Financial regulation is a young science grappling with some very old questions. In excerpts from a lecture at London University, FCA Chief Economist Peter Andrews argues policy makers can better answer some of these questions by adopting a ‘multi-perspective’ approach. Blending traditional economic expertise with insights from other fields and drawing on leaps forward in areas like machine learning and eventually neuroscience.
Speaker: Peter Andrews, Chief Economist, FCA
Location: Conference on behavioural finance held by Queen Mary University of London School of Business, Lancaster Business School and Santa Clara University, 14 June 2016.
This is a conference about behavioural finance so I should say that my themes are wider than behavioural finance itself. I intend briefly to explore:
- the role of the regulator
- how economics is helping us regulators in our role
- innovations in regulatory economics
- challenges for the future
First, though, as a practitioner of regulation, I view behavioural finance as an important part of the set of economic and other disciplines that enable the FCA to pursue its operational objectives of protecting consumers, ensuring market integrity, and promoting effective competition. Behavioural finance is therefore important in all the themes listed above.
The need for a broad perspective arises partly because the FCA’s operational objectives are couched within a strategic objective to ensure that the relevant markets function well. We find in practice that understanding these markets is hard if we take a solely behavioural, neoclassical or any other approach.
In fact, to add some detail, the first theme of my talk is that to make markets work well regulators need to understand all material drivers – including behavioural drivers - of the market equilibria we observe. And understand how to improve them.
The second theme of my talk is that economics is a critical discipline for improving regulatory policy and unique in its ability to help us make markets work well.
The third theme of my talk refers to the question mark in my title ‘Beyond economics?’. Clearly, regulators need to use disciplines beyond economics, for example to ensure the legitimacy and legality of their actions. The question mark refers to the expanding boundaries of economics. What might have seemed beyond economics a few years ago might now appear mainstream. I will discuss how the FCA is pushing the boundaries of practical economics.
My fourth theme draws out the collective implication of the first three.
One can think of economics as the science of decision-making by economic agents and of the markets that result from their decisions. Regulation occurs when the body politic does not like aspects of these markets but prefers to change the markets rather than supplant them with state actors. One can therefore think of regulation as the science of understanding why markets are failing to meet expectations and of how to change the decisions of economic agents so that the markets work well.
This description makes the close connection between economics and regulation inescapable. It also means that to be practically useful economics must explain what is really driving the market equilibria we observe. And this is not straightforward. Hence my fourth theme: challenges for the future.
My perspective is that we must discuss how economics and other disciplines can help regulation because the challenges regulators face are large and chronic. Here is a brief historical detour on the point.
Financial crises have occurred almost throughout recorded economic history. Not least in the UK, with several crises in the eighteenth and nineteenth centuries. Reinhart and Rogoff starting in the 1300s in their book ‘This Time Is Different: Eight Centuries of Financial Folly’. Which looks like a behavioural finance perspective…
And the costs of crisis could be huge even before finance grew so large.
For example, Sheridan in ‘The British Credit Crisis of 1772 and The American Colonies’ concludes with qualified support for the idea that, despite calming initiatives by the Bank of England, the UK bank runs of 1772 and the taxation policy responses to it helped start the American War of Independence. Which I guess was extremely costly, at least from a British perspective!
Moreover, the wily have separated the unwary from their assets throughout recorded time. Consider the rudimentary financial exchanges that developed around London’s coffee houses in the seventeenth century, closely following similar venues in Amsterdam. Koudijs and other show that in these British and Dutch trading venues the ‘insiders’ systematically exploited the ‘outsiders’.
Even the idealistic, revolutionary United States were not immune. In 1792, Assistant Secretary of the newly formed Treasury, William Duer, sought to exploit inside information from First Secretary, Alexander Hamilton. This created a bubble – lots of people watched who bought and who sold what, used this information to trade themselves, and crashed the securities market.
It is hard to escape the notion that financial regulation is a young science – substantive UK conduct regulation started in 1986 - grappling with very hard and very old questions, implying scope for regulatory innovation.
Theme one: using multiple disciplines to understand the drivers of market equilibria
If regulators are to make markets work better, we need to understand whatever materially drives the equilibria we observe. In particular the decisions that firms, consumers and, of course, regulators take.
Nor should we forget market structure, competitive dynamics such as firms’ strategies under imperfect competition, information, culture, values, emotions, beliefs, behavioural biases, and more. All of which influence and in some cases reflect decisions.
And then there are regulators’ powers, public law more widely, regulators’ licence to operate, compliance culture and non-compliance culture, contract design, law of contract, law of tort, consumers’ general rights, consumer advocacy bodies, marketing by firms, institutional structures such as courts and so on...
All this looks intractable for a formalised discipline like economics seeking ‘proofs’ through models that simplify the world, for example assuming that somebody is maximising something perfectly rationally or subject to a single identified bias. But financial contracts are generally about storing or growing financial resources for the future. A future in which personal and general circumstances are unknown. Meaning that models which proceed as if there is a knowable optimum can feel uncomfortable for the practical regulator.
One implication is the critical role of policy analysis and sound judgement: carefully considering whatever is relevant and not considering what is not relevant. This is a formulation of the public sector duty ‘sufficient enquiry’ and today’s discussion suggests meeting this is a big challenge. Which is probably as it should be if regulators are to interfere with property rights.
A behavioural perspective
It is nevertheless useful for a conduct regulator, especially at this conference, to use a behavioural lens to start discussing drivers of financial markets. This does not mean debating whether ‘behavioural errors’ or ‘biases’ really are a market failure providing an economic case for intervention. The FCA is clear that they are. See our occasional paper on ‘Applying behavioural economics at the Financial Conduct Authority’, 2013. Similarly, the Financial Services Authority published ‘Financial Capability: a Behavioural Economics Perspective’ by de Meza et al, 2008.
Moreover, this stance on the importance of behaviour is borne out by our practical experience of regulating markets. For example, the consumer intelligence we gather. And various behavioural experiments have shown that financial consumers are biased in some ways. See various occasional papers on our website.
But there are liberal concerns that using an ever-growing list of supposed behavioural errors can justify almost any intervention. So I’ll share a few observations.
A well-working market likely depends upon a demand side disciplining suppliers directly or through its agents. If the agents are unbiased… But transaction values may not allow such agents to operate at a price people are willing to pay. Or at a price that it would be sensible to pay. So we have to consider the demand side – consumers.
Consumers can discipline suppliers if they have:
- the information needed to discriminate between offerings, and
- either the technical understanding/framework plus processing power (a burden eased by well-designed comparative tables covering features as well as price), plus the time needed to deal properly with the information, plus the self-control and awareness needed to over-ride unhelpful beliefs and emotions
- or, in the absence of these, useful heuristics that lead to satisfactory decisions and are not undermined by clever marketing.
Perhaps this happy state of affairs held in the Garden of Eden, at least until the snake started talking. But it is unlikely to hold in a market for complicated products with a dynamic supply side that learns about and responds to how consumers make decisions.
This is the argument of Akerlof and Shiller in ‘Phishing for Phools’. They say financial consumers are especially vulnerable as many financial products are credence or experience services, given suppliers’ need to pursue profit which, they argue, is both the strength and weakness of the market economy.
Sophisticated suppliers can observe consumers’ heuristics. For example, through live experiments on their websites or in ‘controlled High Street environments’. Then they can design responses to undermine these heuristics. Even use personalised responses based on big data. All suppliers can observe consumers’ decisions under different conditions. Then change marketing strategies to sell more of what they want to sell at prices nearer the ones they want to charge.
Also, complexity is valuable for suppliers who want go to ‘phishing’. It can be increased, for example to make assessment of value harder.
As we can see, ‘behavioural biases’ are not a vague concept to be deployed by regulators at will. They are in fact intimately connected with the traditional market failure that justifies conduct regulation ‘information asymmetry’, with the challenge of using information in decisions when it is available and with the dynamics of the market.
Psychological insights are surely making economics more useful to policy makers. Improved understanding of behaviour – put in the context of other drivers of how a market works – generally improves models. This is why the FCA recently held a conference on behavioural industrial organisation. The point was to recognise how important behavioural experiments are for exploring problems in markets and to what extent, if any, specific nudges will change choices in experimental context, while accepting that they cannot tell us what equilibria will evolve under a nudge that changes some choices: rigorous behavioural and other economic analysis is required to begin to explain this.
