The research has been published by the Financial Services Consumer Panel (FSCP)...
“...an independent statutory body, set up to represent the interests of consumers in the development of policy for the regulation of financial services. We work to advise and challenge the FCA from the earliest stages of its policy development to ensure they take into account the consumer interest.”
This is a body the FCA is obliged to engage with. So what sort of response is the FSCP’s research likely to elicit from the FCA? Are insurers about to see a shift in regulatory strategy around this issue? Let’s explore what the FSCP’s report has to say, and how well it stands up.
The Elusive Link
At the heart of the FSCP’s report is the researchers’ view on what sort of causational link exists between the AI systems and personal data being used by insurers and the bias outcomes that some consumers are said to be experiencing. The report is cagey about this, describing that link as elusive.
Hold on though. They engaged with experts like myself on the issues involved, and supplemented this with desk top research into academic and other such sources. Yet they don’t seem to have engaged directly with insurers. This seems strange, for where else would the data lie to establish that causational link if not with those handling it?
Would a direct engagement by the FSCP researchers with insurers have produced the necessary data though? That’s far from certain. I understand that Citizens Advice have yet to receive any such data from insurers or the ABI. So the reason for evidence about the causational link being elusive seems to be down to insurers keeping it firmly under wraps.
More About Power
What we have here then is an expression of power by the sector. Their narrative will of course be that it makes no sense for them to disclose data that might bring their firm or sector into a bad light. And with bias, that would be a very bad light indeed. Instead, they would seem to be doing two things: saying 'we are good people and so wouldn’t do what you’re worried about', and saying ‘it is our data and we will do what we like with it’.
What we end up with then are two consumers groups (Citizens Advice in the UK, the Center on Race, Inequality and the Law in the US) producing the data and analysis to evidence that causational link, but nothing from insurers or the regulator. My question to the FSCP then is: have you really given due weighing to that evidence from those two consumer groups? It is hardly ‘elusive’, when both methodology and findings have been published.
The FSCP is in danger of placing itself between a rock and a hard place. On the one hand they have insurers using their power to ‘not play ball’ with the data they have. And on the other hand, they have consumer groups using their power to collect and analyse micro-outcome data. Again, elusive is just the wrong way of seeing the situation. It’s more about what evidence you decide to give weight to.
This stand off could of course be undone by the equalities regulator asking insurers to evidence how they are upholding the sector’s exemptions within equalities legislation. As the FSCP’s researchers know, there are practices in the market that point to the conditions upon which those exemptions are based not being taken seriously. Claims optimisation is the most obvious one.
I’m no expert in equalities legislation, but it strikes me that the Equalities and Human Right Commission must have some authority over ensuring that those exemptions are being respected. Will they exercise it? If not, then why not?
Does the FCA do Bias?
Then we have the FCA itself. One reason for consumer groups being at the forefront of investigating bias is because the regulator seems to have done so very little. Their current business plan makes no mention of bias or discrimination. For sure, they’ve got a lot on their plate, but it looks like bias is not a priority for them.
This is despite their then interim CEO telling the Treasury Committee in 2019, in a session on discriminatory pricing in insurance, that it had the expertise and resources to get inside insurers’ pricing models. That’s all very well, but if you then choose not to use them, the boast seems more than a little hollow.
The FCA have famously said that ‘they do not do ethics’. The consequences of this became apparent when the pricing super-complaint was made in 2018. The FCA had had data confirming the loyalty penalty for nearly three years, but not seen it as important enough to do anything about.
If the regulator’s leadership and culture choose not to address social issues like fairness and bias, they should be transparent about this. Instead, their response is centred around the consumer duty resolving such issues. The problem with the consumer duty is that it will take too long to have an impact on issues like bias that could well be systemic.
For this reason, I think the FSCP asked the wrong question. Instead of asking the FCA to look into bias, they should have asked why they haven’t been looking into bias before now. Is it a matter of how they’re interpreting their remit? Is it because they have data science expertise, but not social science expertise? Or is it because they would prefer to do other things? What changed their mind after that Treasury Committee session? The FCA needs to have such questions put to them.
A Debate on Fairness
The FSCP researchers clearly recognise that the whole question of discriminatory pricing in insurance sits within a wider context. This is the debate about personalisation and pooling, and the question of how to resolve the tensions it can create around fairness.
What seems odd to me is that their report quotes several articles of mine, but not the detailed paper I wrote for the Institute and Faculty of Actuaries and which was published in January 2023. It examined fairness in some detail and put forward a framework for allowing that debate about fairness and personalised insurance to get started.
What I emphasised in my paper is that it is not a debate that the regulator should control, but it is one that they can help facilitate. For example, they could provide a forum within which people can come together and discuss fairness, without fear of being accused of being in breach of competition law (more here).
Reports like this add weight to three developments that insurers need to prepare for.
The first development will be a questioning of the FCA CEO by the Treasury Committee about why they’ve done so little to address bias. The CEO will undoubtedly point to the consumer duty and the Committee will undoubtedly be unconvinced by that. Pressure on the FCA to address bias will then open up on yet another front.
The second development will be in the US. If the State Farm case is given class action status (more here), then lawyers on both sides of the Atlantic will tell their insurer clients to batten down the hatches. And this will then spell the end for resolving the problem of bias in insurers’ digital decision systems through some form of debate.
The third development will be the appointment of Citizens Advice as the consumer advocate for insurance (more here). That will give them powers to force insurers to disclose data.
To Sum Up
The FSCP’s report could only ever have been addressed to regulators. Within that focus, it raised some pertinent questions but at the same time, failed to grasp several key aspects of the ‘bias and insurance’ situation. For that reason, it will move the debate forward, but not by much.
In describing the causational link as elusive, the researchers signal their preference for regulatory or sector sourced evidence. Research by consumer groups appears to have been given a lower weighing. Is this right?
And the notional of that link being ‘elusive’ is also questionable. It’s less about it not being found. It’s more that it’s down to a lack of disclosure by insurers and a lack of regulator enquiry. That’s a different thing.
The Financial Services Consumer Panel now has the opportunity to be influenced by their research but not be held bound by it. It needs to be more challenging of the FCA’s position, moving their questions beyond the ‘could you’ sort and into the ‘why haven’t you’ sort.