Behavioural fairness is a serious risk to the future of insurance

An influential narrative has emerged within insurance in recent years, centred upon the notion of behavioural fairness. It is a narrative that could determine how insurance evolves over the next decade. The problem is that it is a narrative with significant flaws.

In this post, I am going to use recent academic research into insurance to highlight inherent weaknesses in this narrative and to highlight the political risks that it will expose the sector to.

Actuarial Fairness

I’ll start with another tenet of modern insurance thinking, the notion of actuarial fairness. This has the insurer seeking to match premiums paid by policyholders as closely as possible with risk exposures. Actuarial fairness has its origins in notions of moral hazard, which sought to differentiate those policyholders who presented their risks and claims honestly, from those policyholders who used dishonest means to extract lower premiums or higher claims settlements.

In their 2017 paper ‘Enacting Actuarial Fairness in Insurance…’, Gert Meyers and Ine Van Hoyweghen chart the emergence of actuarial fairness from its conceptual origins in the early 1960s, to its position at the heart of insurance thinking in the 1980s. Building on earlier research by the legal scholar Tom Baker into how insurance’s handling of moral hazard evolved over time, the picture that emerges is one of adaptation and evolution. Concepts and principles have been made to serve the needs of the market as it evolved to meet both opportunities and threats.

Insurance has not been alone in this. Wider economic thinking has been changing too. For many years, economic thinking was centred around the rational, risk averse, utility maximising individual. It has however had to evolve to accommodate the reality that individuals don’t always make decisions that are rational. And as a result, behavioural economics has changed our understanding of how markets and their participants work.

What these two cases (insurance and markets per se) show us is that central tenets like actuarial fairness and rational individuals are less representations of external realities, and more enacted assumptions to deliver anticipated performance. Bear this in mind – it’s important for later.

Anti-Discrimination Challenges

Let’s look now at how actuarial fairness has moved from being a background assumption, to a ‘much talked about’ principle. Meyers and Van Hoyweghen find that this change happens against the backdrop of a number of anti-discrimination challenges that the sector has faced. Prominent amongst those challenges have been HIV, genetic testing and gender.

As more and more countries enacted anti-discrimination legislation, insurers positioned actuarial fairness as a core principle that would be endangered if the sector was not given exemptions to such legislation. It was presented as a reality and utilised for two purposes. Firstly, to recognise the sector’s understanding of societal concerns about discrimination. And secondly, and more importantly for the sector, to protect the insurer’s ‘right to underwrite’. That reality became a powerful playing card, put on the negotiating table to orientate perceptions around how the sector would be too exposed without it.

The result was a series of exemptions to insurers within equalities legislation. For example, in UK legislation, insurers could only use sensitive information when it was “… both relevant to the assessment of the risk to be insured and from a source on which it is reasonable to rely, and it is reasonable to do that thing.” It had to be a “proportionate means to achieving a legitimate aim”.

A Flexible Perimeter

There were exceptions to these exceptions though. Race can never be used, and from 2012, gender as well. So actuarial fairness had a perimeter, and Test Achats showed that it wasn’t a fixed one. Just as it had been developed and refined by the sector to meet anti-discrimination challenges to their right to underwrite, it could also be reshaped to reflect the direction and strengthening of societal concerns.

This societal push back on gender, Meyers and Van Hoyweghen argue, led the sector to consider its options. One option was to strengthen the sector’s resistance to further challenges to their exceptions within equalities legislation. A second option was to evolve the debate in a different direction, by developing and emphasising a new narrative based around the concept of behavioural fairness.

