!VAR HAGENDOORN

Financial Markets, Networks and Complexity

I have always been a sucker for novel analytical perspectives from outside the mainstream. When I was at university I did a course on chaos theory and neural networks in finance. This was long before high frequency data became available so we didn't find any signs of chaos and our neural network simulations were pretty feeble. Some ten years ago I read some papers about financial markets as a complex system coming out of the Santa Fe Institute. The problem is that it never got far beyond the observation that financial markets are complex adaptative systems. Yeah, well obviously.

So when I came across some papers which apply network analysis and a form of actor network theory to finance I was intrigued enough to be briefly distracted from what I should be working on and have a glance.

Financial markets form an intricate network of interrelations. Networks have empirical properties which can be analyzed. Some networks are resilient to small disruptions but vulnerable to big disruptions, other networks are vulnerable to small disruptions if they can spread through the entire network. This is what the mathematics of networks tells us (some years ago I wrote a long essay review about some popular books on network analysis, but it's in Dutch). So the question is: what are the network properties of different financial markets? And is there anything we can learn from it? Since these are initial approaches they merely confirm by alternative means what is already known, that some banks are systemically important etc. The conclusion at this moment appears to be that "more research is needed" (as always). However, I thought the picture was quite interesting.

The mathematical analysis of networks leaves out the (behavioural) mechanisms through which the nodes are connected and through which information flows. It merely observes that if let's say a node is removed from a network, the connections change. It does not say anything about the decisions people make that cause the change in connections.

Actor network theory as a sociological anti-methodology takes a completely different approach to networks. I thought I'd add "anti" because otherwise I'll receive dozens of emails from angry sociologists informing me that ANT is not a methodology. Yes, I know...

Decisions are made by people. Individuals have cognitive, emotional and cultural biases. People act on information, which can be incomplete and on knowledge, which can be imperfect. They organize themselves into groups which have their own dynamics, influence individuals within the group and interact with other groups. The interaction between individuals, between individuals and groups and between groups is governed by principles.

People don't just make decisions, they also consume, produce and design products. These products have their own laws, dynamics and so on. There is a difference between oil and oil futures and between plain vanilla and OTC markets. This is because the markets are organized differently, because the products are different and because the players are different.

Economics created a series of concepts such as utility, value, labour, capital, interest, consumption and production, which made it possible to focus on a specific type of action and interaction without going into the details of the person, product, organization etc. In physics and chemistry scientists don't worry about the legal ownership of the substances they study either.

Economic sociologists have now begun studying how people in financial organizations actually make decisions and how information actually flows between people, departments, organizations etc. They therefore study the same networks, but from a different perspective. Donald MacKenzie has been at the forefront of this research and recently wrote another op-ed article for the Financial Times (or pdf-version here).

I think the two types of analysis are and should be complementary. The picture shows different network densities at different moments of the day. The sociologist might then ask what this means for the people making decisions, who is in the office, whether work piles up, whether the decision process is different. And now I should be going back to what I should be working on...

Links

The ECB organized a workshop on Recent advances in modelling systemic risk using network analysis. The proceedings of the workshop were published in January 2010.

The topology of the Federal Funds Market. ECB working paper No. 986, December 2008.

On the Informational Properties of Trading Networks. Lada Adamic et al. 2009.

Networks in Finance. Franklin Allen and Ana Babus . Wharton Financial Institutions Center Working Paper No. 08-07.