Naïve Cynicism

Adverse selection is often described in pejorative terms which suggest that higher risks strategically game the insurance system to obtain some unwarranted benefit. Naïve cynicism is the belief that this ‘micro’ speculation about the intentions of a few individuals provides comprehensive insight into the ‘macro’ operation of an insurance system.

Naïve cynicism seems to have psychological roots. In understanding complex human systems such as insurance, the human brain seems particularly adept at focusing on isolated instances of cheating or freeriding, rather than dispassionate observation of the broad sweep of reality.Footnote ¹ This might be because isolated instances, such as one individual gaining an unwarranted benefit, are much easier to observe and comprehend than the overall performance of complex systems such as insurance.

The naïve cynic’s obsession with freeriders rather than overall outcomes treats the insurance system as a sort of morality play, in which the most salient actors are the ‘good’ low risks and ‘bad’ high risks. This perspective is not helpful because in general, economics is not a morality play. Economic arrangements which produce good ‘macro’ results may often entail tolerance of some ‘micro’ features which run against the grain of everyday notions of justice.

Actuarial Paranoia: Prioritising Policing over Epidemiology

In the past three decades, two small groups have attracted disproportionate concern in insurance circles: people with HIV infection and people with genetic predispositions to illness. In the 1980s and 1990s, actuaries became preoccupied with the perceived importance of ‘policing’ insurers’ clienteles so that such people could be surcharged or excluded. Substantial analytical effort and political capital was expended on these preoccupations, as evidenced by contemporary statements (some of which were quoted in Chapter 2), publications and the establishment of professional working parties concerned with AIDS and insurance, and genetics and insurance.Footnote ² However, neither people with HIV nor those with genetic predispositions actually turned out to be financially significant to the overall solvency of insurers in the UK.

A development which has turned out to be financially significant to the overall solvency of some insurers in the UK is the widespread reduction in mortality at higher ages, not in a specific subgroup but across the whole population of annuitants and pensioners. For the past quarter century, this development – the management of which is primarily a matter of epidemiology, not policing – has generally been underestimated by actuaries. There was no annuities working party – or rather there was, but only briefly, when it was already too late.Footnote ³

Actuarial paranoia is the tendency of actuaries to direct their efforts disproportionately towards the policing of minorities rather than the epidemiology of populations. It is the professional manifestation of naïve cynicism.

The Fallacy of Composition

Where insurers are permitted to use any risk classification methods, a new risk classification invented by one insurer can be a competitive advantage. An insurer which introduces a new risk classification – either charging less for some type of low risk or charging more for some type of high risk – can gain a least a temporary advantage over other insurers. Other insurers may then be forced to adopt the new classification. This is because if they continue to charge a price undifferentiated to the new classification, they may attract an increasing proportion of the high risks and a reducing proportion of the low risks. This is the ‘competitive adverse selection’ which was discussed in Chapter 8.

The previous paragraph describes the dynamic of a particular insurer competing against other insurers which may use different risk classification methods. But it is not the dynamic where a regulator bans certain methods of risk classification for the whole market. If a classification method is banned for the whole market, there is no possibility of an insurer being disadvantaged because rivals offer prices differentiated by that method. For adverse selection to be a problem under this scenario, there needs to be selection against the whole market (‘informational adverse selection’ in Chapter 8), not just adverse selection against a particular insurer (‘competitive adverse selection’ in Chapter 8). The latter does not necessarily imply the former. The common elision of these two concepts of adverse selection is a fallacy of composition.

One example of this confusion is contained in a paper commissioned from economic consultants Oxera by the trade association Insurance Europe. One paragraph accurately describes competitive adverse selection against a particular insurer: ‘If the insurer does not offer lower premiums to consumers who are less likely to claim, these consumers are likely to shift to an insurer who does recognise their lower likelihood of claiming and charges less for the same product.’ But in the next paragraph, this mutates without elaboration into informational adverse selection against the whole market: ‘A ban on the use of age and disability would be likely to result in serious problems arising from adverse selection for all of the insurance products considered in this study, and could lead to a breakdown of the market for certain insurance products.’Footnote ⁴

The Fallacy of the One-Shot Gambler

One justification often given for expecting adverse selection to be devastating to insurers is the idea that anyone with some private knowledge of their higher risk will wish not just to buy insurance, but to buy very large amounts of insurance. For example, in 2014 the Canadian Institute of Actuaries lobbied the Canadian senate with a research paper which assumed that 75% of people who receive any adverse genetic test result would immediately purchase convertible term life insurance providing cover of 1 million Canadian dollars. Among other reasons, the paper states that ‘They see the price as low enough to constitute a good investment which will benefit their heirs.’Footnote ⁵

But consideration of plausible probabilities and premiums suggests that in most realistic scenarios, buying large amounts of life insurance on the basis of private knowledge is probably not an attractive investment. This is for two reasons: (a) life insurance pays out only on low-probability (in gambling terminology, long odds) events, and (b) life insurance purchasers are typically ‘one-shot gamblers’ who can make their bet on these odds only once.

