In Markets’ Tuned-Up Machinery, Stubborn Ghosts Remain

A generation ago, when the stock market crashed on Oct. 19, 1987, the Nasdaq stock market appeared to have done much better than the New York Stock Exchange. While the Dow Jones industrial average fell 23 percent that day, the Nasdaq composite index was off just 11 percent.

It was not, it turned out, that Nasdaq stocks were more highly regarded. It was, instead, a question of the technology used.

At the New York Exchange, trading was still largely done by people, in person. At Nasdaq, trades were done by phone. The difference was that the market makers at the Big Board could not escape dealing with the flood of sale orders. But many Nasdaq market makers could, and did, decide not to answer their phones.

The Nasdaq market appeared to be operating, even though it really wasn’t.

Markets are far more automated and far more fragmented today. That makes trading much faster and — when everything works — more efficient.

And when it doesn’t? Chaos can briefly emerge while people try to figure out what went wrong with the computers.

That has been demonstrated twice this week. In the most publicized failure, the Nasdaq market found itself unable to put out stock quotes and halted trading in all its listed stocks for more than three hours on Thursday.

Most such problems last only a little while. On Monday morning, options markets were briefly roiled by a computer error at Goldman Sachs, which caused it to send out ridiculous trade orders for options on stocks whose ticker symbols began with the letters I, J or K.

In a less publicized problem, soon before the Nasdaq market had to be shut down on Thursday, the Arca electronic exchange, operated by the same company that runs the New York Exchange, NYSE Euronext, was unable to report trades on Nasdaq stocks whose ticker symbols came after TACT in the alphabet. It canceled some orders during the period, which lasted less than nine minutes. Other markets routed orders away from Arca.

Just what went wrong in each case will be sorted out eventually, as were the technical failures that caused the “flash crash” on May 6, 2010, when some stocks fell to $1, and the disastrous first day of trading in Facebook shares on May 18, 2012, when Nasdaq’s computers were overwhelmed. It was just more than a year ago that Knight Capital, a large market maker in Nasdaq stocks, suffered significant losses when its computer system caused numerous incorrect trade orders to be submitted.

Why is this happening, and happening so often?

One reason is the need for speed. Another is increased competition.

The need for speed comes from a market in which high-frequency traders expect to be able to get in and out of positions within a second. Any market that cannot offer such speed will be at a competitive disadvantage. But such speed is not compatible with safety features that could cause suspicious orders to be delayed while someone — a slow person, perhaps — checked to see whether something was amiss.

The competition comes from the fact that there are now numerous exchanges for every stock. That has caused the cost of trading to plunge. But it has also meant that each exchange is under pressure to keep costs to a minimum, which itself could be a deterrent to safety features.

While Nasdaq’s failure on Thursday appears to be one of the most significant technology problems to strike the markets, it was less important than the earlier errors in one key way.

When Nasdaq determined it was unable to distribute quotes on all stocks listed on its exchange, it asked that other markets that trade Nasdaq stocks also halt trading, and they did. As a result, no one could trade. Traders who have grown used to the idea they can get in and out of positions quickly were frustrated, and Nasdaq suffered another humiliation. Investors curious about the market reaction to specific news events had to wait. But it does not appear that any bad trades were made.

In the earlier Arca problem, it is possible that some trades that had been sent to the market were not executed, and that as a result someone missed a brief market opportunity to send the order somewhere else. But, as with Nasdaq, there do not appear to be any trades that someone will want to cancel.

That was not true on Monday, during the period when Goldman’s computers were spewing out mistaken orders, just as it was not a year ago when the Knight computers went haywire.

When that happens, exchanges have to decide which orders to cancel, and they have developed rules about just how ridiculous a trade has to be to justify canceling it.

Myron Scholes, who shared a Nobel Prize for developing the Black-Scholes options pricing model, told the Financial Times this week that no options trades should be canceled. If Goldman and other firms “internalized all of the losses associated with program errors and bad algorithms, they would be more careful,” he said.

That might work in options, where Goldman’s errors appear to have led Goldman itself to make bad trades. But in other cases, it would create its own injustices. During the “flash crash,” erroneous sell orders caused some $30 stocks to fall to $1. Innocent individual investors who had put in orders to sell shares at the market price lost money when their trades were executed. Others who had put in “stop loss” orders, to sell a stock if it fell below a given price, found they had similarly suffered.

None of that would have happened back in 1987. Then there were people involved, and if there was a crescendo of sell orders they would not all get executed at lower and lower prices. On the Big Board, where the specialists could not avoid the orders, they could — and in some cases did — halt trading to sort things out.

Those halts, as it happened, helped to end the crash. It had largely been caused by foolish institutional investors using something called “portfolio insurance,” which required them to sell stock index futures when stock prices fell, and to sell more when prices fell further. The market makers who bought the futures then tried to hedge by selling stocks, which drove prices down further, and so on and so on.

When the Big Board began to halt a lot of stocks on Oct. 20, the Chicago futures exchanges threatened to halt trading in index futures contracts. Only then did Goldman and Salomon Brothers, the two largest brokerage firms serving institutional investors at the time, step in and offer to put up capital to reopen trading in stocks. Prices began to recover.

Now, in most cases exchanges are not willing to halt trading just because prices have defied all logic. They figure that would send business to their competitors.

What distinguishes Thursday’s Nasdaq mishap from other recent computerized trading malfunctions is that it involved the dissemination of prices, not the submission of bad orders. With no prices available, trading had to halt. Brokers and exchanges lost business, but it does not appear that anyone lost money from making bad trades.

In that sense, it was a return to what was good about 1987. When things were out of control, markets could stop until sanity returned. There’s no guarantee that will happen in the future.