Last Thursday, US equity market traders and other participants had an unusually exciting day, as the value of many stocks declined sharply, as reported in the New York Times. The value of the Dow-Jones Industrial Average fell more than 1000 points within about 15 minutes, although it later regained most of the lost ground, ending the day down 347.80, still a sizable decline, about 3.2%, but not a disaster. Many individual stocks temporarily lost huge chunks of their value: Proctor & Gamble, for example, showed a decline of more than 33% within minutes.
As is customary, there has been no shortage of explanations for this event. One suggestion is that it was initially triggered by an erroneous trade entry, perhaps a trader entering a ‘B’ (for billion) rather than an ‘M’ (for million), and was then amplified by an unspecified computer system problem. There may be some truth in this: I was working in New York City when the October 1987 crash occurred, and vividly remember the price of IBM stock, as displayed on my computer monitor, jumping around by $20 per share or more. In that case, the trade and price reporting system simply could not keep up with market activity, and more or less went crazy.
However, in the post mortem analysis of that 1987 crash, one “culprit” was identified: the use of automated trading strategies that were designed to provide “portfolio insurance”. These strategies are designed to gradually reduce a trader’s equity exposure as the market falls, so that his potential loss is limited to a pre-determined amount. (In essence, the strategy uses dynamic trading to replicate the payoff of an option.) The problem, of course, is that when too many people are trying to get out at the same time, the “door” may not be wide enough, causing prices to fall, which then requires those following the strategy to sell even more. This is a key reason why so-called “circuit breakers” were put in place on the New York Stock Exchange [NYSE] and other exchanges, to halt trading temporarily if a suspiciously large price change occurs, to provide time for human intervention.
A lot has changed since 1987, and automated trading, particularly very short-term trading strategies known as high-frequency trading, has become a very big business. It is estimated that this trend is largely responsible for the 164% increase in daily trading volume on the NYSE, as well as the rapid proliferation of alternative, computer-based “exchanges”. The time frames used in these strategies are in some cases so short (measured in milliseconds) that firms aggressively bid for computer locations physically close to the exchange’s data center: the network propagation delay (at the speed of light!) has to be taken into account.
This is obviously a game that is open only to well-heeled institutions; the average retail investor certainly cannot compete in this sort of contest. Defenders of the practice say that it increases liquidity. That is certainly true, in that trading volumes have undoubtedly increased significantly. However, it’s not clear to me that this benefits anyone other than the other high-frequency traders very much. If stocks could only be traded one day a year, then prices would be stale, and liquidity would be a problem. Going to weekly, or daily, trading would undoubtedly bring economic benefits. But it seems clear that this process is at some point subject to diminishing returns; the benefit from being able to execute a trade every 10 milliseconds, as opposed to every 1-2 seconds, seems open to question. I somehow doubt that the overall allocation of capital in the economy is greatly improved.
There is another practice associated with high-frequency trading on some exchanges that is even more suspect. This is so-called flash trading. (Note for geeks: this has nothing to do with Adobe’s Flash, Steve Jobs notwithstanding.) In this setup, certain privileged traders get, for a fee, an advanced look at the incoming orders to the exchange, before those orders are executed. This gives them the chance to gain an advantage when, for example, a large buy order has been submitted, by buying the stock in question before the large transaction tends to cause the price to rise. If a stock broker does this sort of thing for his own account based on his knowledge of customers’ orders, it is called front-running, and is against the law. The SEC has proposed banning flash trading, but no final action has been taken.
It seems to me that once again, we are seeing the consequences of failing to recognize what Charley Ellis called a Loser’s Game. These trading strategies are not irrational: for the early adopters, they have to potential to make a good deal of money. But if most of the market participants are trying to do the same thing, much of their value is eroded; and they were never a net positive for the overall economy to start with. The rules need to be set to prevent insiders from grabbing short-term profits from this kind of activity, while leaving someone else (like the US taxpayer) holding the bag when things fall apart.