Risk Variables Are On Their Way Out

The core cover provided by a motor policy has not really changed in the last 50 or more years. And the nature of the risks encountered by a policyholder whilst driving haven’t radically changed over that period either. So why are motor insurers now using a thousand or more variables to price such policies? What’s changed?

It’s a question worth asking, for risk has shaped the narrative around insurance pricing for a great many years. Yet it is a narrative becoming increasingly out of kilter with people’s experience of insurance. And that disparity (between what people experience and what the sector talks about) matters, for it causes awkward questions to arise in policy making circles.

In essence, and speaking as if I was a policy maker, why am I hearing this from the public and that from the sector? I can’t see how they fit together. Whose narrative do I find more convincing? Is it really pricing sophistication, or could it be unfair use of data?

The Big Switch

At the moment, the old sway that sector lobbyists had over policy maker mindsets is slipping. I believe part of that loss of sway is down to the downgrading of risk as the key variable in insurance. The public understand a lot about risk in round terms. It influences their views on the fairness of what’s on offer from the sector (more here). If the sector detaches from it, and concentrates more on a myriad of statistical correlations, then the public will struggle to understand how fair it all is.

The switch that seems to have happened is that insurers had views on risk and backed these up and measured them with statistical calculations. So, risk and then correlation. The trend that now seems to have taken hold in the sector is the reverse of this. The insurer looks first at statistical correlations and clusters, and then labels them as risk.

The vast increase in available data and the sophistication of analytics has facilitated this. And the outcome is motor insurers with a thousand plus rating factors. My favourite is whether you drink tap or bottled water as a rating factor in motor.

Yet those 1000+ rating factors are not 1000+ types of risk. They are 1000+ correlations that have been found to influence outcomes and from that, the portfolio. They’re then labelled as risk and to this is then applied a postscript narrative around such correlations being indicators of character, and character being an indicator of moral hazard, and moral hazard being an indicator of insurability.

Loosing More than Risk

The problem though is that by this time, the sector has lost both the public and policy makers. The latter could understand labels like occupation being a proxy for a greater or lesser risk: banker compared with bar keeper, for instance. The sector loses them on tap or bottled water, on how you shop, on what words you type into social media, and so on.

What the pricing review here in the UK recently established is that character is of interest to insurers not just in relation to moral hazard, but increasingly in relation to pricing opportunity. You shop that way so it means you’ll carry this level of price rise compared with that level of price rise. What such thinking falls very short on is an appreciation of the sophistication of a lot of human decision making. So for example, many people don’t drink bottled water for of economic reasons, but for environmental and social reasons. Such motivations are usually missed by digital systems, or at least mislabelled and so misinterpreted.

Risk is on its way out. In its place are factors that individually produce micro variations in outcomes, with the hope that they will then accumulate into macro variations in portfolios. Insurers will be under pressure to continually extend the scope of such factors in order to maintain a competitive advantage. So those micro variations will become even more micro, and those macro variations even harder to achieve.

In the meantime, consumers and policymakers will hear the sector’s narrative on risk and fairness, but, to be blunt, simply not believe it. Once a sector loses those two stakeholders, it faces an uncertain legislative future. The EU AI Act could well be the bellwether of this.