Time for Regulators to Impose Order in the Markets

If your machine makes a mistake that the dumbest human would never make, then maybe you don’t have a very good machine.

The regulators are still trying to figure out just what set off the crazy trading a week ago Thursday, but some facts are obvious. If a stock goes from $40 to one cent to $40 within a few minutes, somebody messed up.

Getty Images

Or, it appears, some thing.

As Mary Schapiro, the chairwoman of the Securities and Exchange Commission, said in Congressional testimony, “Automated trading systems will follow their coded logic regardless of outcome, while human involvement likely would have prevented these orders from executing at absurd prices.”

The computers that now dominate trading are amazing. Their speed of trading is measured in milliseconds. They can post and withdraw an order faster than you can read this word. If a fast computer is on the other side, then a trade can be made before any person could even notice a trade was possible.

That trade can then cause other traders’ computers to determine that momentum has shifted, and trigger huge orders to sell at the market price. The orders can also cause some brokerage firms to back away from making markets, which they do by posting a “stub quote,” such as a bid of 1 cent for a stock.

As the regulators try to work their way out of this, there are several things to keep in mind:

1. Markets are utilities for investors. We used to think that utilities are natural monopolies, which should be regulated rather than competed with. Changing that view, and allowing lots of new markets that could compete with one another on the basis of cost and speed, required much better regulation, but it did not happen.

2. One of the most important functions of markets is price discovery. For most major stocks in the United States, that role used to be performed by the New York Stock Exchange, which, despite a lot of automation, remained under the dominance of actual people. An important goal of the new regulation must be to assure price discovery functions well.

3. Don’t blame the investors. There is apparently no evidence of a “fat finger” trader who ordered a sale of billions of shares, when he meant millions. But even if there were, the systems should have prevented chaos. They did not.

4. To have a functioning market, someone must provide liquidity. Maybe we need to go back to paying people and brokerage firms to do that.

In the old days — whether they were good or bad depends on your perspective — nothing like the May 6 follies could have happened to a stock listed on the New York Stock Exchange. Confronted with a large order imbalance, the Big Board halted trading and tried to sort it out. That might not prevent a stock from plunging, of course. But the delay gave time for cooler heads to leap in if they wished to do so.

In those days, the exchange had a monopoly on trading in stocks it listed. So it could delay without costing it any business.

Also in those days, commission costs were far higher than they are now. Big brokerage firms made markets in stocks for institutional investors, and charged them for it through commissions and, if the order was large or hard to place, through price discounts.

An institution that wanted to unload a large block of stock would call one of a handful of brokerage firms. If it wanted to get out fast, it knew that Goldman Sachs or Salomon Brothers or Morgan Stanley would agree to buy it without lining up customers to take the stock.

How big a discount the broker demanded would depend both on its assessment of the market and on the institution’s reputation. If an institution was known for sandbagging brokers, perhaps by selling on news that was not yet public, or by selling blocks simultaneously to several brokers, it would find that subsequent sales were much harder to pull off.

Now we have competition between exchanges and computers with programs aimed at allowing an institution to parcel out stock in small blocks, so no one knows it is unloading. Liquidity is largely provided by so-called high-frequency traders, who have their own superfast computers and models that are supposed to allow them to get in and out rapidly, making a small profit on almost every trade.

Costs of trading have collapsed, and volume has soared. Ms. Schapiro pointed out that the 1987 stock market crash produced trading of 600 million shares a day — a figure that seemed incredible at the time. On May 6, the number was 10.3 billion, much of it from those high-frequency traders who had no desire to own — or short — stock for any length of time.

Here is what appears to have happened on May 6.

First, prices fell for real reasons. But as the selling became worse, buyers became scarce. The Big Board started to invoke what it calls “liquidity replenishment points,” or L.R.P.’s. Those are not trading halts, as in the old days. The exchange could never hope to get away with them these days. But they are pauses, usually for a few seconds, intended to “facilitate more accurate price discovery and prevent the market from a sudden and significant move,” as Larry Leibowitz, the exchange’s chief operating officer, testified.

