BATS's aborted IPO on Friday, March 23, 2012, is an excellent example of the stupid side of computerized trading. While the precise course of events isn't entirely clear, BATS (short for Better Alternative Trading System) tried to initiate trading in its own stock around 10:45 a.m., at a price of $15.25 per share, or $0.75 per share below the IPO price of $16. Not a good sign when an IPO begins trading below the offering price. But that wasn't the really bad news. Rather, trading in BATS stock stopped moments later for technical reasons.
Around 10:48 a.m., BATS announced it was looking into "system issues" concerning stocks having symbols between A and BFZZZ. Nine minutes later (10:57 a.m.) Apple stock, which was trading in the high 590's, suddenly traded on BATS at $542.80 a share for a single 100 share transaction. This was a 9% drop from the previous price, even though there was no intervening news or other public event. Trading in Apple stock was halted. Five minutes later, it resumed back in the high 590's.
Around 11:07 a.m., BATS informed other exchanges that it was having problems and they need not send orders to it. Since competition for order flow is the essential dynamic between exchanges, this instruction was like making an unexpected pit stop in a NASCAR race.
At 11:14 a.m., BATS resumed trading in its own stock with a large trade at $15.25 per share. Within seconds, BATS shares were trading for pennies per share. Trading in BATS stock was halted again, and didn't resume. Near the end of the day, BATS announced that it was withdrawing its IPO.
In essence, we had flash crashes in Apple (which rebounded within minutes) and BATS (which cratered in the most embarrassing way on its IPO day). These morasses were apparently the result of computer glitches. Both were plain stupid, perfect examples of stupid is as stupid does. It's inconceivable that a human trader would have sold 100 shares of Apple at $542 on no news when the market was in the high 590's. There's no way a human trader would have sold BATS stock for pennies seconds after BATS traded at $15.25. That these trades happened is because a computer has no way to tell when it's doing something stupid. It can only follow the instructions in its program coding. And if the coding is deficient, the computer will blithely plunge forward however stupid the result.
Of course, human traders are eminently capable of stupidity, and are stupid more often than not. But humans trade much slower than computers. And few humans control as much order flow as computers at the large high-speed trading firms. So when a computer does something stupid, it can do it really fast and perhaps in really large quantities. No human is fast enough to stop the computerized stupidity before it happens. And no computer has been programmed to have the judgment and common sense to realize that it's about to do something really stupid.
Friday's morass didn't affect the larger market, which ended the day up. But it serves as a reminder that the baseline problem in computerized trading--how to prevent it from having unexpected and undesired consequences--is hardly any closer to solution than before the Flash Crash on May 6, 2010. Unless high speed trading firms can find a way to prevent stupid trades, their animal spirits will likely be curbed. And it's hard to argue against constraining them when they produce results as ridiculous as last Friday's.
Showing posts with label Flash Crash. Show all posts
Showing posts with label Flash Crash. Show all posts
Sunday, March 25, 2012
Sunday, October 3, 2010
What to Do About the Flash Crash?
The Oct. 1, 2010 joint SEC-CFTC report on the May 6, 2010 flash crash revealed that the market plunge was triggered by a single large sell order in a derivatives contract called the E-Mini, which is a futures contract that tracks the S&P 500 index. The story from the report, which is summarized here, is that the big sell order, placed by a mutual fund complex after the market had fallen about 4%, was for 75,000 contracts, having a value of around $4 billion. The mutual fund complex, unnamed in the report, wanted to hedge a large existing stock market position. It apparently hoped that selling the E-Mini contracts would protect its investors from further loss if the market continued to fall.
The mutual fund complex, which hasn't been officially identified although only those without Internet access can't figure out its probable identity, chose a automated computer execution process based on an algorithm that simply sent out the amount of orders it calculated to be 9% of the previous minute's trading volume. These orders were placed without regard price or time. That made the algorithm insensitive to the impact its orders might have on the market. It was an automaton that simply measured the preceding trading volume and pumped out 9% of that volume in new orders. If trading volume fell, the number of new sell orders the algorithm would send out would fall. If volume rose, so would quantity of new sell orders the algorithm issued.
The early sales from the 75,000 contract E-Mini sell order were purchased by other institutional investors, mostly high speed traders that don't hold their purchases for very long. As more sell orders came in from the algorithm, the early purchasers got nervous about the E-Mini contracts they'd already purchased and tried to sell them. Thus, they added to the trading volume generated by the algorithm's sales. The algorithm took this increased trading volume as a signal to place even more sell orders (since 9% of a larger number calls for more orders than 9% of a smaller number).
Buying interest, however, shrank as more and more E-Mini sell orders came on the market. Some high frequency traders began to buy and sell from each other, because there were fewer and fewer other buyers. But that only increased trading volume, which led the algorithm to dump yet more sell orders onto the market. This only made things worse.
