The researchers surveyed three of the four Italian insurers offering motor telematics policies, as well as data scientists at software providers. Three findings from this research stood out for me.
Firstly, their finding that the individualisation of underwriting seemed to be progressing at a far slower pace than current sector narratives would imply. In short, the digitalisation of insurance is less widespread and having less impact than the “often very emphatic narrative” would have us believe.
Secondly, telematics has been more successful at addressing miscalculations within actuarial assessments than at rewarding policyholders through behavioural assessments. In other words, the actuarial way still sets the pace and structure.
Thirdly, the way in which insurance has traditionally acted as a facilitator of mercantile and social risk taking could now be in jeopardy. So what are the public policy implications of this, and what sort of pressures might this create for the market?
Let’s explore each of these in turn and consider their implications.
Tweaking within Segments
In wide ranging interviews, the researchers, Prof. Alberto Cevolini and Prof. Elena Esposito, found that the Italian insurers were sorting drivers into three segments based upon traditional actuarial thinking. Insurers were then using behavioural assessments to understand why some drivers in any one of those segments displayed lower than average estimated probabilities, and why some drivers higher than average estimated probabilities. Insurers were then using those behavioural understandings to reward or penalise members of the same segment more selectively based upon their actual behaviour.
Two thoughts struck me about this. Firstly, I recall the period many years ago when motor insurers here in the UK went from around seven vehicle segments to forty segments and then into the many dozens of segments. And by many years ago, I mean before the internet. So why are these 21stcentury telematics insurers working with just three driver segments? If you’re collecting a river of data from the telematics app used by policyholders, why underwrite it through just “advanced, normal and reckless” driver segments. It didn’t make sense to me.
And secondly, I wonder whether the three insurers were being fully open with the researchers. Some ‘competitive advantage’ thinking may have made them reluctant to fully open up about their activities, and may also have stopped the fourth and biggest telematics insurer cooperating with the researchers. There was little for them to gain and much more for them to lose.
So while I acknowledge that there is a fair amount of myth creation when it comes to talk of individualised and predictive underwriting, could there also be a gap here in sector narratives, between that for non-market audiences and that for internal and market consumption?
Actuarial Still Rules
An insurance executive made this interesting point to the researchers...
“...the great advantage of behavioural tarification is that it provides an objective, rational and structural way to industrialise discounts that maximise customer retention.” (my underlining)
So while behavioural data from telematics may be telling insurers lots of things, these insurers are still only using it to vary discount levels in relation to actuarially based premiums. That’s a very subsidiary use of data that is expensive to collect and expensive to analyse.
The reason for this may lie in how telematics policies were first offered to Italian policyholders with an immediate and automatic flat premium discount. This would then have driven lots of policyholders with middle to high level premiums to adopt the new type of policy. What emerges then are discount driven insurers selling to consumers conditioned by distribution channels offerings to focus on price.
There’s also strong support for good driving to result in rewards. These can range from “little gadgets to cashback when filling the tank, vouchers for buying trips and so on.” Clearly this acts only one way – rewards can’t be returned for bad driving. Yet it is also an approach that is orientated around the insurer retaining premium income.
What emerges then is a tension between insurers using behavioural data to lower losses, while wanting to avoid it lowering premium income. The mixed message it sends will not be lost on many policyholders.
Insurance Innovation and Risk Taking
“Historically speaking, insurance was introduced not to induce individuals to keep their exposure to risks under control, but to relieve individuals of their worries about possible future damages.”
“The traditional purpose of insurance... has never been that of reducing risks, but it could be said to have been more that of multiplying them, guaranteeing the possibility of managing their consequences.”
These two quotes from the researchers’ paper raise a critical question for those seeking to shape the future course of insurance. If the traditional insurance mechanism for managing the consequences of risks (pooling) were to be diminished by the digital transformation that the sector is undergoing, what does this mean for our social and economic futures? If society then takes less risk, or some risks not at all, how will this impact our capacity to develop as individuals, communities or even nations? Could our development trajectory start to flatten out, perhaps even go into some form of reverse?
