Over the past three decades, the global financial system has become more dynamic and interconnected, more concentrated and complicated than ever before. Financial engineering seems to know no limits to creating new instruments that link institutions in new ways.
Is that a good thing? Or could the resulting financial network be too complex? Or, perhaps, complex in the wrong way?
A look at biology - which has been tinkering with network designs for billions of years - suggests the answer to the last question is most likely yes.
In "The Architecture of Complexity", an extraordinarily original paper published 50 years ago, the economist, psychologist and artificial-intelligence pioneer Herbert Simon asked the question: why does nature so consistently organise itself into hierarchies? Why, that is, are so many of its creations designed as systems of systems?
In biology, for example, cells organise into tissues, tissues into organs, organs into larger systems. The cell itself contains a nucleus and a cell membrane, ribosomes and mitochondria. Our human organisations obviously also follow hierarchies, as do our buildings, technological devices, even our writing - words make sentences, which build paragraphs, which then make up essays or chapters.
Scientists and philosophers since Aristotle have noted as much, but Simon, one of the most creative minds of the 20th century (he died in 2001), was perhaps the first to ask why. He also proposed an answer.
For one thing, he pointed out, structures such as this are easier to make and also more amenable to beneficial alteration. We might, in principle, build computers as enormously complex assemblies of billions of individual transistors, linked in some exquisite design. Then, however, every device would have to be built as a whole. We simplify construction by designing computers as assemblies of subunits that can be linked - a memory chip, central processor and keyboard, for example. The units can be built and tested separately, and they can be linked in different ways to make different kinds of computers. We can reach in and alter one component - changing the memory - without worrying that we have wrecked the keyboard. As a result, computers become easier to improve.
Hierarchy, in other words, is a way of limiting complexity in the interest of both stability and the ability to evolve. Simon argued that systems structured in this way possessed a basic, competitive simplicity.
We are only beginning to appreciate how much, as living beings, we rely on this architecture. Take ordinary bone, for example, which is remarkably tough, yet lightweight, with properties that our technology still cannot match. The secret is hierarchy. Within bone, small molecules bind together into proteins, which then link into filaments, which in turn organise into larger structures. When a bone suffers a blow, the hierarchy provides a variety of mechanisms by which it can pass along the excess energy it absorbs, without creating lasting damage. Bone, like most other structures in biology, is not just complex, but complex in a highly organised way.
What about structures in economics and finance? The growth of modern finance seems to have violated the principle of hierarchical structures, and with gusto. Two trends in the past 30 years - the merging of banks into huge institutions and the explosion of derivatives that link them around the globe - have made the network much less modular. We have created a vast web of interconnections with extreme complexity but little organisation. And this does appear to have made the system less resilient.
For example, in a study last year, economists from the University of Auckland, New Zealand and the Bank of England used computer simulations to explore how failures might cascade through the interbank network, the system that banks use to manage their day-to-day financing demands by making loans to one another. This network normally functions fairly well, with funds flowing easily, but it can experience sharp liquidity crises - as it did following the collapse of Lehman Brothers Holdings Inc in 2008. For short-term cash, banks rely heavily on "repos" - overnight sales of stocks or other assets, which they agree to repurchase later. How much cash a bank can get for a specific asset depends on the "haircut" - a reduction in the cash lent against it, which lenders demand to protect themselves against risks, or losses they may face if, in the case of default, they have to sell the asset themselves.
Haircuts fluctuate with time and perceptions, and the simulations show that the interbank network's resilience to such fluctuations depends on its architecture. The more the network is concentrated in and dominated by big banks and the higher the overall density of links among banks, the less modular the system is, and the less stable. That is, both these trends make it more likely for financial distress to cascade through the network.
Specifically, huge banks that account for a disproportionate share of all links act as potential epicentres for trouble. This is a way of describing "too big to fail", although it would be more accurate to say "too central to fail". Meanwhile, a high density of interconnections in the network creates ever more channels along which contagion can move. This problem encourages banks to "hoard" funds in times of stress - the least desirable behaviour in a network of banks trying to share resources to meet their momentary funding demands. Unlike organisms, of course, financial systems have not undergone evolutionary competition from which only the fit have emerged. We have little reason to expect that what exists would be anything like optimal, or even reasonable.
To counter these developments, we could try to manage the way lending occurs - control the amount of leverage used and the haircuts involved - so as to prevent dangerous contagion. More boldly, we might try to set up constraints on the very concentration of our networks, on who is linked with whom and how strongly.
Both high concentration and high interconnectedness contribute to an "everything is linked to everything" outcome that is the very opposite of modularity, and a likely recipe for instability. Financial engineering should learn to avoid this architecture, just as surely as biology has.