Some of the traders buying E-Minis sold individual S&P 500 stocks to hedge themselves (i.e., to limit their risk from holding E-Minis). This put downward pressure on stocks, which led to the plunge that the investing public saw.
It was not until a computer system at the Chicago Mercantile Exchange, where the E-Mini contract is traded, implemented a temporary trading halt that the downward motion of the derviatives market was stopped. When trading resumed five seconds later (a long time in the world of computerized trading), E-Mini prices stabilized and then rose. The algorithm actually sold some contracts into this rising market.
Meanwhile, back at the ranch, stocks were still plunging. The early stock price drops had triggered more sell orders, as traders and their computers far and wide interpreted the market volatility as a signal to bail out. Buying interest exited stage right, exacerbating price plunges. Some stock sell orders were executed for a penny per share. Eventually, trades more than 60% above or below the 2:40 p.m. prices (which the exchanges and FINRA, a stock market regulator, deemed to be pre-lunacy prices), were cancelled. Why 60%? It's not very clear in the report.
Although the flash crash was over very quickly, it produced a intraday drop of 5-6% in stock prices (on top of a 4% drop that had already occurred that day). This scared the bejesus out of many investors, especially individuals, who subsequently moved more money out of stocks and into bonds even though bonds have absurdly low yields.
The report doesn't include any prescriptions for the future. No doubt, the SEC's and CFTC's enforcement divisions are sniffing around for violations of law. But there is scant indication in the report that anyone in either agency sees a likelihood of enforcement action. A crucial question is why did the selling mutual fund complex choose an algorithm that issued sell orders based on trading volume only, without any assessment of price impact and without regard to how quickly its orders were landing in the market. It had previously sold such a large quantity of E-Mini's, but much more slowly and without the sudden price drops. The report doesn't cast any light on the seller's thinking.
Based on the information available to date, there is a serious chance that neither the SEC nor CFTC will do anything on the enforcement front. On the regulatory front, the SEC has approved tighter triggers for trading halts. Sudden price moves in the most frequently traded stocks of 10% or more within a period of five minutes now result in a 5-minute trading halt. The SEC has also tried to make the process of canceling trades at seemingly off-market prices more transparent.
But the SEC and CFTC don't seem to think they can prevent another flash crash. At least, if they think they can, they surely didn't make that point in the report. Realistically, it would be quite difficult for them to "prevent" another flash crash, because they'd have to control the volume of orders reaching the market, trying in some way to balance buying interest with selling interest. Any such effort by the government would be destined to failure, as it would require the government to define supply and demand, fundamental market forces that governments can't effectively define. So it seems that the regulators can only soften the impact of future flash crashes.
So is there no legal consequence? A large institutional investor can just wallop the market and everyone who is clobbered learns the hard way that passbook savings are so bad?
There may be an answer in the history of American business. About 150 years ago, businesses began incorporating under newly enacted state laws that allowed anyone to create a corporation. Because the corporate form of business protected investors from unlimited personal legal liability for the business's liabilities, it became the dominant form of business enterprise. Incorporated businesses attracted large amounts of capital and grew quickly. Their reach became regional and then national. Companies in one state sold products to customers in other states 2,000 or even 3,000 miles away.
Some of these products were shoddy or defective. When customers tried to sue, they were hindered by a variety of legal doctrines, some of which dated back to medieval English law. Many state legislatures and some state courts took steps to modernize the law, resulting in the evolution of today's law of products liability. This body of law is based on principles that lawyers call "tort law," which hold that a person who is negligent can have civil monetary liability for the foreseeable consequences of his or her acts, even if the person didn't intend to cause injury. For example, early in the 20th century, courts began to hold auto manufacturers liable for defects in cars, even when they were thousands of miles away and didn't directly sell the car to the injured person. This was an outcome that the courts of the Civil War era would have considered unspeakable. But it quickly became the law of the land when commerce grew to be national in scope.
The financial markets have evolved way beyond the current legal structure. And the snail's pace of legislative reform, with the Dodd-Frank Act coming two years after the financial crisis of 2008, offers little hope that the top-down government regulation of the financial markets from Washington will keep pace. Maybe it's time to think about applying the principles of tort law to players in the financial markets. The mutual fund complex that evidently triggered the flash crash chose a forceful way of executing a massive quantity of sell orders in the E-Mini that it perhaps should have foreseen would cause disruption, chaos and losses to innocent investors. The threat of financial liability for losses and damage might well make market players pause and think before using potentially injurious trading tools.
A major advantage of applying tort law is that it doesn't try to regulate conduct. It creates liability that leads people to regulate their own conduct. Tort law applies to drivers on the roads. While many drivers don't seem to believe in scrupulous adherence to the rules of the road, almost all drivers try in their own way to be careful because an accident that injures others can lead to a jump in their insurance premiums. In other words, negligence costs them money so they exercise care.
