The future of insurance was the topic of a two day workshop I attended at the University of Bologna last week. The workshop brought together insurance researchers of many backgrounds from around Europe. It is part of a five year EU research project into the changes happening in insurance because of data and analytics. My role was as the discussant to two academic papers that examined in very different ways the issue of fairness in relation to digital insurance.
What I’m going to do in today’s post is set out the key points from my discussant speech. I won’t go into the detail of the two papers (they’re not published yet), but I will look at key themes I drew from them and what these mean for the future of insurance.
The first paper examined three well known frameworks for delivering algorithmic fairness in the context of calculating insurance premiums. Given that insurance pricing already has its own actuarial fairness criteria built into it, the way in which algorithmic fairness was controlled for had to dovetail with it. It’s complicated, but workable.
The second paper looked at the different ways in which trust in insurance is conveyed. Trust in the past was often conveyed through architecture (big solid buildings), while today that solidity is often conveyed through data and analytics. Insurance start-ups appear as rather rebellious, seeking trust through a narrative of being modern and clever. Yet it’s a narrative that can at times trip them up, creating social media storms about the fairness of what is being done with those clever and powerful algorithms.
The Scope and Depth of Fairness
What these two papers convey is the scope and depth of the fairness narrative around insurance. The algorithms themselves need to be fair – that’s a necessary step which data scientists are working on. At the same time, the overall purpose of which those algorithms are part also needs to be fair. And this is what makes insurance so unique, because in some circumstances, those two levels of fairness can act together, while in other circumstances, they can act in opposition.
The overall narrative for fairness in insurance is therefore wide and complex. Eight years ago, underwriters would talk of their premiums being fair because they were being personalised for each customer. This was ‘fairness of merit’, where higher risk paid higher premiums and lower risks lower premiums.
It is a view that has come to dominate discussions and perceptions about the direction in which insurance was heading. Even the UK regulator signed up to it (see their 2020 Sector Views report), despite also acknowledging it would cause consumer detriment.
We know however that people have started to question that fairness of merit narrative, in large part because of the downsides that it has been generating. Concerns that the direction in which insurance has been developing could leave increasing segments of society ‘uninsurable’ has begun to take shape. Only last week, the Financial Times carried a prominent article about just that problem, based upon joint research by a consumer group, Fair by Design, and the Institute and Faculty of Actuaries.
That research looked at the ‘poverty premium’, being the amount which low-income families paid for essential services just because they were on low-incomes. What I think caught the FT’s eye was the finding that insurance contributed more than any other service to that poverty premium.
More than One Fairness
Research like that builds upon earlier work on vulnerability and access to insurance. Consumers groups are now watching the market carefully, looking at how it is dealing with these issues. Some have a hopeful eye, seeking to influence the market. Others have a critical eye, seeing some form of challenge as inevitable.
One very significant output of this was the recent ban on the personal lines market’s principal pricing methodology. Lifetime value modelling was deemed unfair by the regulator, in very large part because of a successful challenge by a consumer group. Using both data and politics, Citizens Advice brought about what I think has been the biggest change in underwriting practices since Test Achats.
What this highlights are two things. Firstly, the singularity with which the fairness of merit narrative favoured by insurers has dominated thinking about the future of insurance – that is now over. And secondly, the fairness narrative emerging in its place is being shaped not just by insurers and their regulator, but by a wider range of what academics call actors, of which consumers groups have so far been most influential.
Insurers and regulators are beginning to realise that fairness of merit will have to co-exist alongside two other fairness dimensions – fairness of need (drawing on vulnerability work) and fairness of access (drawing on access to insurance work).
Fairness is Complex and Intuitive
Back in 2013, the then CEO of the UK regulator talked in a speech about “the dominant theme for 21st century financial services fast turning out to be a complicated question of fairness.” And since then, I think the regulator has struggled to take on board that complexity.
I agree that fairness can be complex, but at the same time, it is something we all have to grasp early on in childhood, especially if there are siblings around. Parents teach it to children and not long after, children remind parents to respect it too. What this means is that the public has a greater sense of fairness than the insurance community often gives them credit for. Sure, it’s complex, but it’s also important to get right, and to be seen to get right.
Time Matters
One aspect of fairness that I think modern insurance thinking lags behind that of the public is around what I call the fairness of time. The public know that things cannot always be fair all the time, but that over time, it often ends up as swings and roundabouts. A bit of unfairness today is likely to turn into a bit of fairness tomorrow (and vice versa). Yet the direction of insurance is very much towards assessing fairness of merit in increasingly smaller time slots.
What this means, I think, is that the fairness narrative for insurance needs to accommodate a fourth dimension, the fairness of time, in order to reflect public sentiments about fairness. I realise of course that many in the market will question the need for this, given how long fairness of merit has been at the heart of actuarial thinking. Yet it’s a question that itself is being redefined in the context of the influence that data and analytics are having on the market. After all, can the market really rely on the singularity of old traditions to scope the weighing up of new innovations?
Remember from earlier how I talked about how, in the real world, people think of fairness over time. We recognise that we sometimes loose a bit now, but gain a bit later. And we accept this, even if that ‘gain a bit later’ may not always be certain. We experience fairness not just individually but collectively as well. We put into something, with some expectation (but far from certain) that we’ll get something out at some point in time, yet in the knowledge that collectively, we’re stronger for doing so.
The Lost Twin
What we have here is something that I believe needs to be positioned as a parallel element to the fairness of merit – a bit like a ‘lost twin’. I refer to it as ‘the fairness of crowds’ (more here) and believe that it will become a key element of how fairness in insurance comes to be weighed up. And weighed up by whom? All those ‘actors’ who will influence the future of insurance.
I believe that the future of insurance is not a debate about algorithms, nor will it be a debate about themes like personalisation and hyper-personalisation. The future of insurance will be defined by how equality in fairness is achieved – the bringing together and negotiation of the various components to the public’s sense of fairness.
The author of one of the papers I was discussant to made the very important point that personalisation was a social trend, not a technological one. In other words, it’s not going to go away, but equally, it will evolve and be shaped by a range of ‘actors’. What insurers need to be prepared for then, is that while personalisation will be a big feature of the future of insurance, the way in which personalisation is realised through the data and analytics that underpin it, is very much a thing for negotiation.
If then the primacy of fairness of merit will end, the market needs to collectively start preparing for engaging with the process that will shape the new equality of fairness. Sure, it will be a complex process, but better to start on it now, rather than await the challenges that inaction will trigger.