We find Mark Armstrong’s paper ‘Interactions between Competition and Consumer Policy’ very useful for analysing the operation of markets in the presence of imperfect consumer decision-making. Our occasional papers such as No. 12 on showing the prior premium on insurance renewal notices and No. 3 on how the add-on mechanism for insurance sales influences consumers show our approach in practice.
The length of the list of influences on decisions of economic agents and markets at the start of this theme means I cannot discuss in any detail the role of other disciplines in understanding markets. The key point is that the relevance of these disciplines is real and significant. The markets we observe are the product of our society, our people, the ways our brains work and the structures that make pursuit of profit a central function of most service providers. How these forces interact in specific cases may be very complex.
Another important point is that economics, sensibly, has a history of drawing on other sciences that are helpful for understanding markets.
Most obviously, economics uses mathematics and statistics and arguably physics. Drawing on psychology to improve analysis of demand (and supply) side behaviour in behavioural finance is a natural step in the evolution of economics. Recently the Bank of England used evolutionary dynamics to explore development of crises. Today, some are optimistic about use of neuroscience in economics.
Further, economics draws on developments in law, risk analysis and public sector management, and reflects social policy and political context. The famous economics textbook by Samuelson and Nordhaus was through many editions optimistic about the prospective triumph of Soviet five-year plans. Apparently adverse weather kept intervening. Who’d have thought it, in Russia?
Anyway, at the time of writing, the observations probably seemed routine to the authors. Not much later, it felt very strange to be required to absorb this material when undergoing my initial training in a City firm, in the era of Margaret Thatcher and Ronald Reagan. But soon enough there were new economics books on monetarism supplanting Keynesian ideas…
How economics might use other sciences to develop itself further is the future challenge to which I will refer in my fourth theme.
Theme two: the unique role of economics
None of the forgoing is meant to play down the fundamental importance of economics for understanding the level of efficiency of equilibria in markets and changing them for the better.
Economics enables regulators to deliver true benefits, making society as a whole better off rather than just making transfers between interest groups.
Some ways in which economics helps regulators assess market efficiency
Improving regulation raises difficult questions. What are the right criteria for intervention in markets? Do we mean to improve total welfare or the welfare of specific groups? Or is welfare the wrong metric? Many of these issues are discussed in Stiglitz’s, ‘Regulation and Failure’, which informs chapter one of the Tobin Project.
And there are more questions in this vein. Which remedies work best for which problems? What is the optimal combination of regulatory tools? How does this differ by regulatory goal? What is the right unit of account for designing interventions: it might not be an economic market or even a sector if multi-product firms with huge fixed and common costs are operating across sectors. Or if consumer inertia allows waterbed strategies in some realisation of Demsetz’s model of competition FOR the market. What is the effect of the regime overall relative to the sum of its parts?
All of these questions require economic analysis.
Nor should the value of economics be under-estimated due to weaknesses in what the discipline had to offer. Unrealistic demand-side behaviour, models without causation, inadequate data, and so on. These issues are being addressed.
For example, the FCA and others are integrating behavioural insights into industrial organisation as mentioned earlier. On causation or, rather, lack of causation David Spiegelhalter’s ‘Cambridge Coincidence Collection’ is interesting. Again, the FCA seeks to develop econometric models that can show causation through use of instruments or regression discontinuity design.
Moreover, even if regulation cannot be solely about economics, it also cannot be solely about any of the other disciplines.
Law matters as a vital constraint on what regulators can do. Contractual analysis can help us understand whether consumers are getting a good deal. But it does not tell regulators how to maximise societal benefits.
Risk analysis matters because it makes sense for regulators to forfend foreseeable harm to consumers, where possible. But risk analysis by itself cannot be the sole basis for regulatory strategy to the extent it uses the status quo in the way prices were used in the ‘Cellophane Fallacy’ case under the Sherman (competition) Act in 1956.
In that case, the analysis, which was accepted not by fools, far from it, but by the US Supreme Court, was that Du Pont did not have excessive market power because it could not profitably raise prices from their current levels.
