The challenges for insurance and regulators in a Big Data world

Speech by Andrew Bailey, Chief Executive of the FCA, delivered at the Association Of British Insurers (ABI) annual conference.

Speaker: Andrew Bailey, Chief Executive
Event: Bishopsgate, London
Delivered: 22 November 2016
Note: this is the speech as drafted and may differ from the delivered version

Highlights

  • Big Data means that insurers can move the boundary between risk assessment based on aggregate modelled behaviour and risk assessment based on the observed behaviour of the individual.
  • How the impact of Big Data affects the boundary of pooling risk characteristics and individual observation.
  • The issues the FCA is considering as part of its review of the Financial Services Compensation Scheme.

It is a great pleasure to be at the ABI Conference today. Insurance is very important to the effective functioning of our society and economy. It is therefore very important to the FCA. On a slightly selfish note, I have to admit also that it is a great pleasure to talk about insurance today and not speak about Solvency II. I will go out on a limb and say there is more to life than Solvency II, important though it is. 

So, off that leash, what am I going to talk about, is there anything left? Yes there is. To prove that, or try to at least, I want to talk today about a number of challenges to the process of pooling and pricing risk, that is insurance. I want to put it into the context of public policy and in particular our role as financial conduct regulator. And, in doing so, I want to give some explanation of the recent publication of Our Future Mission by the FCA. 

I am going to start with what is often known as Big Data. My comments draw on the FCA’s recent paper on Big Data though I don’t do justice to the full scope. 

There has been a revolution in our ability to capture and use information in our lives.

There has been a revolution in our ability to capture and use information in our lives. Insurance provides the service to manage risk in many areas of our lives. It is a fundamental building block of society – much more than just a financial service. This emphasises its importance, but also puts it squarely into the world of public policy.

Why? One important reason is to enable risk sharing and to allocate the costs of risk management, to price risk in society and to allocate risk by pooling it and thus enable diversification. The cost of risk management for all of us is lower through pooling than in an undiversified system. This much is all very well known. 

So, why re-state it?  And why does Big Data matter? On the face of it, Big Data is simply more data. It doesn’t change the fundamental laws of behaviour – specifically it doesn’t change the key assumption that we all typically behave rationally in response to news from outside. Rationality here means that we maximise our welfare. Models are then built - and of course it is arguable that the oldest users and developers of models in the world are insurers - to capture and aggregate the act of maximising welfare. Then we are left with the task of how to build and maintain systems to process all these data.

But, I think there is more to it than assuming that Big Data is just more data. I do not subscribe to the view that we no longer need the theory of aggregated rationality that underpins our models of people’s behaviour because we have access to a lot more data. 

But it does mean that insurers can move the boundary between risk assessment based on aggregate modelled behaviour and risk assessment based on the observed behaviour of the individual. It therefore takes us somewhat away from pooling. This point about pooling – just to warn you – is one that I will come back to several times in this speech. 

Let me give a first example of how the impact of Big Data affects the boundary of pooling risk characteristics and individual observation. It concerns driving. It is possible now to use telematics to collect real data on how each of us actually drives. In terms of the boundary between theory and empirics it reduces the dependence on the assumption that

I am a person with the following representative characteristics and puts more emphasis on how I actually behave. That strikes me as a good thing. It prices risk more accurately, and importantly, it should incentivise improved driving as a means to reduce the insurance premium. In contrast, when the premium is based on aggregate factors which I cannot change, my incentive to improve is cancelled out. 

Regulation should be outcomes based and incentives comparable.

This is important because I strongly believe that regulation should encourage the operation of incentives that help to deliver the outcomes that we want to see in society. Regulation should be outcomes based and incentives comparable. I would however emphasise here that, in my view, empirical analysis needs to be grounded in theory even when the empirics are much better identified.

Let’s go back to the driving example briefly. We know a lot more about my past driving, but that information still needs to be translated into a premium – my insurer still needs to position me on a distribution within the pooled set of people. So, the next step depends on pooling and that still requires modelling and theory. 

Let’s keep going with this train of thought and introduce a second example. Access to big data can yield a wide range of information. Suppose – as happens – that it allows us to identify people whose behaviour over time shows inertia in the sense that they do not shop around for a better premium. They are not prone to switching. There may be many reasons why a customer chooses to stay with a particular provider, but big data could be used to identify customers more likely to be inert, and insurers could use that information to differentiate pricing between those who shop around and those who do not. The latter pay more and thereby can cross subsidise those who do shop around. 

