What to do about Discriminatory Pricing? Algorithmic Bias, Consumer Outcomes and the Future of Regulation
In March 2022, we published research that found that people of colour spend on average £250 more than white people on car insurance. We used a combination of research methods to get under the hood of car insurance and build a robust evidence base. Our results are clear, but there isn’t a simple solution to this complex problem.
Many factors are likely to drive these unequal outcomes: from individualised pricing in insurance to algorithmic bias and structural racism. However, the multifaceted nature of the issue shouldn’t prevent action from being taken. Instead, insurers must recognise that - in line with the Equality Act 2010 and, as we highlight, the FCA’s new Consumer Duty- they have a responsibility to explain and justify any unequal outcomes or risk illegal indirect discrimination. Since publishing our report we have been working closely with stakeholders from the Financial Conduct Authority (FCA) and the Centre for Data Ethics and Innovation (CDEI) to develop a practical solution to this complex problem.
For insurers to take serious steps towards providing an adequate explanation of unequal outcomes, they must go beyond platitudes, assertions or hypotheses and build a robust data driven understanding of the problem. This is no mean feat. It will require grappling with challenging questions around data collection and protection, algorithmic decision making and fair value for consumers. However, this is work that all firms will need to undertake in line with the FCA’s new Consumer Duty. The FCA’s guidance outlines both the requirement to monitor outcomes for “customers with characteristics of vulnerability or customers who share protected characteristics” and provides innovative suggestions for collecting the data firms might need to do so.
The Consumer Duty does not replace the requirements of the Equality Act 2010 but rather strengthens firms’ responsibilities towards customers with protected characteristics. Whether we continue to find evidence of discriminatory pricing will depend on the success or failure of how the Consumer Duty is implemented.
Our policy proposals below outline a framework that industry and regulators should use to address discriminatory pricing and an outline of how this approach interacts with the requirements of the Consumer Duty. There are different approaches firms could take to monitoring for discriminatory outcomes. Firms will first need to decide between collecting ethnicity data from customers or using proxies (like postcodes) to understand consumer outcomes. They will also need to choose whether to ensure algorithmic models are fully explainable or to concentrate on carefully measuring outcomes like claims rates and costs to serve and checking they explain any differences in premiums charged. It is vital that, whichever specific mechanisms are used, we secure a market in which people of colour are not charged more for their insurance.