For example, suppose that an insurer offers me a term life insurance premium based on an assumed risk of dying of p = 4% over the term of an insurance policy (say the next 25 years), but I have private information (e.g. a genetic or other test result) which tells me my real risk is 5 times the normal level, that is p* = 20%. If I ‘invest’ in over-insurance, then my private knowledge means I pay a ‘favourable’ price (4% rather than 20%). But despite the ‘favourable’ price, on a one-shot gambler it is still very likely (80% likely, on the true probabilities) that I shall just lose the premium.

This one-shot gambler against long odds seems to me an unattractive investment proposition for a large bet. It would remain unattractive for a wide range of plausible probabilities and premiums. It may be worth buying a ‘normal’ level of cover for insurance purposes, such as protecting my dependants from destitution in the unlikely event of my early death. But the notion of massive over-insurance based on a private genetic test for investment purposes generally does not seem attractive for realistic probabilities and premiums.

For a large bet using private information to be attractive for investment purposes, I need either (a) a large ‘information edge’, defined as the difference (not the ratio) between the true probability based on my private information and the probability used by the insurer to set the premium or (b) the ability to engage in multiple independent transactions, either in sequence (repetition) or in parallel (diversification). These conditions are not satisfied by typical ‘private genetic test’ scenarios such as that described above.Footnote ⁶

Cartoons (Exaggeration for Effect)

Insurers and their advocates may present exaggerated cartoon-like accounts of adverse selection because they believe that an accurate account would be insufficiently compelling, and so fail to influence public policy in the direction they desire. This is the most straightforward form of strategic misrepresentation.

One manifestation of exaggeration for effect is that references to adverse selection are often qualified by epithets such as ‘severe’ (as in ‘severe adverse selection’), but almost never qualified by epithets such as ‘mild’, ‘moderate’, ‘slight’, ‘low’ or ‘weak’. In January 2016, a Google search on the single string {“severe adverse selection” “insurance”} returned over 5,000 hits. The five search strings {“mild | moderate | slight | low | weak adverse selection” “insurance”} returned around 100 hits in total. The overwhelmingly higher frequency of ‘severe’ compared with plausible alternative epithets suggests a rhetorical motive, rather than a calibrated assessment of exceptional circumstances.

In other cases cartoons take the form of principles without proportionality, where a technically valid principle or result is presented without relevant context which would highlight its practical insignificance. For example, when the Institute and Faculty of Actuaries acknowledged in 1999 that early actuarial models for genetics and life insurance suggested the effect of banning genetic tests would probably be to raise term life insurance premiums by much less than 10%, it immediately added that this represented ‘a material effect on premium rates for term assurances’.Footnote ⁷ It omitted to mention that typical term assurance premiums had fallen by around 25% in the preceding decade, or that the range of the cheapest few quotes at any time from different insurers for any particular term assurance risk varied by around 25%. In the context of these substantial inter-temporal and inter-company variations, an increase of much less than 10% could not reasonably be described as material.

Barricades (Defending Industry Interests)

Insurance lobbyists may believe that by exaggerating the impact of adverse selection and arguing against all risk classification restrictions, they are protecting the industry’s interests against external critics. The validity of this ‘barricade strategy’ depends on the interpretation of ‘industry interests’.

On the one hand, if ‘industry interests’ means the financial interest of shareholders, the barricade strategy may be misconceived. As noted in Chapter 3, maximising loss coverage is equivalent to maximising premium income. The models in this book assume that insurers make zero profits in equilibrium, but in practice insurers hope to earn profits. If these profits are (roughly) proportional to premiums, shareholders should welcome some restrictions on risk classification, at least to the point where loss coverage is maximised. More generally, restrictions on risk classification are generally motivated by social objectives, but as a side effect they reduce the intensity of competition in risk classification. This should tend to raise aggregate industry profits.

