(Funny, my wife made this point months ago.)
Duncan Watts, principal research scientist at Yahoo Research and expert on human networks and complexity, makes this point in the June Harvard Business Review (”Too Big To Fail? How About Too Big To Exist?“).
Watts looks at the financial market as a complex system and compares it to another complex system: the power grid. As an outage in one power plant can cascade and cause outages regionally, so the financial system failures (such as the collapse of Lehman Brothers) cascaded to a general meltdown in credit and prompted unprecedented governmental intervention.
He points out that in a complex system, the actors (financial firms, power system components) affect each other to the point that one’s own risk profile can change dramatically depending on what happens to others. Meaning, your risk department’s calculations are dependent on assuming the other guy is stable and rational (risky assumptions those are).
Government coming in after a disaster and resuscitating the surviving firms is one approach. A better approach, according to Watts, is to make certain that each actor is small enough that its failure has a limited effect on the other actors.
This discussion reminded me of the evolution of robustness in computer systems in the past thirty years. In the 1980’s, the best way to achieve robustness was to build a huge computer with redundant components and very complex software. Such computers were protected in military-style data centers with concrete walls and fire suppression systems. In case a piece of the computer failed, the software helped the machine use other pieces to continue operating. Tandem (now part of HP) was the market leader here.
Of course, relying on one huge computer (too big to fail) exposed you to lots of other risks. For example, what if the power went off? What if there was a localized weather disaster? etc. There were limits to the “too big to fail” computer architecture–exposed most notably in the 9/11 disaster, where reliance on Lower Manhattan data centers put the stock markets and other financial markets on hold for days till their data services could be relocated.
Another approach to computer redundancy was created in the internet space, perfected by Google. Rather than having one or two huge servers with complex software managing redundant everything, Google has created a worldwide network of hundreds of thousands of small, pretty dumb servers, and software that allows transactions to be moved across those servers depending on their health. If a Google server goes down, nobody notices because its traffic is quickly spread over the remaining zillion servers that are working.
And that seems like a better model for our financial systems, too. I agree with Watts: too big to fail is too big.
On Duncan Watts’ “Big Seed Marketing” idea