There is a choice for society – do we permit this sort of behaviour to go on, or not? This is the essence of financial conduct regulation – the job of the FCA. Parliament has given us an objective to ensure markets function well and, in order to achieve that, an operational objective in respect of consumer protection. We are therefore asked to exercise judgement on whether as a society we should or should not allow this type of behaviour. To simplify, our view is that we should not. Why? If you take the argument apart, it is because we think that to do so would be to exploit a feature of individual behaviour which should not be exploited in this way. The reason is either because some of the public act on the basis of ignorance or naivety (and I am not using either of these two words in a critical or pejorative context) which should not be exploited, or because some of the consumers involved are more vulnerable (in whatever way) and cannot reasonably be expected to act in a way that prevents the exploitation of the information on their behaviours. 

Let me offer a few observations on this second case. It is rooted in a judgement by the regulator. In my roles as a prudential regulator and now more as a conduct regulator, one of the common elements is that what we do involves the use of forward-looking judgement against a framework of rules. The second observation is that the framework of rules has to be anchored in public policy objectives which are shaped by intended outcomes, hence the forward looking piece. And, of course, they have to be consistent with the objectives given to us as a public authority by Parliament. 

Recently, I read a commentary piece on one of our forthcoming policy announcements which took the approach that even though we might announce superficially benign requirements for increased transparency from firms to consumers; such an approach would have an undesirable consequence for firms. If I generalise this point, commentators seek to decipher whether our judgements are more pro-firm or more pro-consumer. But, our objective is neither of these; it is rooted in public policy and the public good, which has to consider the welfare of consumers and users as well as the role of firms as providers of financial services. That is what Parliament has asked us to do. 

Let me now move on to the third case involving the use of Big Data. Consider the following: greater availability of information enables better pricing of risk, in other words the first case of driving quality. The incentive created in this third case is, however, to reduce risk in a way that cuts across another objective of public policy which sits outside the objective of the FCA. This is a rational but undesired outcome which therefore creates a problem in public policy.
I have just described the case of flood insurance in the UK. Big Data enables more precise individual modelling of the potential for flooding of properties. By the way, I am for the sake of avoiding an argument heroically abstracting from the debate around climate change and whether it is raining more now or will do in the future. 

The incentive is therefore going in the direction of not living on flood plains thereby reducing flood related costs and risk. But, in another area of public policy outside our remit, and rightly so, there is a desire to increase the supply of housing and its stock and that to do this is some parts of the UK means building more on the flood plain. Here in is an obvious potential conflict of public policy objectives. 

The answer to this dilemma has been to create Flood Re. The essence of this is to create a body which enables the achievement of a public policy objective which requires people who value living off the flood plain and in less dense conditions to make a payment through a higher insurance premium to those who are prepared to live on the flood plain but only if the price of insurance is lower than it would be in the absence of this provision. 

This is a very good example of how we always need to solve what I describe as boundary issues – where different public policy objectives come together and can create conflicting objectives. My view is that resolving these boundary issues is properly a role for Government rather than a sectoral regulator on one side or other of the boundary. 

I have slightly laboured this point for a couple of reasons. First, because failing to have a clear understanding of both the issue of risk-sharing and where the responsibility for dealing with it should lie, will lead to less good outcomes. 

Second, it seems to me that insurance is an area of financial services most likely to run into more of these issues in the future. Let me give just one example of what might happen. I don’t know whether it will by the way, so don’t read this as more than a “what if”. Let’s suppose that genetic identification really does revolutionise the prediction of life expectancy and each person’s probability of suffering from dementia. 

The implications for life insurance are potentially profound. It goes beyond the scope of my first case – we cannot simply react by changing our lifestyles and then insurers can model a revised distribution of life expectancy. Now, maybe the reaction to improved identification is to say “it is what it is” and we accept the implications for purchasing life insurance. Or, maybe we say that it creates unacceptable divisions within society – outcomes that are not acceptable for society in terms of access to insurance. By the way, I don’t know the answer to this question is obvious. What I do know is that this is a question that needs to be answered by involving Governments rather than solely regulators or firms. 

Let me summarise the argument so far and then put in a plug. I have identified three cases which draw on the impact of Big Data. The first is beneficial all round – if indeed Big Data refines the measurement of risk and incentivises better driving. The second creates unacceptable incentives – to charge more to naïve or vulnerable consumers. The FCA has a responsibility to tackle this under its public policy remit. The third example creates a potential boundary conflict among different objectives of public policy. It is in my view for Government to solve such challenges. 

The Mission document is not about re-defining and changing our role, but rather about explaining what we do, in terms of both the “why” and the “how”.

So now for the plug. Last month we published a document on the FCA’s Future Mission. It is out for consultation until late January; please give us your thoughts. In it you will see quite a few of the points I have just made and some more. The Mission document is not about re-defining and changing our role, but rather about explaining what we do, in terms of both the “why” and the “how”. It deliberately focuses on the big issues we face – how we think about vulnerable consumers and whether we place more emphasis on remedying the problems they face rather than focussing more evenly on all consumers. It also considers how we deal with the boundaries of different areas of public policy while preserving the essential operational independence of the FCA. 

