Banking as an ecosystem

Most of my interest in the use of biology in economics concerns humans being subject to the forces of selection like any other biological organism. With this starting point, it is natural to use many of the tools, models and methods of analysis that evolutionary biologists use.

But sometimes those models and tools are of value without the biological underpinnings. Evolutionary economics is one area where this is done, with the concepts of selection applied at the level of firms (as discussed in my last post).

Another instance of this crossover was in an article published by Andrew Haldane and Robert May, who have proposed that analysis of complexity and stability in ecosystems (dating from the 1970s) is useful in examining financial systems.

Haldane and May’s starting point was the recognition that complex ecosystems are not necessarily stable, with instability increasing with the number and strength of interactions. As an example, they noted recent work by Caccioli and colleagues which suggested that very strong fluctuations in the volume of trading in derivative markets could occur in a complex but complete market. As long as there is a positive premium to trading, banks will supply new financial instruments despite the lack of demand from non-banks. This expansion in derivatives comes at the cost of stability. There is no benefit to this expansion as market completeness has already been achieved.

Haldane and May developed a model which examined how banks may fail in response to a shock. In their model, each bank is linked to the same number of other banks and each has the same size of loans, capital reserves and ratio of loans to total assets. The more banks each bank is linked to reduces the number of failures following from the first bank failure, as the losses are spread more broadly. However, when later failures do occur, they will involve more banks. The model also showed the potential of small liquidity shocks to amplify through the system. Liquidity “hoarding” can have significant effects, as we saw in the recent crisis.

They also noted that their model reflected earlier work that had shown that as banks become increasingly homogeneous in their holdings (as they seek to cut their risks through diversification), the probability of the entire system collapsing increases. Once they are the same, the probability of one bank failing is the probability of all banks failing.

Haldane and May list a number of policy implications of their model. The first is that there is a broader role to minimum capital requirements for banks than simply reducing risk to each bank. Capital requirements could increase the entire system’s stability. Regulators should set capital limits with the broader systemic implications in mind.

A second implication concerned the goal of regulatory intervention. Typically, regulation might seek to reduce the probability of failure of all institutions to below a certain threshold. Haldane and May suggest that particular institutions that pose broad systemic risk should face higher regulatory requirements.

The most interesting suggestion concerned the desire to shape the topology of the financial system. As banks diversify, they became homogeneous. Accordingly, Haldane and May noted that a diversity objective of regulators may have merit. Trying to introduce “modularity” to prevent cascades through the entire system may also be desirable.

Nature published two responses to Haldane and May’s article: one in support of the use of such analysis by Thomas Lux, while Neil Johnson suggests that a model as simple as that used by Haldane and May will produce unreliable predictions that are only as robust as the assumptions used to prepare the model.

I do not have much sympathy for Johnson’s argument. While it is appropriate for models to contain health warnings about how broadly applicable the model is, models should by their nature have some simplicity. The question is whether any concepts are usefully illustrated. Haldane and May’s paper has several. Without a doubt, further work on these models by adding more elements and testing the robustness of the assumptions could be useful. That is often the way that science progresses. But to suggest that we cannot scale up a paper plane to a full-scale 747 does not mean that a paper plane can teach nothing about flight.

We may see more of these types of studies, or at least in Nature, as an editorial in the same issue suggested that where economic research has significant implications for fields such as behaviour, conservation biology, systems biology or physics, they would be happy to publish it. The editors suggested that this could benefit both economic science and natural science. My instinct is that economics has the most to gain.