But this inability in fact arose because Du Pont had already set its prices as high as it sensibly could, making further increases unprofitable – and this was precisely because Du Pont had huge market power as the monopolist in the cellophane market. Thus the market status quo was not the right benchmark for assessing harm.
Similarly, any approach which, say, focuses on non-compliance with existing rules or on dangers arising from new products or technology implicitly sets the status quo as the benchmark for measuring (and trying to prevent) harm. This matters because when the status quo is fundamentally suboptimal as a result of inherent market failures, using it in this way will prevent recognition of scope for improvements, for example through correcting market failures to empower the process of competition.
Clearly, the alternative disciplines have significant gaps and cannot be the sole basis for regulation.
In fact, even if a regulator wishes to focus on non-economic perspectives, economics still has a critical role in securing legitimacy (operating licence). This is because it is needed to ensure that the regime is proportionate. Legitimacy will be lost as stakeholders see net costs mounting at regime level from regulation which does not correct market failures. There could also be significant challenges under Administrative Law.
Theme three: how has the FCA been using developments in behavioural finance and other branches of economics to improve regulatory policy?
Let’s look at three disciplines being used by the FCA to improve regulatory policy: computational, behavioural and public sector managerial. Some papers I’ll mention today will appear on our ‘Insight’ website:
My examples of these disciplines – and I hope some of them will spark useful ideas – are, first, computational:
- big data – an application in a Market Study of competition
- innovative mathematics – an application to Supervision
- machine learning – an application to understand the mortgage market
Second, two behavioural examples:
- neuroscience – the dog that did not bark (yet)
- behavioural analysis to promote firms’ compliance – a new paper
Third, two examples from public sector management:
- behavioural analysis to improve regulatory performance
- bringing it all together: insights from ‘public sector management’ to improve regulatory policy-making and cost-benefit analysis
In our big data work on credit cards we have combined transaction and Credit Reference Agency data to produce a customer level view of indebtedness covering over 2.5 billion monthly cycles. Through mapping postcodes to longitude and latitude co-ordinates, we have used visualization techniques to build understanding of the relationship between debt, income and arrears.
In deploying innovative mathematics in Supervision, we are exploring compliance with suitability requirements in a way that builds on work by Lewis and Hoberg for the US Securities and Exchange Commission.
The SEC’s problem was that large numbers of firms reported financial information and it mostly did not know which reports were false. It used latent dirichlet analysis to detect language clusters: what patterns were in the free text disclosures of (ex post) known liars; then who else used these patterns (of exaggerated normality). Targeting supervisory forensics on this basis increased the rate of detection of false reporting.
We are using machine learning to research how well the domestic mortgage market is working. For example, how far do consumers’ choices diverge from a ‘financially good’ deal? This requires us to assess credit risk of borrowers. We therefore combine two large datasets of mortgage transaction reporting and performance to design a model for predicting risk of loan problems as a function of facts known to firms ex-ante (demographics, location, etc).
Many predictors of credit risk are available but we need a parsimonious model for interpretability and to avoid poor out-of-sample performance from over fitting the data. We use machine learning algorithms for data driven variable selection to pick risk predictors that are tractable and optimally trade off the in and out of sample predictive power.
We have not applied Neuroeconomics but have considered whether it could help us, given some apparent usage by firms.
The underlying rationale would of course be limited understanding of the drivers of people’s financial behaviour: there are multiple plausible explanations of the same economic phenomenon. Let’s say that Neuroeconomics is the application of neuroscience and its techniques (brain imaging, reaction times, eye tracking, etc.) to individual economic behaviour. As such it may tell us the drivers of decisions and something about welfare weights, biases and preferences (e.g. risk, valuation, reward, cooperation, fear...).
An interesting example of retina scanning is work by Marotta Wurgler and others on consumers reading on-line financial contracts… almost not at all. Surprise, these contracts changed over time to the advantage of suppliers!
Another application could be analysis of the effects of firms’ marketing strategies. However, some techniques (brain imaging) are more costly/less readily available than others (surveys) and interpretation of results is very difficult. In terms of when to expect what, I hope Herbert Simon’s paper on what the ‘digital computer’ might achieve is not informative. His ‘Theories of Decision-Making in Economics and Behavioral Science’ appeared in the American Economic Review in… 1959.
