Experts examine systemic risk at Oxford Martin School conference

18 February 2014

Can full-scale system collapse occur even when individual components of the system are robust? The pernicious nature of these 'systemic risks' has attracted the attention of academics and industry alike, leading to a research collaboration between the Oxford Martin School, Oxford University's Future of Humanity Institute and Amlin plc, a major reinsurance company.

At the programme's inaugural conference on February 11 and 12, experts in ecology, complex systems, catastrophe modelling, psychology, and economics all shared different perspectives on how systemic risk can be understood through their respective fields. Lord Robert May discussed how the structure of connections between financial institutions mirrors that of ecosystems, and how modular and nested topologies tend to foster stability in both contexts. Professor Didier Sornette, who has established a 'financial crisis observatory' in Zurich, argued that complex systems provides some of the best tools since certain catastrophes can be predicted from superexponential movement of prices and a suspicious flattening of volatility.

One fundamental discussion came from the role of individual psychology. Cognitive biases and heuristics of individuals have been well documented in laboratory settings, but how do we apply these lessons to a complex set of human organisations? Should we take a top-down approach by producing predictive models of data, or should we take a bottom-up approach by modeling the individual agents in the system and examining the data it produces?

Professor Doyne Farmer, co-director of the complexity economics programme at the Institute for New Economic Thinking at the Oxford Martin School, explored the agent-based modelling approach, arguing that the technique allows one to predict how regulations can impact the market. For example, in some situations leverage ratio requirements can push traders to sell into a falling market, creating a destructive feedback loop. Although the model is not a perfect representation of real data, the insights from the agent based model come from the ability to understand why agents act in a certain way.

While the techniques for understanding these problems are diverse, so too are the problems themselves. Professor Ian Goldin, director of the Oxford Martin School, stressed this point in his opening keynote, arguing that the increased interconnectedness creates conditions ripe for pandemics, cyberattacks, market failure, and terrorism. But he also argued that despite these problems, we should also recognize that these connections are the basis for our opportunities in the 21st century. Preserving these opportunities while simultaneously navigating the risks will be our challenge for the years to come.