It is the latter option that has gained precedence, in large part because the sector was enjoying a surge in the scope and depth of data it could use in underwriting. Meyers and Van Hoyweghen’s paper references an interesting interview in which a senior reinsurance manager explains this realignment. Speaking about the interest of insurers in tracking policyholders’ behaviour…

“It comes from the background that ultimately, in the long run, insurers will be blocked from using things that people have no influence over. For example, we are not allowed to use genetic testing because you are born like that. You have no say in it. No matter what you eat, you have no influence at all on your genetics. So in the long run the regulator and consumer interest groups will probably shut down us using things that people have no control over. Now, the one thing people have clear control over, is their behaviour. And that is a very easy discussion to have with people to say: ‘You know you are doing wrong. Yes I do. Why don’t you change it? I don’t feel like it.’ Then you say: ‘You understand, I cannot actually reward you for that or accept you as a client. There is not much you can argue because all you have to do is act in a positive way and that solves the problem.’ It is something they have full control over.”

Behavioural Fairness

So behavioural fairness is creating a new narrative for modern insurance. It’s one that moves away from the technical language of how insurance worked (pooling, high risk, low risk, selection, etc.) and towards the language of personal responsibility. People could control how they behaved, and take responsibility for the outcomes that their behaviour produced for them.

Insurance people talk about this as the new and modern form of fairness. And in doing so, they’ve found that this is a narrative that could resonate with consumers. Meyers and Van Hoyweghen illustrate this through the example of a Dutch telematics based motor insurer (Fairzekering), with statements such as this from their website:

“Other peoples’ bad luck, recklessness and carelessness determine how much you pay for your insurance. Even if you never cause any damage, you end up paying for people who do. At Fairzekering, we don’t think that is fair. How badly or, more importantly, how well you drive should make a difference.”

This is fairness as a choice, specifically here in terms of driving, but more widely in terms of lifestyle. And it is a choice centred around the individual, rather than some chunky and (to the consumer) vague risk category. In contrast to the uncertain and unpredictable feedback when moving between those technical risk categories, the customer could easily get feedback on how their behaviours influence their insurance outcomes, from the choices they made while living their lives.

What results, in terms of the premium you pay, the cover you have, is a redistribution of responsibility. Your behaviour determines your premium, your cover, not the technical calculations of some back room insurance person. It sounds so obviously ‘a good thing’, yet as I mention at the outset, it is a narrative with some significant flaws.

Significant Flaws

We can see these flaws on five levels. Firstly, take the narrative so often constructed by insurers around behavioural fairness. Fairzekering’s presentation of its insurance offering is full of ‘hidden in plain sight’ assumptions. Bad luck does happen to other people, but it just as likely to happen to you too. After all, that is what bad luck means. And never causing damage is not the same as never suffering damage. All this is premised on the short term, not the long term.

And it presents a relatively binary choice. Do you want a ‘fair’ policy like ours, or do you want to carry on paying for other peoples’ claims? This is a choice that encourages you to assess your sense of identity before even thinking about your behaviour. In other words, are you stupid enough not to want such a policy? So this is a narrative constructed to deliver a ‘win win’ for such insurers, as it seeks to push those with a less confident view of their (driving and other) behaviour away from (in this case) the telematics insurer. It’s insurance for the confident and optimistic.

Still Just a Wrapper

A second level on which the narrative is flawed is a more market orientated one. It belies what is actually happening in the market. The plain truth is that telematics insurance is often little more than a wrapper around an underwriting model firmly premised on actuarial fairness. The benefit you earn from how you drive is shaped around a discount off a group rate. The default at both the policy and portfolio level is still actuarial fairness.

And every telematics type insurer knows this, for the reality is that their business model relies heavily on the reinsurance market. And that reinsurance market is underwriting with actuarial fairness at its heart. Show me an insurer willing to buy reinsurance orientated around behavioural fairness. And show me a reinsurer promoting their product with a narrative along the lines of ‘you won’t have to pay for other insurers’ bad luck, recklessness and carelessness’.

Testing the Assumptions

A third level on which the narrative is flawed is in how it ascribes, on the part of the individual, particular outcomes with particular behaviours. How realistic though is this emphasis on the individual?

Our behaviours can certainly be influenced by our own decisions, but they can most certainly be influenced as well by the decisions of others. Our behaviours do not stand in isolation. They’re the product of our own actions and the product of our interactions with others. And that applies when we’re behind the wheel of a car, as well as in a million and one other activities.

