will be priced out of the market or unable to even access cover. That also seems unfair.
Thus a key ethical question is where to find a balance between pooling and personalisation of risk. This is not a question of science or technology, but of judgement. I’ve gone into this in greater depth in a recent article for the CII’s Journal, from which I’d like to draw one particular point. Decisions around what is or is not fair need to take into account all members of society, all of whom are of equal worth and should have an equal opportunity to access something like insurance. At the same time, it’s fair that on-going access to insurance should also take some account of merit, so that someone who continually claims on the pool could find themselves excluded from it altogether. Only governments have the capacity to guarantee unconditional access to certain risk pools, such as the state pension provision in the UK’s National Insurance system.
Just as risk pooling can be seen as a financial mechanism for spreading risk, it can also be seen as a social mechanism for sharing risk. This sharing mechanism helps bring the benefits of insurance nearer to those who need it. It also helps make it more accessible to all members of society. So some risk pooling is fair, but not if it is taken too far, for otherwise that ‘merit’ dimension of fairness suffers.
Let’s move on to that other ethical dimension of how underwriters set premiums: differentiation. This involves looking at how particular attributes of risk are categorised and given relative weight within rating structures. Underwriters look for certain attributes of a person or a property that appear to correlate with certain claim outcomes. So thatched roofs would be allocated to a separate risk category from properties with tiled roofs. Unfortunately, underwriters have not always been able to (or have shown less enthusiasm and energy to) gather risk information that allows more direct correlations between risk attribute and likely claim outcome to be used in rating. Instead they’ve relied on imperfect correlations (of the ‘it will do’ type) or proxies that are convenient substitutes. Proxies by nature are imperfect, so some policyholders may have that proxy attribute, but not that likely claim outcome. It’s unfair that they have to pay a higher premium as a result.
Society accepts that insurance must involve some differentiation of risk, but only when there’s a reasonably strong correlation between risk attribute and claim outcome. When that correlation is inadequate or its proxy outdated, that differentiation can turn into discrimination. The most recent example of this is gender, and the most serious example is what became known in the United States as ‘red-lining’.
Underwriting can be a highly complex activity, relying on increasing inputs from science and technology. Even so, it is still a human activity, full of judgements, approximations and sensitivities. Categorising policyholders through the level of risk pooling, and differentiating between attributes, will remain a human activity, irrespective of the level of technology introduced. As such, it will always be open to ethical, and unethical, decisions being made. Insurers need to have a clear picture of where those judgements, approximations and sensitivities play more than a nominal role in the underwriting of a line of business and make sure that a suitable level of critical and informed oversight is present.