Behaviour in firms
Compliance with the spirit of financial regulation by UK financial services firms is sometimes remarkably low. We have seen systematic and apparently indefensible mis-selling scandals over decades. These include pension transfers, mortgage endowments, pay-day loans and PPI. We are, therefore, actively considering behavioural insights into firms’ compliance.
Our paper explores the standard ‘credible deterrence approach’ and behavioural drivers of compliance. How to make detection and punishment highly salient? Or to de-bias decisions - for example how to reduce the sometimes negative impact of loss aversion and endowment effects on the quality of governance and internal controls? Or how to increase ‘compliance morale’ and the impact of positive moral codes?
We are also seeking compliance lessons from the experience of UK and other tax authorities, in a paper for us by Professors Gareth Myles and Chris Heady. An interesting question raised is whether we can find an equivalent of HMRC’s tax take metric. This assesses the level of tax compliance by comparing actual tax revenues with expected tax revenues based on levels and types of economic activity. It is not clear that an equivalent can be developed for financial services regulation but it merits consideration.
Another important aspect of Myles and Heady’s paper is compliance morale. The idea that circumstances can be created in which people actively want to do the right thing rather than do something legal but obviously wrong.
Sean Hundtofte’s research on the consequences of ‘rescheduling deals’ offered to people with second mortgages in the United States shows this can be important in financial services. Briefly, psychologically attractive deals were offered to consumers in arrears. In principle, these deals provided a path to redemption. Statistically, however, the likely outcome of each deal was known by the firms but not by the consumers to be that the consumers would make higher payments to the firm before defaulting. And indeed this is generally what happened.
Behavioural analysis to improve regulatory performance
We are using the work of David Hirshleifer and others to consider whether and how the behaviour of regulators might be improved. What are our biases? And how do they affect our problem diagnosis, design of interventions and ex post monitoring?
For example, diagnosis may be affected by salient or vivid information, or confirmation bias may lead from regulators’ beliefs about regime efficacy to uncritical designs of interventions, which might be reinforced by over-confidence. How could this be mitigated?
Possibilities include external scrutiny, CBA, adjustments to address over-confidence, and sunset clauses to reduce salience.
Public sector management more broadly
My final example is ‘Bringing it all together: insights from ‘public sector management’ to improve regulatory policy-making and CBA’. This is really about uncertainty in public policy and difficulty of measurement both of problems and of effects of supposed solutions. It draws for example on the work of Charles Manski (‘incredible certitude’). What we are trying to do is develop a realistic approach to analysing markets and developing solutions to market failures that, to the extent possible, do not require us to pretend to know things we do not or even cannot know.
Our starting point was being sceptical about quantification and some other requirements in various Cost-Benefit Analysis and Impact Assessment Guides. Our extensive review of these guides and published work based on them showed that very few institutions follow the standards they set. We considered reasons for this – problems of welfare and well-being analysis, incredible certitude and uncertainty, for example about the counterfactual, and so on.
We see possible help with this in the public sector management literature on ‘wicked problems’. That is, problems which do not admit of formulaic remediation or tractable measurement; instead requiring a multi-perspective approach. This literature also discusses alternatives to Impact Assessment and Cost-Benefit Analysis such as measuring a wide range of factors closely linked to regulatory objectives.
We have brought together some insights from ‘Public Sector Management’, traditional economics, behavioural economics and big data approaches in a new guide on, occasional paper No.13.
Theme four: regulation as the science of understanding why markets are failing to meet expectations and of how to change the decisions of economic agents so that the markets do work well
Under this view of regulation, what can we do to improve it? As discussed already, we need multiple perspectives from within and beyond economics for regulation to be effective. And the FCA is drawing on economics, including new developments in behavioural finance and big data, as well as on law, risk analysis, consumer and market intelligence, Public Sector Management and Futures, to develop a new and better regime.
This does not cover all drivers of market equilibria set out in theme one. Some remain to be explored, and some drivers may not require great attention. However, there might usefully be increased granularity of analysis in pursuing the core function of economics in conduct regulation: to explain how the markets work.
I believe that improved data, combining transaction and personal data sets and improved software and Cloud processing will facilitate more granular analysis.