Think about it. What part of our work life, our social life, is spent in isolation? Very little. I’m not the only person driving down that motorway, walking down that street, working in that factory or office. Humans are social animals and our lives reflect this. A narrative of behavioural fairness built around the individual may resonate, but it is seeking to construct a ‘reality’ that doesn’t exist.

The fourth level on which the narrative is flawed is around the completeness with which the notion of behaviour is presented. By this, I mean: what exactly do insurers mean by behaviours?

What is Meant by Behaviour?

What an insurer views as a behaviour is dependent on what it wants to observe. And what it wants to observe is down to what it wants to measure. Certainly what it wants to measure is down to what it wants to achieve. And we know that what it wants to achieve is down to a mix, typically, of market share, combined ratio and return on investment.

My point is that it is the insurer who defines what is meant by behaviours – what they are, how to measure them and how they are interpreted. And the party to a contract like telematics insurance, who is able to define what behaviours count and to define what the outputs mean, is the party who is then also defining what is mean by fairness. It is this that has caused insurance to be labelled as a ‘political technology’, and none more so than telematics type insurance.

But surely, some of you may ask, this is something that science has made clear. After all, doesn’t science create a single, unified objective understanding of what our behaviours are, what they mean, how they can be measured? Some of science is of course so structured, but a lot of it is more like a work in progress. Indeed, the history of science is littered with ‘objective structures’ that ended up being pulled down and replaced with something else.

The danger here is that people could accept (and rely on) the science behind how certain human actions are measured, without questioning whether the underlying science is in its infancy, undergoing a ‘revolution’ or universally accepted. These different stages of development clearly present different risks to adopters like insurance.

Behaviours as Opportunities

The fifth and final level on which insurance’s behavioural fairness narrative is flawed has to do with wider market developments. Behavioural fairness is not the only development looking to overtake actuarial fairness as a core concept guiding the sector. A more prevalent one at the moment is optimisation – the determining of the premium charged, of the cover provided, of the claim settlement, by what the customer is willing to pay or willing to accept.

This is about behaviours being used not simply in relation to risk, but in relation to opportunity as well. In simple terms, optimisation goes something like this. If I buy V and W, and stop buying X, then that means I may accept paying Y more in policy premium, or I may accept receiving Z less in settlement. Fairness doesn’t get much of a look-in – it’s all about the financial optimisation of incoming premiums and outgoing settlements.

The danger for the sector is that it is using a narrative that emphasises the relationship between behaviours and fairness, while engineering underwriting and claims decisions around the relationship between behaviours and optimisation. That’s a reputationally dangerous strategy.

The Shape of the Future

Let’s move on and look at behavioural fairness as a market tool that will undoubtedly be developed over time. So I want to take where we are now (with telematics / device based insurance continuing to grow) and look at what that trend means for the future.

And this is a future that I would roughly bracket around three to seven years ahead. I know some people would challenge me on this, on the basis that this is a future that is happening now. My view however is that while it may be happening now, it is not yet happening at scale. So this is a future that might still take shape in ways that are other than as it looks now.

I’ll run through some of the assumptions about this future. Firstly, behavioural fairness becomes more deeply ingrained within insurance models, positioned itself as the dominant underwriting philosophy. Secondly, the data and analytics that underpin this development continue to expand in scope and sophistication. And thirdly, there are few obstacles outwith of the sector to these two developments.

The Morality of Inclusion

I’m going to look at this future through a lens put forward by two Swiss researchers, Michelle Loi and Markus Christian. In their paper, they have explored developments in data analytics and insurance through a conceptual framework – the ‘morality of inclusion’ – put forward in the 1990s by the philosopher Allen Buchanan.

I’ll briefly outline Buchanan’s framework (bear with me – it’s worth it). It’s perhaps best illustrated with an example or two. Buchanan recognised that some disabilities result from the interaction of natural components (such as impairments in biological function) and social components (such as the contexts in which we work, travel and go about our daily lives).

