The Computer Glitch Felt Round the World

So this is what high-frequency trading looks like. Just before 3 p.m. today, the Dow Jones industrial average went from being down about 180 points, mostly off continuing fears over the Greek debt crisis, to falling off a 900-point cliff, to then rallying back another several hundred points. All in the span of about 20 minutes. During that span, share volume spiked to its highest levels in more than a year, to more than 4 million shares being traded in the Dow Jones alone.

Stocks in every sector mimicked the roller-coaster dip and climb. Shares of Accenture dived from $41 all the way down to 1 penny, and then back to around $35 in under five minutes. Shares of Proctor & Gamble crashed from about $60 down to $48 and then up past $60 in a few minutes. The big news surrounding Proctor & Gamble today: an investigation into high instances of diaper rash in Pampers. Painful, yes. But enough to cause a massive sell-off and then buy-back? No way.

As precarious as the situation is in Greece, it needed some help in prompting such a huge movement. The real culprit it seems, at least in the minds of many market analysts, is computer-driven trading algorithms that now account for more than 60 percent of all stock-market volume in the U.S. While high-frequency trading certainly brings efficiencies to equities market, it can also exaggerate things enormously. When you're dealing with such volume and speed, movements can be bigger and faster than predicted. Today's volatility is an interesting blip in the yearlong debate over whether high-frequency trading is a dark, sinister practice that needs to be reigned in, or a benign technological evolution.

Here's how it works: firms load complex trading algorithms into supercomputers, hook them up to stock exchanges, and then trade vast amounts of stocks at warp speeds, literally millions of shares in a matter of milliseconds. With each trade, the codes gobble up about one tenth of a penny. Eventually, a few billion pennies add up to serious money. A lot of people have been warning of a stock-market correction for weeks. And with supercomputers doing the trading, it can happen in the blink of an eye, faster and bigger than anyone ever expected. The one concern about HFT seems to have played out today: when everyone runs in the same direction at once, the market can take sudden, unprecedented swings. It should be a point in favor of those, many of them on Capitol Hill, who have been rattling for more regulation and transparency of HFT. The most adamant of them has been Sen. Ted Kaufman, who has cautioned of a similar market dislocation.

Of course, there are rumors that human error played a part here today. One rumor alleges what traders call the "fat-finger discount," where someone typed billion instead of million into a sell order. When you tell a computer to do something, no matter how irrational, it doesn't ask, "Are you sure?" It doesn't say, "That's insane." It simply executes the order.

A Chicago-based market observer and expert in high-frequency trading, tells NEWSWEEK that rumors are circling about a computer glitch hitting Citicorp's automated trading system this afternoon—that basically, it went haywire. "This is why quality control is so essential on these automated systems," says the market observer, who asked to remain anonymous due to his position in the market.

For more on Greece and other countries in the post-crisis poorhouse, check out our photo essay on how the financial meltdown is playing out around the world.