The reason for publishing the Mission document is very simple: the FCA is a public authority carrying out public policy and our effectiveness depends substantially on our ability to set out and explain the framework in which we operate. This is an important and I should say not easy task. 

Let me now move on to the second area of my remarks today, concerning future risk sharing in the Financial Services Compensation Scheme (FSCS). FSCS is a mutual insurance scheme in its own right, covering most of the scope of financial services in the UK with limits on the scale of insurance. It covers therefore a very broad scope of financial activity of different forms. 

I want to start by briefly going back to my former world at the PRA and the Bank of England (though let me quality “former” by pointing out that it is a great pleasure for me to still be a member of the FPC and the PRA Board). One way to define the separation of prudential regulation between the PRA and FCA is to point to the fact that the PRA is in the business of regulating for prudential purposes firms that take the money of consumers as principal – i.e. with the risk on their own balance sheet. This broad proposition holds across deposits, savings and premium paid. Most FCA only firms manage the money of consumer as agent or where they take risk as principal these activities are not a threat to the overall system (I am using the term FCA-only carefully here to define the scope of FCA prudential regulation, of course, FCA conduct regulation goes across the whole piece of financial services).

In the world of firms acting as principal for consumers’ money, their capital base forms the front line of defence. A lot of Solvency II is about that. The FSCS is the backstop if you like, what is there to insure up to a defined point in the event of the insolvency of the firm.

Let’s now turn to the world of the FCA only firms. Here, the FSCS is called upon where the firm has acted to the detriment of its customers in an agent or advisory capacity and it has become insolvent so that it cannot make good that detriment.

Two points stand out for me here: first, while losses as principal dominated the financial crisis – in other words, the failure of banks – and the interest costs of these failures remain even now the biggest costs being covered by the FSCS, failure of firms acting as agent are more common, if on average smaller. 

But size has to be related to the resources of the FSCS, something I will come back to. For agency firms, capital is not where we should start - as holding capital is an inefficient way of mitigating the risks posed by firms who are not typically a threat to the system. The front stop should be commercial insurance in the form of Professional Indemnity (PI) Cover, much as it is for lawyers and other professions. But PI cover is not by experience always reliably performing the role, particularly in the IFA and investments world, – the contracts are framed often in ways that rule out loss absorption in the context we are dealing with here when the firm fails. What is the consequence of this? It is that the protection of client assets and ultimately the FSCS become the primary lines of defence, and this is what has happened. 

You can see from this that in the principal world, the pooling of risk and thus the insurance occurs in the backstop, whereas in the agent world it occurs in the frontstop. The latter in my view requires further thought. On the face of it this is a public policy issue in terms of the availability of insurance and where the pooling should occur. Pooling can here take one of two forms. It can be pooling through private insurance (PI cover) or pooling through mutualised industry risk cover via the FSCS. It is important to note however that there is a very important difference between private insurance with pooling for which a risk-based premium is charged – assuming that insurers can obtain sufficient data to charge a risk-based premium - and the FSCS, in which a risk-based premium is not currently charged. The frontstop problem in the agency model is really caused because there is no risk-based cost. This creates a very marked incentive problem. 

Why is this relevant today? It is because we are currently conducting the review of FSCS pooling on the agent side of its activities. This is not primarily a review about whether there should be insurance for consumers but rather about the sharing of the burden of the pooled risk across the agent world. Here, there are two tensions and no easy answer. The first tension is how to divide the provision of risk cover among the various sub-sectors of the finance industry. The tension is between spreading the cost enough to avoid damaging other firms seriously versus spreading it so much that firms are paying price for failures that are a long way from their area of activity. The second tension is how to spread the cost in the agency world between the creator of the product and those who sell and/or advise on it. 

But these tensions would be at least less pointed if they were not trying to cover a world where there is no risk-based pricing of insurance in the front-line.

This is rightly a public policy issue, but it is also a private issue too because many advisory firms are meeting substantial bills for FSCS pay-outs. With all this in mind, my request today to the insurance industry is to help us to think through how we might solve this problem. 

Conclusions

Insurance is a very important part of a well-functioning economy and society.

Insurance is a very important part of a well-functioning economy and society. It matters a lot for the successful pursuit of public policy, and in turn it can produce some important public policy challenges, I have set out what I believe to be some important ones in today’s world. 

My two cases are different. The first – on Big Data – illustrates how insurance affects society and draws upon the data from it. Understanding the effect and significance for insurance of Big Data and how it evolves requires a clear framework to disentangle the issues. It emphasises the work we have set out in Our Future Mission as a means to explain our part of the public policy landscape. 

The second case study involves the provision of a particular form of insurance which matters for the FCA to pursue its objectives. At the moment I would say that the insurance scheme we have to limit the cost of financial conduct risk is highly limited outside the FSCS provision. If there is a desire to solve this problem and I hope there is, we need to work together to determine what is possible on professional indemnity or related insurance.