Ideally, we might consider which people sign what contracts. What choices will the suppliers offer? Will the right choices be made? And what will these contracts say? Critically, for a competition authority, will the prices of the contracts represent good value – that is, be cost-reflective – or will the prices be excessive for the quality offered? And, again, what will happen to efficiency and welfare overall?
The central problem of decision making
Conduct regulators want to change consumers’ decisions. So we need to understand how decisions are made. As already described, this is complicated.
But we should be optimistic. There was economics before information economics. There was information economics before behavioural economics. Behavioural economics is a valuable step forward. What could a sharper focus on ‘decision economics’ mean? Can the evolution of economics be hastened?
For example, how long will it be before neuroscience can tell us how interactions between brain areas happen, what influence they have on decisions, what specific emotions and thoughts are being experienced, and so on? Again, let’s hope we are past the stage of Herbert Simon’s 1959 remarks on ‘digital computers’!
Akerlof and Shiller in Phishing for Phools propose the stories people tell themselves as a way of unifying the disparate elements of decision making. A central question is whether regulators can use this idea or other ideas about how decisions are actually made to design remedies that will work. Here is a brief reflection on it, based on the idea that financial decisions are often ‘uncertain’ and that this matters.
Smith and Stern in Philosophical Transactions of the Royal Society list four types of uncertainty:
- Imprecision – the outcome is not known precisely but decision-relevant probability statements are possible.
- Ambiguity – this is also widely known as Knightian (1931) uncertainty; here we simply cannot make probability statements.
- Intractability – we cannot formulate or, if we can formulate, we cannot execute the relevant computations.
- Indeterminacy – there is indeterminacy with respect to quantities relevant to policy-making for which no precise value exists.
David Tuckett argues that neoclassical economics typically addresses situations as cases of imprecision and treats more radical uncertainty as ‘risk’ in order to apply probability calculus.
Similarly, he argues that behavioural economics uses heuristics and biases to describe potentially ‘bad’ (less reflective) use of evidence, implicitly as if there is a knowable optimum. It follows that cases of uncertainty that do not fall into the category ‘imprecision’ – and there are plenty in finance – will be challenging for disclosure and other ‘nudge’ remedies. Indeed David Laibson and other behavioural scholars have shown that typically ‘disclosure’ does not ‘work’.
One approach might be to categorise consumers’ decisions by type of uncertainty. Then decide what decision-making process is most likely to kick in for each category. Then decide what type of remedy is best-suited to each type of decision-making process.
Interestingly, Akerlof and Shiller’s work aligns well with David Tuckett’s psychoanalytical work on the role of ‘phantastic objects’ in finance. He finds intriguing examples in asset management and characterises financial markets as ‘markets in stories’, distinguishing ‘emotional finance’ from ‘behavioural finance’. It is easier for firms to create ‘phantastic’ objects than for regulators to do so, and the policy implications of this need consideration.
Again, the ‘Darwinian’ approach of Gigerenzer to decision-making may be useful. Our development as social animals clearly has had profound effects on how our brains work.
Together these perspectives suggest another categorization for analysis and policy. Rational use of information seems important in choices such as pension versus ISA. Emotions might determine choice of fund manager. Perhaps social norms prompt decisions to save rather than consume. And loss aversion bias helps people to choose to insure risks. While choosing gold cards or very high cost but low value cards like the Kardashian card presumably reflects man as social animal.
I trust my description of the FCA’s work shows we are serious about using the knowledge frontier to improve regulation.
A central challenge is to understand consumers’ decision making well enough to work out the real causes of problems in markets, how big the problems are and what is the most effective remedy for reach problem. The last point is critical to prioritisation.
The work by Akerlof and Shiller, Tuckett’s and Laibson mentioned above together form a big challenge for a competition regulator like the FCA. Disclosure is typically a pro-competitive remedy, while more intrusive remedies may not be. But they might sometimes be the best way to improve the markets we regulate. We shall learn more.
Meanwhile we seek approaches that appear best now. We draw on traditional economic ideas about incentives and equilibria, ideas from psychology about realistic behavioural assumptions and other insights into the markets we observe and how to make markets more efficient. We exploit large data sets, powerful computers and new analytical models. We see economics as constantly evolving and constantly becoming more useful. Behavioural finance is an important contribution to this.