Take dyslexia. It’s a biological dysfunction that is only a disability in a society that relies on written communication. Change that social environment and our views on that natural component change too. So the way in which social institutions are configured, and the way in which they adopt and utilise technologies, influences how individuals are perceived within what Buchanan called the ‘dominant cooperative scheme’.

Schemes of Cooperation

So what is this ‘dominant cooperative scheme’? It emerges from the choices that determine the interaction of those institutional and technological factors over time. Now, everyone has an interest in being able to participate in dominant cooperative schemes, as it helps us have more productive and rewarding lives. Yet that dominant cooperative scheme requires trade-offs between what features it has, and the cost of providing them.

Consider wheelchair access. For a long time, little was done to facilitate access for wheelchair users to buildings, spaces and facilities. Now, albeit with some way still to go, the ‘dominant cooperative scheme’ here is that such access is the normal expectation, paid for by the organisation seeking to provide access to the public. Choices and trade-offs changed, and the inclusive nature of the scheme changed with them.

Let’s turn now to insurance and the future I outlined at the start of this section. It has behavioural fairness as a feature of the ‘dominant cooperative scheme’ for how policies are underwritten. And that ‘dominant cooperative scheme’ for how policies are underwritten is also relied on by other ‘schemes’, such as how houses are bought, how jobs are obtained, and other features of how we live our lives.

Conditions for Participation

Now, as part of the ‘dominant cooperative scheme’ for how policies are underwritten, prices and cover are determined by insurers’ analysis of social, financial, travel, shopping and other such life data. The conditions of inclusion in underwriters’ dominant cooperative scheme are of course, that you’re able to provide the data that insurers now rely on, that you’re happy for insurers to acquire such data, that you’re prepared to rely on the results produced by their analytics, and so on.

And because behavioural fairness relies on higher and higher risk policyholders being underwritten at higher and higher prices (and vice versa for low risk policyholders), this will result in different people having different experiences of that underwriting ‘dominant cooperative scheme’. Differences in my biological functioning will result in differences in how I engage with that underwriting scheme. An example would be around mental health. And differences in my social functioning will also result in differences in how I engage with that underwriting scheme. An example would be whether I wear a tracking device.

Schemes are not Neutral

Now these ‘engagement thoughts’ must also be seen in the context of the institutions and technologies controlling this ‘dominant cooperative scheme’. The institutions are private insurers and the technologies predominantly data and analytics. Neither are neutral.

What insurers judge to be significant, what they see as worth monitoring, and the decisions they make from this, are based on their social, philosophical, economic and political priorities. How technologies are put to use to gather, sort and interpret data are also based on similar such priorities.

This configuring of that dominant cooperative scheme by those institutions and those technologies will determine the kinds of people who can then engage with those related schemes, such as buying a house, etc. To put this in another (and more pithy) way, it enables or disables people’s engagement with key life events. This future world of insurance will therefore generate new forms of disability. And this raises all sorts of political and social justice questions. Questions that the sector has to recognise and seek to address.

And as part of addressing those political and social justice questions, insurers will have to respond to the fairly obvious question marks hanging over how they have gone about utilising the behavioural fairness that is at the heart of underwriting’s new dominant cooperative scheme. The five flaws in the narrative of behavioural fairness outlined early will just be the starting point.

Summing Up

The implications of this likely future for insurance need to be debated, in a balanced and inclusive manner. The danger is that unless the insurance sector is proactive in this debate, it will find those political and social justice questions forcing the sector into a scenario that feels more like a court than a debating chamber.

That debate also requires sector attitudes such as ‘leave insurance to the experts’ and ‘this is the new normal’ to be dropped, for they blind insurers to the moral challenges that lie ahead. Indeed, it is such challenges that have lead to insurance and the calculative techniques associated with it being labelled as ‘political technologies’.

The insurers who grasp this, and I have seen some do so, are the insurers who will ride these challenges to their advantage and succeed in the future world of insurance. The insurers who can at least start to understand this, and I hope many will do so, are the insurers who will recognise what they need to do to build enduring relationships of trust with consumers.