Why discrimination is an enormous ethical challenge for insurance

  • 14 January 2020

Insurers are going to be challenged on how they are managing the risk of discrimination across their business. Some insurers will be surprised and upset by this. Yet it is important that they move beyond this initial reaction, for the challenge has powerful backing. Insurers who are able to take a critical look at how they operate will take it in their stride. 

Let’s be clear on language first. Insurers often think in terms of discriminating between risks in their pricing. It would be more apt to call that differentiating between risks. Everyone else thinks of discrimination in terms of the issues covered in equalities legislation. And that is how I’ll refer to it in this blog post.

It will be natural for some insurers to be upset  by the implications that such a challenge represents. Yet surprised they shouldn’t be, for questions around discrimination have been signalled to the sector for some years now.

Plenty of Signals

I picked up on those questions back in 2015 and conducted a short survey to scope the problem – more on that later. Discrimination was one of the seven ethical risks associated with pricing that I wrote about in 2017. And the writing on the wall became even clearer in October 2018 when the Financial Conduct Authority had this to say in a ‘Dear CEO’ letter to insurers:

During the review (of household pricing), we identified the following issues regarding firms’ pricing practices that could cause significant harm and poor outcomes for consumers:

  • The risk of discriminating against consumers through using rating factors in pricing based (directly or indirectly) on data (including third party data) relating to or derived from protected characteristics.

I believe that most insurers read this and thought: ‘we would never contemplate doing such a thing so it’s not a problem for us’. That would not have been a sound move, for a few months later, the issue of discrimination in insurance pricing was debated in a parliamentary committee. The outcome was this telling off for the regulator from the Treasury Committee:

The Committee is concerned that, despite the FCA telling the Committee that a number of firms could not give it assurance straight away that their pricing data is compliant with the Equality Act, the FCA did not choose to ask for more information.

So is this just another dimension of the current pricing review? Yes and no. Yes, it has emerged from the research conducted by the FCA’s data science team into retail general insurance pricing. No, because it’s now seen as pertinent for claims, marketing and counter fraud, as well as underwriting. Insurers need to see this challenge in relation to their overall operations.

What do Insurers have to draw on?

What do insurers have to draw on then? Recent years have seen a lot of attention being given on two fronts. Firstly, diversity and inclusion issues in relation to employees, and secondly in how insurers engage with individual consumers. This is important stuff, but only part of the issue. What many insurers have not been addressing are their responsibilities under equalities legislation for consumers as a whole.

Is that last sentence rather too ‘pot half empty’? Not if you consider the findings of a short survey I conducted in 2015, when I looked at how six insurers were addressing, in policy terms, those wider responsibilities. Insurers were not referencing customers in their equality policies in anything other than a very passing manner. And by that I mean something along the lines of ‘customers benefit  from our approach to diversity and inclusion with employees’. There was also a marked absence of equality policies in the public domain. The thinking was that as they were aimed at employees, there was no need to make them more widely available.

Is this just a problem with policies, monitoring and reporting then? In other words, write a policy, gather the evidence and get on with business as usual. After all, we’re good people and discrimination is not something we’d ever contemplate. Well, I wish it was just that.

Many Good People in Insurance, but…

Of course, there are a great many good people in insurance. I worked with some of them during my 18 years in the market. However, I’ve been told of practices that, on the face of it, very much justify this challenge on discrimination. And of course, what I’ve been told may only be an isolated case, a bad apple. Two points here. Firstly, as the saying goes, even one bad apple can be bad news for the barrel (in other words, the market). And secondly, a blip could be just a blip, but it could also be a tip, underneath which is an iceberg seized problem.

What does this point to then? Is it just a problem about the poor handling of proxy data throwing up some out-of-line outcomes? And by proxy data, I mean not data directly about protected characteristics, but data that indirectly indicates (in other words, proxies for) a protected characteristic.   Unfortunately, no. There are signs of the problem being wider than just indirect discrimination. There could be direct discrimination as well

This is not an AI Problem

Perhaps then we seeing here an incident or two of the growing recognition that artificial intelligence tools like machine learning might be picking up discriminatory practices through the historic data they’ve been trained upon? Again, unfortunately no. This is about humans making decisions that, yes,  might involve the use of AI tools to put into operation, but which still remain, very firmly, judgements by people. This is not an AI problem. 

So what is happening then? The explanation I believe lies in the sub-cultures that can exist within insurers as large and often complex organisations. These sub-cultures can sometimes generate particular beliefs about what is justified and what is not justified, about what is reasonable and what is not reasonable. And this can lead to ways of thinking about ‘how things get done round here’ that drift out of line with that organisation’s purpose and values.

This field is often referred to as behavioural ethics and I wrote a series of articles about it four years ago, looking at it from an underwriting angle (starting here), and a claims angle (starting here). Read them and you’ll begin to get a picture of how good people can end up making bad decisions.

Six Functions with a Role to Play

Let’s look forward now. How should firms prepare for this challenge? Here are six functions that have a role to play in addressing this challenge:

  • The internal audit team need to take a detailed look at where the firm stands now and where it needs to do more.
  • Senior management needs to weigh up the implications that this discrimination exposure represents.
  • Digital people need to look at their data and algorithms, and assess how well discrimination is being controlled for.
  • Risk management people need to put discrimination on their risk register and assess the firm's risk exposure.
  • Oversight people need to track how accountability for tackling discrimination is being implemented across the business.
  • Learning and development people need to equip senior people with the right capabilities to handle a discrimination challenge.

And then all this needs to be weaved together into some form of discrimination prevention plan. It adds up to quite a set of tasks, but it also correlates with the scale of reputation and financial risk that discrimination represents. The time to act is now.

If you have any questions about this post, please get in touch

Related Posts 

These two posts will also be of interest...

  1. Data, Equality and Insurance: a small survey
  2. The Dangers of Unintentional Discrimination