The imperceptible but inevitable emergence of crises

This short paper argues that economies cannot be simply viewed as the sum of well behaved but isolated individual decision makers. Even if we accept the notion of an equilibrium state, thinking about the economy as a global dynamic system, it may be true that there are several such equilibrium states. It may be the case that small changes in the parameters of the system result in a switch from one equilibrium to another. We argue that the changes in the parameters or structure of the system are not exogenous but arise because people adopt rules which become the norms around them. For example, if a rule is adopted by the majority of one’s neighbours it may become acceptable or, alternatively if people see that changing their current rule leads them to obtain greater gains, they may adopt modified versions of their previous rules. However, as rules develop and spread they may have consequences at the aggregate level which are not anticipated by the individuals. Indeed, the emergence of new rules or the modification of old ones may fragilise the whole system which may then essentially cease to function. Finally we develop a simple model using some structure from statistical physics to show how this may happen. The standard vision of the economy remains one in which there is a static equilibrium state or where the economy is basically in a steady state to which it returns when perturbed by an external shock. If the economy deviates suddenly and considerably from such a state this can only be, in this view, due to the occurrence of some major exogenous shock. A first problem with this view is that many models of the economy or specific sectors are characterised by multiple equibria (for a comprehensive survey see Cooper (1999)). Although it has often been argued that one could hope for some sort of local stability which would prevent radical changes as a result of small modifications in the structure of the system, a seminal paper by Morris and Shin (2001) argued that the opposite may well be true. Furthermore it is widely asserted, but not so widely explained, that systems with such multiple equilibria may shift from one equilibrium to another. Small changes in underlying parameters, or small « shocks » may lead to such a transition. Moreover it may well be the case that such an evolution may not be simply reversible. Economists are familiar with the idea of a system which evolves along a surface and then may, as a result of a continuous change in variables slide suddenly on to a different surface. The applications of catastrophe theory and later chaos theory, to economics are illustrative of this, (see Rosser (2000)) A phase transition of this type does not put the economy temporarily off track in such a way that, with the appropriate measures, it can be brought back to its original path. If the changes mentioned happen, then there is no steady state or equilibrium path but rather an evolution through a set of states. The major question here is just where such shifts in some of the parameters in the model come from. As I have said, we have become accustomed in the macroeconomic literature to the idea that periodically there are exogenous shocks which temporarily knock the economy or market off course but that it eventually returns to the steady state path. This vision has led to the study of « impulse response functions » which are supposed to measure the consequence of the external shocks. However, none of this fits well with the accounts of the evolution of the recent crisis, where words such as « trust », « contagion » and « network » are commonly used. Each of these terms suggests that we have to consider the economy as an interactive system, which does not mean that markets cease to exist or be important, but that peoples’ behaviour in these markets is conditioned by the behaviour of others around them. This is what Shiller (2008) refers to as « social contagion ». Furthermore, this evolution does not require a class of « dishonest » individuals who break the existing rules or norms because doing so is to their advantage and they have fewer scruples than their honest counterparts. In such a case one might just reflect on more effective regulatory systems to control and punish such behaviour. What is, in fact at work, is the process by which what is acceptable changes, and it is this that underlies the collapse of such systems. In all of this there is no villain, no proximate cause, just a system evolving with a collective logic which does not correspond to that of the people who make it up. This sort of process may happen at different, but interlocking, levels.Consider, for example, the financial sector or more particularly the mortgage market which was at the origin of the recent crisis. How much consumers will be willing to borrow as a proportion of their income may evolve, and this may well be a function of what the majority of similarly placed individuals are doing. At the same time this will be catalysed by the changing willingness of banks to lend to such individuals. A small but gradual evolution in this direction can lead to major consequences at the aggregate level, as the system becomes more fragile. At another level there is the network of banks who lend to each other and across which risk is, in principle, diversified. Yet here, as we know, small change in the price of the underlying assets can lead through a process of contagion and diffusion to a freezing or even a collapse of the system. This would seem to be at odds with the claim that globalisation has increased the possibility of diversifying risk and therefore diminished the latter. The reason for this has been known for a considerable time and it is important to keep in mind that such « diversification » of risk is often illusory and depends on the underlying assumptions on the stochastic processes governing the evolution of the underlying assets. Even without correlation of risks if asset returns do not follow a Gaussian process diversification can yield to an increase in portfolio risk as Fama (1965) pointed out a long time ago. Already at the turn of the twentieth century, Poincaré (1905) explained clearly that, in his view, the work of Bachelier which is at the origin of the « efficient market hypothesis » could not be applied to financial markets, for, as he said the actors in such markets do not look in isolation at the information available to them but tend to infer information from the actions of the other actors in the market. This will lead to « information cascades », see Chamley (2005), and can generate price bubbles. This is one aspect of the type of contagion that can happen in financial markets but there are many other aspects, all of which are associated with the fact that it is unrealistic to envisage the actors as isolated and independent whether they be traders, consumers, firms or banks. What is central to this vision is that an economy or sectors of it can undergo radical changes and that this may be due to a number of factors, all of which can be subsumed under the idea of the economy as a complex adaptive system. Suppose that such a system does have stationary or equilibrium states or « basins of attraction ». Then the question arises as to how such a system might shift from one such state to another. Suppose moreover that individuals use rules of behavour and that these evolve in function of those used by those around them. This, rather than some external shock, is what causes such phase transitions. Such a vision does not correspond to an equilibrium model in the standard sense since the state space through which the system travels is changing as new rules develop. The idea of modifying rules or routines harks back to Nelson and Winter (1982) but in the illustrative model developed below it is formalised rather differently. The other analogy is to the analysis of repeated games by Lindgren (1991) who allowed for the constant evolution of strategies as the history of the game develops and showed that the system would settle temporarily on one strategy configuration but would have periods of high volatility as a newly emergent strategy temporarily took over the system. This sort of approach can, of course be applied to other social and economic systems but what is essential is the idea that the « myopic » vision of the individuals taking decisions can lead to catastrophic changes at the aggregate level. An example from financial markets. In the light of the current crisis it is worth looking at the financial sector in particular. In a recent paper, Rajan et al. (2008) develop a model of the financial system which has two very different equilibria which correspond to the degree of securitisation. They estimate a statistical default model from loans issued in a period with a low degree of securitization (1997–2000), using what they refer to as « hard information » variables about borrowers. They show that the statistical model which is widely used, underpredicts defaults on loans issued in a regime with high securitization (2001 onwards). The degree of underprediction progressively worsens as the securitization increases, and they observe that this must mean that at the same hard information characteristics, the set of borrowers receiving loans has worsened over time. Their argument is that lenders who are going to sell their loans on will not find it worth collecting « soft information » about borrowers. They therefore, expect the prediction errors to be particularly high when soft information is valuable; that is, for borrowers with low FICO scores and high LTV ratios. Indeed this is exactly what they find, a systematic variation in the prediction errors; they increase as the borrower’s FICO score falls and the LTV ratio