On the other hand, if ‘industry interests’ means specific job roles, the barricade strategy may be valid. In particular, lobbying on risk classification policy is often devolved to personnel presently employed in risk classification. They have a personal interest in maintaining elaborate systems of risk classification, which is quite distinct from the interests of insurance company shareholders. Upton Sinclair’s famous aphorism is apposite: ‘It is difficult to get a man to understand something, when his salary depends on his not understanding it.’Footnote ⁸

One example of this barricade strategy in action is a paper ‘Genetics and insurance – some social policy issues’ promoted by the Institute and Faculty of Actuaries in 2003.Footnote ⁹ The paper gives a total of eight separate arguments why insurers should be allowed to use genetic tests under the heading Insurers’ lines of reasoning. Despite the ostensible ‘social policy’ focus of the paper, there is no equivalent enumeration of arguments for anyone else’s lines of reasoning. In the published discussion of the paper, actuaries variously complained that the UK genetics moratorium was ‘big enough to drive a lorry through’ (p. 842), ‘unduly generous’ (p. 857) and that ‘once you have given that game away you will never get it back’ (p. 870). But recall from Chapter 2 that even complete bans on genetics, more comprehensive than the terms of the UK ban, have been estimated to imply an increase of less than 1% of aggregate premiums.Footnote ¹⁰ In the light of this likely insignificance, many of the actuaries’ comments seem overblown.

Signals (Beliefs as Attire)

Insurance requires durable institutions, and this requirement implies a generally conservative culture. This is not a criticism: it is an observation of a necessary characteristic of institutions which are designed to reliably fulfil very long-term contracts. However, an unfortunate corollary of this conservatism is that for an actuary, it is only right to be right when other actuaries are also right. Being right early, when other actuaries are not yet right, is not just uncomfortable at the time; it also carries no credit when one’s earlier rightness is subsequently revealed.

Belief as attire is not principally concerned with making correct predictions; it is concerned with signalling that you are a ‘sound’ actuary who holds ‘responsible’ views. Belief as attire is particularly prevalent in contexts where no penalty arises for being wrong. This is broadly true of insurance insiders who express exaggerated beliefs about adverse selection. Exaggerated beliefs about adverse selection may not do much harm to a profit-making insurer, except to the limited extent that they distract attention from more realistic concerns. Exaggerated beliefs do cause real harm when they influence public policy, but industry personnel promulgating exaggerated beliefs have no accountability for public policy.

The trajectories of actuarial concern about HIV testing and genetic testing in the context of insurance illustrate the phenomenon of beliefs as attire. As recounted in Chapter 2, the predictions typically made by actuaries on these topics in the 1980s and 1990s have turned out to be substantially wrong. But for many actuaries, these predictions were probably made or endorsed largely as attire. There was no subsequent discredit for actuaries who made or endorsed grossly wrong predictions, and no credit for those who made more accurate ones.

Semantic Sophistry

In an official monograph On risk classification, The American Academy of Actuaries notes that public policy sometimes imposes limits on risk classification, and then opines that:

Is this the best the American Academy of Actuaries can do? One might as well defend apartheid in South Africa or the Nuremberg laws in Germany by saying that they did not place people into groups, rather their skin tones or grandparental origins were placed into groups.

One-Way Hash Arguments

To a person who has never given much thought to insurance, the concept of adverse selection may not be immediately salient. But most people of reasonable intelligence – including those with no affinity for mathematics beyond basic numeracy – can quickly grasp the concept when it is explained to them. In this sense, adverse selection stories are pitched at an optimal level of complexity: easy enough for almost anyone to grasp, but clever enough for them to feel a little insightful and astute on account of their understanding.

Reality is more nuanced and harder to explain. As we saw in Chapter 8, tests for adverse selection show evidence of adverse selection in a few insurance markets, but not in many others. And as we saw in Chapters 3–6, a modest degree of adverse selection is beneficial, in the sense that it increases loss coverage, but too much adverse selection reduces loss coverage.

The orthodox presentation of adverse selection as a universally negative phenomenon is thus an example of a one-way hash argument. A one-way hash is an argument which is flawed or incomplete, but nevertheless intuitive, plausible and appealing to non-experts (especially naïve cynics). Understanding the limitations and half-truths inherent in the one-way hash argument requires a deeper level of understanding, a level which the non-expert in a hurry may not be able to attain.

One-way hash arguments are common in lay discussions of complex phenomena which attract political controversy, such as climate change, evolution, environmentalism or adverse selection. The term ‘one-way hash’ is a metaphor from public-key cryptography, where the one-way hash which encrypts a message relies on the property that given two prime numbers, it is easy to multiply them together to find their product (convince a layman by an oversimplified argument); but given only their product, it is hard to find the two prime numbers (explain to a layman why the oversimplified argument is wrong).

