Fighting financial complexity with simple rules?
Day by day we experience that football bets get won or lost, cardiologists’ patients survive or die and investors get rich or ruin themselves. According to Andrew Haldane, the common trait of all these situations is that they require decision-making under a high degree of complexity. In principle, to succeed in these cases would require a huge amount of information about a number of influences whose interactions cannot be reduced to simple forecasting models. In a breath-taking tour de force Haldane argues that here the optimal response is not to fine-tune the reaction to every eventuality, but to follow simple rules of thumb. In a twist of Herbert Simon’s bounded-rationality approach, behavioural economics along the lines of Daniel Kahneman and Amos Tversky, and his own ideas about complexity which I discussed elsewhere, he states that the advantage of those rules is that they are more robust to ignorance. Citing experimental evidence from various disciplines Haldane writes: “In complex environments, decision rules based on one, or a few, good reasons can trump sophisticated alternatives. Less may be more.” (Haldane 2012, p. 5) In his view, the same holds for financial regulation responding to financial crises, and he lists several alternatives to the current regulatory framework which in his view would increase regulators’ potential for crisis control. This is where I’m beginning to have my doubts.
The world financial system is vast and opaque. There are large numbers of banks. Some of them are huge in size and linked to wide networks of smaller institutions. There are shadow banks of all sorts, often established in reaction to regulatory efforts and located in offshore financial centres, which are beyond the reach of policy. The same holds for the intra-firm activities of multinational corporations and conglomerates circumventing banks and markets.
Size and opaqueness pose a big problem to financial supervision and regulation. So, why additionally worry about complexity?
Because complexity may make things worse.
Generally, in a non-complex environment, there is a stable and unchanging direct relation between cause and effect which can be observed or at least inferred from data analysis. For example, this allows to build a model and make statements about how in general a system will react to a policy measure, even if the system’s internal responses are unobservable.
In complex systems, generalization and reduction to simple causal relations which hold once and for all, and inferences from the past, may not be possible. The emphasis here is on processes of information acquisition, learning and interaction of so-called complex adaptive systems (CAS) which are constantly changing and evolving. The Frisbee-catching dog in the Haldane paper is a complex adaptive system in this sense – and a very popular example. Murray Gell-Mann (p. 19) tells the following story:
“Isaac Asimov, the late biochemist, popularizer of science, and science fiction author, told me that he once had an argument with a theoretical physicist who denied a dog could know Newton’s law of motion. Isaac asked indignantly, ‘You say that, even after watching a dog catch a Frisbee with its mouth?’ Obviously, the physicist and he were using ‘knowing’ to mean different things” …
Complex adaptive systems are said to be self-organized. Their components may interact in ways that lead to a phenomenon known as “emergence” – a collective property of the system which makes the whole become more than the sum of its parts. Andrew Haldane mentions the example of avalanches. To cite Per Bak (p. 50 f.), the “master of sandpiles”:
“Consider a flat table, onto which sand is added slowly, one grain at a time. … Initially, the grains of sand will stay more or less where they land. As we continue to add more sand, the pile becomes steeper, and small sand slides or avalanches occur. The grain may land on top of other grains and topple to a lower level. This may in turn cause other grains to topple. The addition of a single grain of sand can cause a local disturbance, but nothing dramatic happens to the pile. In particular, events in one part of the pile do not affect sand grains in more distant parts of the pile. There is no global communication within the pile at this stage, just many individual grains of sand.
As the slope increases, a single grain is more likely to cause other grains to topple. Eventually the slope reaches a certain value and cannot increase any further, because the amount of sand added is balanced on average by the amount of sand leaving the pile by falling off the egdes. … It is clear that to have this average balance between the sand added to the pile, say, in the center, and the sand leaving along the edges, there must be communication throughout the entire system …
The addition of grains of sand has transformed the system from a state in which the individual grains follow their own local dynamics to a critical state where the emergent dynamics are global. … The emergence of the sandpile could not have been anticipated from the properties of the individual grains.”
As the sandpile example indicates, information, learning and communication may take different forms. Murray Gell-Mann (p. 19) wrote:
It is not only learning in the usual sense that provides examples of the operation of complex adaptive systems. Biological evolution provides many others. While human beings acquire knowledge mainly by individual or collective use of their brains, the other animals have acquired a much larger fraction of the information they need to survive by direct genetic inheritance; that information, evolved over millions of years, underlies what is sometimes rather vaguely called ‘instinct.’ Monarch butterflies hatched in parts of the United States ‘know’ how to migrate, in enormous numbers, to the pine-clad slopes of volcanoes near Mexico City to spend the winter.
Another important characteristic of complex adaptive systems beside the constant interaction with their environment and the emergent property is that they may exhibit a high sensitivity to initial conditions also known from deterministic chaos. One grain of sand may make the difference – or, as the title of a paper by Edward Norton Lorenz, a mathematician and meteorologist, and one of the first proponents of chaos theory, suggests: The flap of a butterfly’s wings in Brazil may set off a tornado in Texas.
