Searching for Causes of Financial Instability

Blake LeBaron

International Business School, Brandeis University

May, 2010

The large price gyrations in U.S. equity markets on Thursday, May 6th, were unprecedented, and very interesting. Almost as interesting has been the response of policy makers, and the public. There is a need to find a single human, or maybe machine (human programmed), smoking gun behind the extreme price gyrations that day. The initial rumors about a "fat finger" that input orders is a perfect example of this. We would just love to find some poor individual who simply punched in a few powers of ten incorrectly on the amount of their order. This explanation would have most everyone from Wall Street, to DC, to main street happy. It's easy, concise, and fits our usual desire to have very simple causal explanations. It connects to other simple mental models we like to build in our heads. It also reverberates with many of us as an oops sort of moment we might have in our jobs, but with hugely magnified ramifications. Unfortunately, with several weeks of analysis this explanation does not appear to be the answer. The search continues for a rogue trader, errant computer, or cyber terrorist. I'm afraid we may never find it.

When looking at Thursday's events from a complex systems perspective, one wonders whether this is simply part of the nature of this system that has evolved. There are multiple markets all connected through high speed electronic trading networks. Some of the strategies employed in the trading algorithms could be destabilizing in the right situations. Timing is crucial in many of these markets, and they operate asynchronously with no common control system. Basically, all the pieces are there for a very rich set of aggregate dynamics that are possible in any complex system. This includes sudden phase transitions from relatively stable dynamics into a region of unstable dynamics verging on system collapse. No simple single cause is necessary in this case. The system is capable of doing this on its own. In a nonlinear system of this type it is possible that just the right sequence of order flow might trigger an unstable avalanche of selling or buying activity. For example it could, in theory, be as simple as the appropriate sequence of buy and sell orders. Like opening a safe, in the rare case where the correct numbers come in, the system can be sent into a state with dynamics that are completely different from normal system behavior. No smoking gun would ever be found.

It seems a little unsettling to throw up ones arms and say, we just don't understand these systems. I think we will eventually see patterns. One is that liquidity providers, both human and machine, tend to have similar and important stabilizing strategies. Their objective is to assess what they view as the current true valuation for a security, and they will sell or buy as trading moves above or below this price point. Their problem is that no one knows what this value is. As market volatility, or general uncertainty, increases they may back off on their position, or disengage completely from the market. An example of this would be machine traders hitting the off switch on their trading systems, or in 1987 NASDAQ dealers taking their phones of the hook. This leaves a market dominated by trend following, or momentum traders. If left to their own devices they have to be destabilizing, and will drive the price to extremes. This basic interaction between stabilizing and destabilizing strategies is key to understanding market dynamics.

Without a better understanding of the dynamics of these large asynchronous markets with multiple trading platforms we will often be left looking for smoking guns when there are none to be found. Hopefully, in the future we may have better tools to add to this discussion. It will be interesting if policy makers and the general public find the answers from these models acceptable or not.