Such pauses, he said, “are analogous to taking the controls of a plane off auto-pilot during turbulence.”

Who knew what they were doing?

Normally, such L.R.P.’s take place without anyone much noticing. But on May 6, there were many more of them than usual, and some lasted more than a minute. During that time other markets could see the Big Board bid, but they knew it might change. They could legally ignore it.

It seems likely that many of the high-frequency traders had programs that caused them to back away from making markets. Liquidity dried up just as stop-loss orders triggered more sales.

Other markets had no similar mechanism to slow things down. Those markets competed on the basis of speed.

“Orders in one stock directed to one market can now ricochet to other markets and trigger algorithmic executions in other stocks and derivatives in milliseconds,” Ms. Schapiro said.

Algorithms have been all the rage with institutions. A simple one might call for an order to sell a million shares of a stock to be executed in small blocks for as long as it took, providing that the total sold by that institution did not exceed a certain percentage of the volume.

So as more trades took place and drove the stock down, the algorithm would call for increasing the pace of selling. There was nothing in the algorithm to stop if the stock fell 50 percent in a few minutes.

None of that is news, by the way. In 2008, some Nasdaq stocks — among them Cisco and Charles Schwab — sold for a penny a share. Those trades were canceled with little public notice, and the S.E.C. did nothing to prevent future occurrences.

What should happen now?

First, the regulators need to impose order. At the least, they should make it much riskier for an exchange to route an order away from the New York Exchange when it has imposed a slow-trading delay. The other exchange would be required to have a person decide that was reasonable, and he or she would be subject to regulatory second-guessing. Or the commission could just halt all trading while the Big Board had an L.R.P. in place. Other exchanges that have listings, such as Nasdaq, would be required to develop similar mechanisms.

Second, the S.E.C. should consider whether Nasdaq needs to return to a system where brokers are required to make markets in stocks, as are Big Board “designated market makers,” which is what specialists are now called. Nasdaq used to require market makers to make markets. But as high-frequency traders came to dominate the market it relaxed the rules, allowing the “stub quotes” that let the market maker back away.

Third, the S.E.C. should return to the ideals of a “national market system.” The best argument for “dark pools,” in which some trades are executed without ever being exposed to the market, was that it was an innovative idea that might save a little money. That innovation now looks to be every bit as good an idea as the concept of synthetic collateralized debt obligations.

Finally, the S.E.C. needs to investigate whether brokers met their “best execution” responsibilities, and brokers need to compensate a lot of customers. (Whether the electronic exchanges should then compensate the brokers is an issue that could also be explored.)

The trades that were canceled were those in which the price was down at least 60 percent. But there were plenty of other stocks whose prices fell 30 percent because the market failed. If a brokerage firm sent an order to an exchange that executed it at an absurd price, then that firm did not live up to its obligations.

Granted, deciding the cutoff from absurd to not absurd will be arbitrary. But real investors deserve protection. Reputable brokers should make restitution without being ordered to do so. Other brokers should face S.E.C. enforcement actions. Wouldn’t it be nice if brokers started to announce voluntary restitution programs?

A retail investor who had put in a “stop loss” order to sell Procter & Gamble , but only if it fell below $50, had trusted no sale would happen unless a functioning market decided the company had lost that much value. It did not. Even at the worst of the panic, Procter traded at $56 or above on the Big Board. But on the electronic markets, it traded as low as $39.37.

It would be unfair to those who stepped up to buy to cancel any more trades. But such a retail investor deserves compensation from the brokerage firm that mishandled his or her order.

On the other hand, institutions and other traders that used their own algorithms, or that knowingly used brokerage-house algorithms that had no protection if markets went crazy, can suffer the consequences. They should have know what they were doing, even if they did not.