The problem here is that we know that these effects may exist but we cannot follow the paths and interactions of millions and millions of butterflies, grains of sand and other influences to foresee the emergent phenomenon, let alone influence the system’s course.
The financial system exhibits all attributes of a complex adaptive system. Constantly exchanging information, learning and evolving its behaviour cannot be reduced to a simple model of few ever-lasting relations. Periods of calm alternate with turbulence. There are local disturbances and global crises, but there is no way to foresee when and under which circumstances a local phenomenon will become a global threat – or fade away without leaving a trace.
How does financial complexity become manifest? To answer this question I would like to distinguish between market processes and institutional relations. Andrew Haldane (and other authors such as Greg Fisher, Stefania Vitali et al., and Stefano Battiston et al. who developed DebtRank, a measure of systemic risk) put the emphasis on financial institutions, on banks and their vast networks of hubs and spokes. The biggest and best connected banks are considered to be “super-spreaders” which in analogy to epidemiology have a high capacity to infect counterparties in times of crisis.
Admittedly, chain reactions within financial networks are a key element in the transition from local to global phenomena and the propagation of crises. But, in my understanding, networks are not the source of financial complexity which makes the system alternate between calm and crisis. To see this, we have to look for the emergent property and its roots.
The emergent property financial regulators have to cope with is the seemingly aligned behaviour of the many which makes a crisis transform from a local to a global phenomenon affecting the whole system. Experience shows that, in general, the first seed or starting point of such a crisis is an individual reaction, a change of view, in the market. Other participants may adopt this new view. Gradually or abruptly an ever-wider community of actors may follow. This change of view may refer to prices, conditions of demand and supply, or individual liquidity or solvency. It may be based on solid new information or simple rumours. In some instances, it may be overruled by other events and have no lasting effect, in others it may spread rapidly with the market beginning to show signs of self-organization and “swarm intelligence” (or stupidity). Herd behaviour and “instincts” take over. Dynamics are reinforced by loops and feedback effects – another important element of complex interactions within the system and the outside world which is lacking in the unidirectional super-spreader analogy. The more actors become convinced of the new view, and the larger the group following the first “deviants”, the more will be tempted to adopt this view as well and jump on the bandwagon. The system is transformed from a state of low activity and local unrest to high turbulence and overall instability.
In this process, some actors win, others lose. To the extent that “super-spreaders” are among the losers, financial institutions are increasingly threatened to become an endangered species – and the challenges to financial regulators grow exponentially with every trading day.
Can simple rules enhance the potential of financial regulation for crisis control?
At least, they may perform better than existing ones. Although I do not agree with Andrew Haldane’s diagnosis of the roots of financial complexity, I agree with the general conclusion. It makes no sense for regulators to try to capture all shades of financial risks and activities. There is no way to foresee where in the system dangers loom. The rules must be robust, making banks better prepared to withstand crisis and less reliant on public support. Furthermore, the rules must enable regulators to fulfill their task effectively.
Haldane argues on the basis of the existing Basel framework. He wants the current regulatory architecture to be simplified including the removal of internal models to measure credit and market risk and the reconsideration of measurement of equity capital. In addition, he calls for a reform of the three pillars on which the Basel framework is based simplifying regulatory rules (Pillar 1), providing greater scope for supervisory judgment based on experience and rules of thumb (Pillar 2) and making risk weight information simpler and more consistent as a contribution to restore market discipline (Pillar 3). Other proposals include the taxation of “complexity” to address the “growth in opaque, intra-financial system chains of exposure” as well as size limits and forced separation of commercial and investment banking.
All this may well help reduce regulators’ uncertainties. Besides, simplifying existing rules and limiting financial activities may well render banks more robust. But I cannot see how the proposals may contribute to reducing financial complexity or controlling crises.
In a complex adaptive system simple rules allow actors to make the right (or wrong) decisions without knowing the way the system works. In the financial markets, there are numerous examples of financial decisions made in this way. For instance, there are countless variants of “chartists” and “fundamentalists” or a combination of the two, with each investor having his or her own recipe.
The question is which role financial regulation can play in this world. In his paper, Haldane does not address the market aspect. Nevertheless, the proposed rules may have an effect on market complexity, although unintentionally (and hard to isolate and measure):
As mentioned, complex adaptive systems show a high sensitivity to initial conditions. There is a chance that altering the rules of the game in the way Haldane proposed will change the system’s nature making it less (or more) prone to crisis. If this is intended, or hoped for, however, we have to ask: Shouldn’t policy rather focus directly on market conditions? Weren’t entirely different, new rules in this case preferable to Basel III modifications to cope with complexity?
The problem of containing manias, panics, and crashes is still unsolved.