The corpus of ideas around ‘radical complexity’ has strong potential in economics, according to research by Jean-Philippe Bouchaud, to be presented at the Royal Economic Society's annual conference at the University of Warwick in April 2019. He says that it should help moving beyond the current framework of macroeconomics where rational agents act in a knowable world to create crisis-less economies.
Bouchaud notes that our understanding of so-called ‘complex systems’ has made giant leaps forward in the last 40 years, due in particular to a flurry of activity in statistical physics, theoretical ecology, computer science and mathematics.
Some of the concepts that have emerged should no doubt play an important role in economics as well, where one of the main issues is the aggregate properties of a large number of partially informed, heterogeneous and interacting agents. These ingredients are essentially the same as those present in the ‘complex systems’ studied by physicists.
One of the most important concepts that have emerged from these studies is that some form of radical uncertainty, as envisaged by John Maynard Keynes, is present in these systems. Even when such systems are perfectly known at the micro-level, their macro-behaviour is unpredictable, even in a statistical sense.
The probability distribution describing their emergent, aggregate properties suffers from a kind of ‘butterfly effect’, well known for chaotic systems. Whereas in the latter case trajectories are sensitively dependent on perturbations, it is the probability of these trajectories itself that is sensitively dependent on perturbations, for generic complex systems.
We may know perfectly well what all agents should optimise individually, but the result of this collective optimisation is uncomputable, and hence unknowable, because ‘too complex’.
Of particular interest is the notion of marginal stability (also known as ‘self-organised criticality’ or SOC). As a rule, complex optimisation leads to fragilities and instabilities. Unpredictable innovations at the macro-level can arise in a fully determined micro-world. Radical uncertainty derives from radical complexity.
In his presentation at the RES conference, Bouchaud will discuss several examples of these general concepts, drawn from statistical physics, ecology, game theory and network economics.
In the simplest case, a system with two equilibrium states (say high output/low output), crises occur with a probability that depends exponentially on the parameters of the model. Hence, even the probability of crises and the price of the corresponding insurance is unknowable – extreme risks cannot be hedged.
A more elaborate example concerns the stability of network economies. Why is the output of large economies so volatile, when fluctuations should average out quickly as the size of the economy grows?
In the 1990s, physicists Bak and Chen and economists Scheinkman and Woodford surmised that large economies spontaneously evolve towards an anomalously fragile marginally stable state of the type described above. But in the absence of a convincing model, this SOC scenario failed to gain acceptance in the economics community.
Research by Bouchaud and colleagues has introduced a general model where SOC occurs naturally, with idiosyncratic shocks being propagated and amplified along a network of interconnected firms.
Using results obtained in the context of disordered solids, they find increased interlinkages, profit maximisation and/or reduced substitutability between firms' inputs drive the system to the edge of stability, as does adding firms to the system.
This scenario is strongly reminiscent of similar ideas in an ecological context, where the disappearance of a single species can lead to mass extinctions mediated by network effects.
“Radical Complexity” by Jean-Philippe Bouchaud
