Wednesday, February 9, 2011

The SEC Tackling Computerized Trading

More than half of all stock market trading is now done by computers. It was inevitable that the SEC would bring enforcement cases involving computerized trading. Late last week, the agency imposed administrative sanctions on a money manager named AXA Rosenberg, for allegedly misleading its clients about a software malfunction in the firm's trading software. AXA Rosenberg managed a "quant" fund that invested based on computer analysis of a variety of factors. According to the SEC, the computer software's risk management process did not work properly, resulting in over $200 million of losses. When AXA Rosenberg personnel uncovered the problem, they did not disclose it to clients when they should have, and instead misled investors by claiming the losses were due to market volatility. Among other things, AXA Rosenberg will pay a $25 million penalty. More importantly, it's reportedly lost over half the money it had under management, as investors apparently headed for the exits after news of the software problem came out. That must have really hurt. Other quant funds will take notice.

Although the AXA Rosenberg case concerns today's elaborate computerized trading, it is a straightforward application of the federal securities laws. Investors were told that their money would be invested through the use of computer algorithms and other computerized analysis, and the law dictates that they must also be informed of risks presented by material defects and deficiencies in the programming. Disclosure of risks has been required since the beginnings of the federal securities laws, and as far as legal theory goes, the AXA Rosenberg case is as traditional as fireworks on the Fourth of July.

The SEC faces much bigger problems with computerized trading, as illustrated by the Flash Crash of May 6, 2010. That day, the Dow Jones Industrial Average dropped 9% in a matter of minutes, only to recover after a few more minutes. Apparently, a large computer-driven sell order by a money manager triggered other selling by computers monitoring the market, which soon led to wide-spread computerized stock dumping. This kind of high-speed chaos, like turning a corner on a highway and driving right into a sandstorm, scared the bejesus out of investors and still keeps droves of them away from the market in spite of the ongoing bull run.

Computerized trading is based on relative price movements: stocks are bought or sold when prices in the near term appear as if they are about to rise or fall. The software senses that a stock is comparatively cheap, and sends out a buy order. Or it senses that a stock is comparatively expensive, and sends out a sell or short sell order. Then, if and when the market moves the way anticipated by the software, the computer then closes out the trade by ordering a sell or buy, respectively. All of this happens very quickly, sometimes in milliseconds.

Trading based on perceptions of relative prices is nothing new. Day traders and other short term speculators have, for generations, tried to profit from relative price movements. Many money managers trying to beat market averages invest based on their perceptions of relative price. But computers have taken it to an entirely new level. With the ability to trade in milliseconds, computers can sense and profit from a price trend before humans have time to blink. Computers probably trade with other computers most of the time and the price movements they attempt to exploit may well be caused by other computerized trading. Sentient beings (i.e., humans) are simply left behind.

But the heart of the stock market isn't found in the upswings and downswings caused by relative price changes. It's in the overarching, long term gains (and losses) reflected in the valuations, perhaps seemingly subjective and imprecise, sentient humans place on stocks over the course of years and decades. Without sentient pricing, if you will, the stock market wouldn't exist. No one would risk their savings in a market where the only hope of profit would be relative price changes based on the short term inclinations of whoever or whatever else might happen to be in the market at the moment.

The real challenge for the SEC will be to preserve sentient pricing's fundamental role in the stock markets. High speed computerized trading to exploit relative price changes cannot take primacy over the human element in the stock market. There will be times when the agency may want to limit or slow down the participation of computerized trading. Such measures would find precedent in historical stock market limits on index arbitrage trading (like the New York Stock Exchange's collars and sidecars). Possibly, position or transactional limits on the size or amounts of some kinds of computerized trading might be necessary to prevent a single firm from smacking the market too hard one way or the other. Other measures, depending on the state of the art of computerized trading, may also be in order.

It's one thing for a computer to beat a human in chess, or even in Jeopardy. But when computers beat humans with the humans' retirement savings at risk, we have a horse of a different color. The "secondary" market, as the day-to-day stock market is called in Wall Street parlance, exists to support the capital formation process and not to serve as a speculators' mosh pit. An entire generation of investors fled stocks after the 1929 Crash, and it's no accident that stocks did not recover their losses from that crash until 1954, 25 years later. Ultimately, what counts is the absolute value of stocks, and human investors are needed to sustain absolute value.

No comments: