One of the most significant changes happening in insurance at the moment is around the perspective of time. Insurers are moving from annual to daily to per minute information about the risks they’re underwriting. Policies are being designed around ‘micro-moments’. It’s exciting stuff, but not without some ethical implications.
These changes in time perspective are being fuelled by data. When your underwriting moves from being based upon proposal form information, to being based upon a vast lake of data, you know not just more about a policyholder, but more about how they and their risk are changing as well. So, for example, what my driving today says about my mental health today, in comparison with yesterday, or last year. And what this month’s visits to the gym say about my physical health, in comparison with previous visits.
This temporality of insurance raises some subtle but important ethical questions. One is around identity and the extent to which one moment is said to represent who we are over any other moment. And what length of ‘moment’ gives that representation greater meaning, compared with another length of moment. If your underwriting is premised around such identifies, is it actually on as solid a ground as you think?
Fairness and Time
Take fairness as another example. How should we weigh up the fairness of a situation, when it is likely influenced by what’s gone before (the historical context) and what will, in turn, happen in consequence of what you decide? If you hesitate over this, just think of how you weigh up the fairness of decisions you make at home. It is almost always in relation to a context, and every context has a time dimension to it.
Then think of how the principle of adverse selection is applied. The extent to which a risk is considered high or low is related to the time period over which it is considered. The shorter that time period, the greater the likelihood of that risk appearing more volatile, of being other than stable. This is because fluctuations in risk will now be falling across more than one, perhaps even hundreds of, time periods.
What this means then is that the time period over which you consider a risk influences how you see that risk. And how you see a risk influences what you think of its nature and the decisions you make as a result. Take telematics. Real time driving data streaming into an underwriting system will provide the insurer with real time opportunities to apply pricing and/or cover decisions. If an insurer hesitates to make those decisions, then the logic of adverse selection labels this as unfair on low risk drivers. Time frames decisions.
This creates a tension, between the customer who tends to judge fairness across a wide time period, and the insurer who is increasingly judging fairness across an ever narrowing time period. Perhaps this is why we’re seeing actuarial fairness being questioned more often now in academic circles.
All Data is Historical
Consider these words by Professor Terras of Edinburgh University’s Futures Institute, from an Alan Turing Institute lecture in 2019: "All data is historical data: the product of a time, place, political, economic, technical, & social climate. If you are not considering why your data exists, and other data sets don’t, you are doing data science wrong”.
What she’s saying is that a piece of data may exist in a particular time period (for example, today), but its meaning is as much determined by how it existed in previous times periods (for example, last year). Isolate that piece of data in too narrow a time period, and your data science is likely to be driving your underwriting more towards a cliff than towards a superhighway.
Customers and Proximity
Let’s look at this temporality of insurance in another way. If the insurer leans more towards the per minute rather than the per month perspective , what does that say about the relationship that insurer wants with its customers? A per minute perspective may be framed by the insurer as ‘getting closer to the customer’, but if the customer finds their premium going up because of a blip in the risk they’re presenting, they’re unlikely to warm to the insurer’s proximity approach . That proximity will have more of a transactional feel to it, rather than a relationship.
So this temporality dimension to insurance frames both how insurers see customers, and how customers see insurers. The question therefore of what temporality to orientate your underwriting around should then be answered by the type of relationship that you want with your customers. If you want your business to be more customer centric, then you need to answer that relationship question first, and have your underwriting, claims and counter-fraud then align with it.
Unfortunately, I suspect that most of today’s market thinks data first and customer relationship second. The focus has been on building relational capital upstream with data brokers and software houses, with an assumption that downstream relational capital with customers, the public and their representatives will just follow suit.
Increasing Concerns about Trust
Life is not that simple, and insurance cannot live off data alone. It needs the trust of its customers, and of the wider public too. And there are signs that the sector is now realising this. Consider what PwC’s 2020 Global CEO survey had to say: “…insurance CEOs are increasingly concerned about public trust. Without winning this trust, it will be difficult for insurers to bring their expertise to bear.”
So the challenge for those insurance CEO’s is around how firmly and confidently they’re doing to respond to those concerns. It’s fine to be concerned about trust, but leadership on ethics is about acting upon those concerns. One of the starting points then is around how this new temporality to insurance is managed. Just because you can underwriting on a real time basis, doesn’t mean that you should do. You can be a digitally driven firm, but should you be? What sort of firm do you want to be?