Complexity and the madness of crowds – lessons from disaster prevention
Recent debates about the crisis vulnerability of international financial networks indicate that in economics complexity is still a widely misunderstood concept. In my critique of a recent paper by Andrew Haldane from the Bank of England I described complexity as a phenomenon rooted in markets instead of networks of institutions and in the interactions of countless individuals with different motives and strategies. The result is an emergent property, a seemingly aligned behaviour reinforced by loops and feedback effects, known in the literature as self-organization or “swarm intelligence”.
According to this view, actors in financial markets can be regarded as a “virtual crowd” to which the following comparison applies, too:
“Psychologists have likened a crowd to a series of intermeshing behavioral cells. Each cell is comprised of a small group of surrounding people, with limited communication between them. Cell members do not have a broad view of what is occurring in the crowd. A dominant cell member may influence the collective behavior of the cell. Chains of cell to cell communication can occur, often with the spread of rumors and incorrect information, potentially inciting inappropriate behavior.” (Fruin 2002, p.4)
Panic arising during a financial crisis is one form of “inappropriate behaviour” resulting from market interactions. Albeit a rare phenomenon panic is probably the most dreaded and the most devastating crisis element. The awareness of the possibility of crashes, bank runs and contagion seems to paralyse authorities in their search for a crisis solution and, at the same time, makes them susceptible to doomsayers, industry pressures and thinly disguised blackmail of stakeholders.
There are several ways to cope with the phenomenon. One is the proposal of Andrew Haldane and others to make rules for financial regulation simpler and to limit certain financial activities. As I pointed out, this may well help reduce regulators’ uncertainties and render banks more robust to withstand a crisis. But whether it may contribute to preventing crisis outbreak and containing panic must be doubted. There is simply not enough information about the functioning of the complex world of financial markets to know.
Another approach is to enhance our understanding of complexity to better get to grips with a complex environment. In principle, the challenges of studying a complex financial system do not differ fundamentally from those of analyzing the spread of diseases or climate change. These days, in each of these cases, agent-based computational models are used to convey a feeling for the interactions and feedbacks involved. Even if the models do not come up with particularly reliable numerical results they demonstrate the scope and limits of policy interventions under complexity and, in a sense, allow to develop computational ‘wind tunnels’, as Mark Buchanan put it, enabling regulators to test policies before putting them into practice.
Similar to climate systems, however, the use of agent-based models in economics is limited. Complex systems, which are characterised by a high sensitivity to initial conditions, defy generalization and reduction to causal relations which hold once and for all. Researchers easily succumb to the temptation to try to grasp complexity by building ever larger models including more and more actors. But, for an emergent system this misses the point. Studying more fish gives no clue to the behaviour of the swarm. The parts allow no conclusion on the whole.
Then, what else can be done to cope with the “madness of crowds” in a financial crisis? As in the case of floods, storms and pandemics, in a financial crisis, authorities are called upon not to stand helplessly watching a catastrophe unfold but to think about prevention policies and measures to mitigate the effects once crisis struck. Again, the analogy to another discipline shows the way, this time to sociology:
Sociologists studying the phenomenon of panic find similarities between situations where physical danger is present, such as people trapped in a burning theatre, and cases where the danger is to people’s financial assets (see explicitly Michael Klausner in Adler and Adler 1984). Then the question is which lessons can be learned from panic and disaster research and practice.
“Most major crowd disasters can be prevented by simple crowd management strategies.”
Does this statement by John J. Fruin hold for financial panics, too? At first view, the idea may seem far-fetched. The situations Fruin refers to include sporting events, musical festivals, riots and religious processions.
This is not the place to develop a research programme on financial disaster prevention and control. But, there are parallels which make it worth to take a closer look at the concepts and principles of crowd management.
The following table lists some similarities and differences. By definition, in physical systems crowds are large concentrations of people. In contrast, in a financial system, no physical closeness is required. Instead, the crowd is the number of interconnected actors who propagate rumours and information and react to the attitudes and decisions of others.
Under physical proximity, panic is “a reaction based on an internal assessment, that the probability to influence one self’s survival in a life-threatening situation is close to zero” (Pajonk and Dombrowsky 2006). In analogy, panic in financial markets can be defined as a reaction based on an internal assessment that in an adverse market or system development the probability to influence one self’s financial survival in a situation threatening material existence is close to zero.
In both cases, the reaction will be flight – under physical proximity from a place, and in the financial scenario out of a market or investment. As a result of the seemingly aligned reactions of the many, physical and/or psychological pressures build up, there is an individual loss of control, “the” swarm/crowd/market takes over, and people get hurt in the one case, and experience heavy financial losses in the other.
In order to characterise the different elements involved, John J. Fruin developed his FIST model of crowd disasters. Again, parallels can be found. FIST stands for:
F: Force, which is the levels crowd forces can reach due to pushing and to the domino effect of people leaning against each other, that are “almost impossible to resist or control”. (Fruin mentions more than 4500 N in this context.)