Presumption of Privilege

Commentators on risk classification typically operate under an implicit presumption that they are members of a privileged class who will not be in any way disadvantaged or threatened by risk classification. Analogous presumptions are usually made by analysts working on the technical details of new risk classifications. These presumptions are generally correct – or at least can be made correct – because the advocates and architects of innovation can always disregard or defer any results or proposals which might be disadvantageous or threatening to their own privileged class.

One example of the deferral of socially unwelcome new results was the slowness of insurers in the 1960s and 1970s to charge more to smokers for life insurance. The health hazards of smoking had been recognised in the medical literature at least since the 1950s,Footnote ¹² but were not recognised in life insurance pricing by UK insurers until the early 1980s. This was possibly because many senior insurance company executives were themselves smokers. I suspect that the delay arose not so much from a reluctance to financially penalise smoking, but rather because it would have been embarrassing to deprecate smoking, at a time when smoking remained common among the most senior insurance personnel.

Devaluation of the Disadvantaged

Commentators who feel secure in the presumption of their own privilege are often offensive to the less privileged. In particular, commentary on adverse selection is often offensive to people who face higher risks. A few commentators are probably deliberately offensive, but many may not realise that they are being offensive.

In general terms, to give offence is to imply that a person or group has, or should have, low status. When discussing risk classification, insurers and their advocates do this in three main ways. First, they talk about people who face higher risks as non-persons, almost always referring to them collectively rather than individuals. Second, they treat people who face higher risks as objects or problems to be discussed, while excluding them from participation in that discussion. Third, they make clear by the nature of their comments that people who face higher risks are also excluded from the intended audience for their remarks.

One example of such exclusion is the routine assertion by actuarial associations that restrictions on risk classification make insurance more expensive ‘for everyone’. Some examples of such claims were quoted in Chapter 2. But the mechanism by which the average price of insurance rises is that coverage shifts towards higher risks, who pay less than they did before. The assertion that restrictions on risk classification make insurance more expensive ‘for everyone’ makes no sense, until one discerns that for many actuaries the ontological concept of ‘everyone’ excludes minorities with higher risks.

Another example of offensiveness towards people who face higher risks, or those who are concerned for them, is the assertion that the fears of these parties concerning insurance discrimination are ‘irrational’. In a position statement about genetics and insurance, the Institute and Faculty of Actuaries said (emphasis added):

Terms such as ‘rational’ and ‘irrational’ are properly used only when discussing cognitive algorithms as algorithms. It is usually imprecise – and in the case of ‘irrational’, gratuitously offensive – to use these terms to describe a person’s concerns.

Affectation of Virtue

Another rhetorical device often used by advocates of risk classification is an affectation of moral virtue on the part of the commentator. Faced with the problems of genetic predispositions to illness, or actual illness, a natural empathetic response is a desire to help, for example by restricting risk classification. The essence of actuarial orthodoxy is that this response is shortsighted: the actuary sternly warns that any attempt to help will lead to a decline in insurance coverage and eventually the collapse of the insurance system. Essentially, actuarial orthodoxy demands immediate pain (discrimination against the ill) for supposed long-term gain (the survival of the insurance system).

Although the implied necessity is almost always overstated, the narrative of immediate pain for long-term gain is appealing to many commentators. Advocacy of immediate pain to secure long-term gain seems to carry a connotation of hard-headed moral virtue. But notice that in general, the commentators are privileged by their affluence and health, while the pain they advocate is to be imposed on disadvantaged others. This makes the connotation of virtue illusory: there is no moral virtue in advocating that pain be imposed on disadvantaged others for the benefit of one’s own privileged class.

Emo-Phobia

Affectations of virtue are often associated with a complementary rhetorical affectation which I call emo-phobia. This occurs when commentators speaking about risk classification preface their advocacy with a reference to public sentiment or sympathy, followed either literally or structurally by some form of ‘but …’ Some examples are given below:

Emo-phobia generally indicates that the speaker wishes to argue against a position which has both popular appeal and moral rightness (e.g. not inflicting further harm on persons already disadvantaged by genetic predispositions), in order to advance some unpopular and selfish interest (e.g. marginal commercial gain for insurance companies). Recognising the weakness of this position, the speaker seeks to disparage popularity and moral rightness as ‘emotional’ properties.

The appropriate response to an emo-phobe is to say yes, this is an emotive subject, and therefore one in which we should take account of moral imperatives, rather than the narrow interests which you selfishly seek to advance. Good judgement requires not the flattening of emotions, but rather the right emotions: those which are accurately calibrated to the contours of truth.