This has sometimes been described as the ‘cold hand of insurance’. We reduce our range and depth of activity because insurance is felt to hold us back, to limit our scope for action. It’s a narrative that I am hearing more frequently now and is one that insurers need to listen and respond to, not necessarily individually but certainly collectively.
This research into Italian telematics highlights some important lines of thinking amongst insurers. Yet at the same time, it also highlights where more research is needed.
First and foremost, it highlights the urgent need to hear from policyholders. Research into any form of behavioural assessment needs to hear both sector and consumer voices. While the former will tell us what insurers want to happen, the latter tells us what was experienced, and the consequences that then resulted.
There are two vital points about which we need to hear from consumers. The first relates to the oft quoted view that they are not interested in carrying the insurance losses of their accident prone neighbours. They prefer to pay their own way. In my view, such attitudes gain prominence because the narrative around them is based upon just fairness of merit, and just on the policyholder being a good risk.
What is needed, and urgently, is insight into what policyholders that are poor risks think about who bears whose risks. After all, if I don’t want to bear other people losses, then de facto, they wouldn’t want to bear my losses. And in that scenario, insurance morphs from providing economic stability to providing economic volatility.
The second point about which we need to hear from consumers relates to the pivotal role that price plays. Just how much does price actually dictate decisions, and why? We know that in both the UK and Italian motor insurance markets, price comparison websites (PCWs) have a big share of distribution. And what PCWs have done more than anything else is emphasise price.
Is that because PCWs found price so easy to communicate? Or was it because that is what consumers wanted to know? Clearly, if over the last 15 years or so, consumers have been fed a price orientated message from the sector, then that is what they will respond to. It’s feels like a classic case of Pavlovian conditioning. The problem is that such conditioning forces issues like fairness to the background, whereupon the sector then thinks it is irrelevant. Cue problems like the pricing super complaint and the poverty premium here in the UK.
We therefore need to overcome two problems: not knowing enough about what consumers think, and getting underneath the conditioning that could be influencing how consumers react to market innovations.
Temporality and Objectivity
From such research will emerge views on two key themes in relation to behavioural insurance. The first is the question of time. All too often research is orientated around the sector’s interest in ever decreasing time frames: a day’s worth of behavioural data for example. Yet in doing so, research tends then to lose sight of how fairness is influenced by time (more here). Just because the insurer collects data on a daily or even hourly basis, doesn’t mean that they can’t then ‘see that data’ over an annual or multi-year time frame. We don’t judge fairness in everyday life just on an hourly basis, so why are we increasingly doing so in insurance?
Another way in which the sector’s approach to behavioural data influences the sector’s engagement with fairness comes from its perception of such data as objective. Recall that earlier quote...
“the great advantage of behavioural tarification is that it provides an objective, rational and structural way to industrialise discounts that maximise customer retention.” (my underlining)
We need to understand more about the objectivity or otherwise of behavioural data. To me (and I’m far from being alone in this), behavioural data is not objective. Subjectivity is built into what data you collect, how you judge its significance, how you then treat it and interpret it, and what you then link it to in terms of actionable next steps. It’s a mass of micro-proxies utilised according to very human decision making.
The more the sector sticks with its belief that what it does in relation to behavioural data is objective, the less attention it will give to the fairness of what it is doing. After all, if I think what I am doing is objective, then there would be no point in listening to subjective ethical things like fairness. That way of thinking needs to be explored and some clarity introduced.
To Sum Up
The way in which Italian motor insurers have adopted and utilised telematics is very revealing. They’re using behavioural data to refine actuarial assessments. Discounts and premium retention are guiding principles. Only recently have they been thinking more pro-actively: “This is our plan for 2021 in terms of products: coaching”. As early adopters, have Italian insurers then been slow to exploit all this behavioural data? Or is the narrative around behavioural fairness still up in the clouds?
As any good piece of research should do, there are important pointers here to the next steps to be taken. Listening now to consumer voices; examining the impact that price conditioning has on behavioural feedback, and looking at outcomes through a more varied temporal lens: these will uncover the next layer of insight that policymakers and the sector need to be consider.