Examples of a financial market analogy would be the volume of sell orders hitting a market under panic or the size of withdrawals of deposits in a bank run and the resulting pressures.
I: Information, including all means of communication (1) between actors, (2) of actors with authorities and (3) in interaction with their environment. Information is an important element of both the spreading of panic and its prevention or containment.
S: stands for space in the Fruin framework, referring to the “configuration, capacity, and traffic processing capabilities of assembly facilities” determining the degree of crowding.
At first glance, there is no analogy to space with regard to the “footloose”, space-transcending financial industry. But, aren’t there parallels between the strategy to separate rivaling groups in a football stadium (Frosdick and Chalmers 2005, p. 131) and the debate to breakup banks and separate financial functions in order to limit contagion? And how about efforts to keep financial traffic flowing during a crisis and prevent bottlenecks? Therefore, the analogy here is defined as structures and institutional characteristics determining the degree or intenseness of financial activities and interaction.
T: Time is another crucial element in both physical and financial crowd management. To cite Fruin again:
“A simple illustration of timing is the more gradual and lighter density arrival process before an event, compared to the rapid egress and heavy crowd densities after an event.” (Fruin 2002, p. 5)
In financial markets, many instances can be found where timing decides about success or failure of panic containment. Think of the urge to come up with a policy decision before markets open in another part of the world. Or, of the “can kicking” of policy in the early stages of the euro crisis delaying decisions thereby giving key financial players time to adjust to the changing environment.
So far the theory.
In practice, there is a need to distinguish between systematic planning before an event, and crowd control, i.e. the restriction or limitation of group behaviour, in a state of emergency. As Pajonk and Dombrowsky 2006 write:
“Interventionen nach Ausbruch einer Panik haben so gut wie keine, präventive Maßnahmen vor extremen Belastungssituationen haben dagegen hohe Erfolgschancen.“ (Interventions after panic breakout have almost no prospect of success, preventive measures before extreme events have a high chance.)
Usually, in financial crisis management, policy decisions are guided – and restricted – by the tools of traditional economics which are implying: Find the source of general disequilibrium which is destabilizing markets and eliminate it. For example, adjust demand and supply of goods and services, raise or cut wages, stimulate investments and exports, restrict imports and capital flows, and so forth. All this may be good and useful to improve an economic situation and market sentiment in “normal” times. It is of no help in preventing or containing panic.
The list, which is not exhaustive, illustrates some key requirements which hold in general:
Information – including monitoring and communication at all stages of the process and all levels of planning, decision making, organisation and execution.
Training – ensuring the availability of crowd management personnel familiar with the situation.
Clarity – of expression, of information provided, of signals and directions given, of paths shown.
Sincerity – staying with the truth and building and maintaining confidence in all circumstances.
These four aspects apply to financial panic management, too. Information is needed, not about macroeconomic instabilities and disequilibria as crisis sources, but about market peculiarities, groups of actors and financial developments. As in the case of floods, storms and pandemics, panic management in this area calls for the efforts and interplay of many actors. The advice of, and close collaboration with, finance experts and crowd management specialists must be sought for, together with that of psychologists, economists and sociologists. Furthermore, monitoring, information exchange and collaboration must take place as an ongoing task and not only in a crisis. Developing a concept for an active financial crowd management, including a permanent institutional framework or kind of interdisciplinary emergency task force, is of utmost importance. One reason for the inability of policy to deal with recent crises, and the tendency to postpone and delay decisions, seems the total lack of rules for preventing a breakdown and preparing for worst-case scenarios.
Training is needed. Concepts of disaster prevention emphasise the importance of staff on site. In the present case, staff should have knowledge of crowd management as well as financial expertise. In this regard, collaboration, training and direct involvent of market participants may be an option. Major financial markets are highly concentrated. As I described elsewhere (taking the example of the foreign exchange market), there are cases were only a handful of big players have a dominant influence. In other cases, the challenge may be to identify the “dominant cell member (that) may influence the collective behavior of the cell” cited in the beginning and involve it in a concept of active crowd management.
Clarity is a weak point of financial crisis management in almost every respect. The cacophony of voices catching the headlines is deplorable – not only in Europe where federal structures and a respective large number of decision makers and stakeholders aggravate the problem. Uncertainties about crisis causes and remedies add to the general confusion, and quarreling experts, with a solid training in macroeconomics and a focus that differs from the one needed here, do more harm than good.
Sincerity is an even weaker point. To give an example: A strategy to play for time and postpone and delay necessary decisions only makes sense when there is a real chance for the situation to calm down. Otherwise, every new wave of market unrest can be expected to result in a greater loss of credibility, further increase of overall uncertainty and a higher probability of panic.
Instead of standing and watching a crisis unfold, only to finance costly bailouts afterwards, policy would be well advised to invest in preventive panic management. With the waning threat of crashes, runs and contagion it would regain a long lost room for manoeuvre, and the decision for or against rescue actions could be based on more rational arguments.