The Two Unknowns Forming Around the Ethnicity Penalty
The Known Unknown
The known ‘unknown’ is claims cost. To what extent are the pricing differences found in the Citizens Advice report down to variations in claims costs? And the natural follow-on question from this is of course why that should be so.
Is such claims data available? Certainly each insurer holds it. Equally, it can be found in the databases managed by the Motor Insurers Bureau (MIB) on behalf of its member insurers. The Claims and Underwriting Exchange is one such database. There will of course be differences but the overlaps with what is needed will be significant.
Then there’s the regulator. In 2019, the FCA was told in no uncertain terms by the Treasury Committee to have insurers provide it with pricing data relating to the issue of discrimination in personal lines pricing. Presumably this was obtained. If it wasn’t, I wouldn’t like to be in the shoes of the FCA’s CEO!
Added to this would be the data obtained by the regulator as part of its investigation into, and subsequent monitoring of, issues around what became known as the loyalty penalty. There would be no use monitoring compliance with the price walking ban without at the same time monitoring the claims that will in some way be influencing the value that consumers are experiencing.
Willingness
So the claims data is there, held by insurers, the MIB and the regulator. What is less clear is the willingness of any of these three parties to share it to the requisite level of detail needed for analysis. Will one of them make a stand? An insurer perhaps? Why not Aviva, whose CEO has been chair at some point of the ABI, the CII and the IFR, CEO of two other leading UK insurers and is involved in a number of Government bodies. If there’s a will, there’s a way, as the saying goes.
I’ve heard talk of issues around competition law. While not a lawyer, I’m not convinced it’s an obstacle that can’t be worked round. After all, if the data flowing into the ABI, the IFR, the MIB and the FCA is not a problem, then why should this be?
As another saying goes: ‘If not me, who? If not now, when?’
The Elephant
The elephant that seems to be in the room around the sector’s response to the ethnicity report is application fraud. It accounts for around 80% of all fraud cases, the remainder being claims fraud. Levels of application fraud in motor insurance are significant: 732,087 cases in 2019, more than double what it was two years previous.
Yet dig a little deeper and a curious angle emerges. Virtually all of that increase in motor insurance application fraud comes from a near tripling in ‘suspected’ application fraud between 2017 and 2019. You can find all the numbers in more detail here.
My point in raising them is this. You can’t look at pricing outputs within also looking at the inputs into pricing of both claims and counter fraud.
Still Short of an Answer
Ok, let’s say that by the end of the year, all of the claims data and all of the fraud data has been made available. The data scientists have sharpened their digital pencils, all ready to start analysing it. The obvious question that remains is the absence of data about protected characteristics. Sure, given insurers’ exemptions under equalities legislation, some of that protected characteristic data will be available, but it will certainly still be gaps.
How does the sector, how does Citizens Advice, get round this? That I believe is where the Alan Turing Institute (ATI) comes in. As the UK’s national institute for data science and artificial intelligence, it also happens to be associated with many of the best researchers on data ethics. This means that the whole question is bias in data and algorithms is bread and butter to them.
It has struck me on a number of occasions that the ATI is likely to have some form of synthetic database that could, to a simple mind like mine, be put through a decision system to establish the level of discrimination embedded in it. And given the seven years or so that the ATI has been working with the FCA, the reason for doing so will be pretty familiar to them.
A FAIR Answer?
Consider their Project FAIR, also known as a “framework for responsible adoption of artificial intelligence in the financial services industry”. One of its anticipated outputs is “an actionable framework for safe and trustworthy deployment of AI in financial services, underpinned by foundational research methodologies formulated by cross-disciplinary and cross-sectoral teams”.
The ATI does have a tendency, like the FCA, to work more closely with big players in consultancy and the finance sector, and less closely with consumer groups and the like. Putting that aside however, it strikes me that their Project FAIR should have some contribution to make to testing the reproducibility of the ethnicity penalty research.
One Final Word
I’ll end with one final thought. Citizens Advice are asking insurers to share their claims data. I can’t imagine insurers not then asking Citizens Advice to share their research data. Anonymity is of course vital for the many thousands of people who provided Citizens advice with data about their motor premiums. I’m sure that is something that their data science team can address.
Summing Up
Data underpinned the research that lead to the ethnicity report. Data should underpin how the present situation is moved on. In the absence of such data, the likelihood is that politics will take its place. And should that happen, the issue doesn’t evolve into an ‘if’ – it remains a ‘when’, just much more difficult.