Many on Wall Street would be aghast at the idea of tort law being applied to financial market players. The liabilities, they might proclaim, would destroy the financial system. But the same arguments could have been made about products liability law being applied to auto companies and all manner of other manufacturers. In general, that hasn't happened. And when it threatened, many companies ducked into bankruptcy court and worked out ways to compensate injured persons while continuing as businesses. The courts applied tort law in measured ways that allowed injured persons to obtain recompense without destroying American commerce.
The threat of tort liability to actors in the financial market could lead them to monitor and moderate their behavior. No government would tell them how to trade. They would decide for themselves how to trade. But they couldn't think only about themselves. They'd have to be concerned about smacking the corn flakes out of other market participants. Tort liability would motivate them to design and use trading algorithms and other trading tools in kinder and gentler ways, making the markets a better and safer experience for all.
The mutual fund complex, which hasn't been officially identified although only those without Internet access can't figure out its probable identity, chose a automated computer execution process based on an algorithm that simply sent out the amount of orders it calculated to be 9% of the previous minute's trading volume. These orders were placed without regard price or time. That made the algorithm insensitive to the impact its orders might have on the market. It was an automaton that simply measured the preceding trading volume and pumped out 9% of that volume in new orders. If trading volume fell, the number of new sell orders the algorithm would send out would fall. If volume rose, so would quantity of new sell orders the algorithm issued.
The early sales from the 75,000 contract E-Mini sell order were purchased by other institutional investors, mostly high speed traders that don't hold their purchases for very long. As more sell orders came in from the algorithm, the early purchasers got nervous about the E-Mini contracts they'd already purchased and tried to sell them. Thus, they added to the trading volume generated by the algorithm's sales. The algorithm took this increased trading volume as a signal to place even more sell orders (since 9% of a larger number calls for more orders than 9% of a smaller number).
Buying interest, however, shrank as more and more E-Mini sell orders came on the market. Some high frequency traders began to buy and sell from each other, because there were fewer and fewer other buyers. But that only increased trading volume, which led the algorithm to dump yet more sell orders onto the market. This only made things worse.
Some of the traders buying E-Minis sold individual S&P 500 stocks to hedge themselves (i.e., to limit their risk from holding E-Minis). This put downward pressure on stocks, which led to the plunge that the investing public saw.
It was not until a computer system at the Chicago Mercantile Exchange, where the E-Mini contract is traded, implemented a temporary trading halt that the downward motion of the derviatives market was stopped. When trading resumed five seconds later (a long time in the world of computerized trading), E-Mini prices stabilized and then rose. The algorithm actually sold some contracts into this rising market.
Meanwhile, back at the ranch, stocks were still plunging. The early stock price drops had triggered more sell orders, as traders and their computers far and wide interpreted the market volatility as a signal to bail out. Buying interest exited stage right, exacerbating price plunges. Some stock sell orders were executed for a penny per share. Eventually, trades more than 60% above or below the 2:40 p.m. prices (which the exchanges and FINRA, a stock market regulator, deemed to be pre-lunacy prices), were cancelled. Why 60%? It's not very clear in the report.
Although the flash crash was over very quickly, it produced a intraday drop of 5-6% in stock prices (on top of a 4% drop that had already occurred that day). This scared the bejesus out of many investors, especially individuals, who subsequently moved more money out of stocks and into bonds even though bonds have absurdly low yields.
The report doesn't include any prescriptions for the future. No doubt, the SEC's and CFTC's enforcement divisions are sniffing around for violations of law. But there is scant indication in the report that anyone in either agency sees a likelihood of enforcement action. A crucial question is why did the selling mutual fund complex choose an algorithm that issued sell orders based on trading volume only, without any assessment of price impact and without regard to how quickly its orders were landing in the market. It had previously sold such a large quantity of E-Mini's, but much more slowly and without the sudden price drops. The report doesn't cast any light on the seller's thinking.
Based on the information available to date, there is a serious chance that neither the SEC nor CFTC will do anything on the enforcement front. On the regulatory front, the SEC has approved tighter triggers for trading halts. Sudden price moves in the most frequently traded stocks of 10% or more within a period of five minutes now result in a 5-minute trading halt. The SEC has also tried to make the process of canceling trades at seemingly off-market prices more transparent.
But the SEC and CFTC don't seem to think they can prevent another flash crash. At least, if they think they can, they surely didn't make that point in the report. Realistically, it would be quite difficult for them to "prevent" another flash crash, because they'd have to control the volume of orders reaching the market, trying in some way to balance buying interest with selling interest. Any such effort by the government would be destined to failure, as it would require the government to define supply and demand, fundamental market forces that governments can't effectively define. So it seems that the regulators can only soften the impact of future flash crashes.
So is there no legal consequence? A large institutional investor can just wallop the market and everyone who is clobbered learns the hard way that passbook savings are so bad?
There may be an answer in the history of American business. About 150 years ago, businesses began incorporating under newly enacted state laws that allowed anyone to create a corporation. Because the corporate form of business protected investors from unlimited personal legal liability for the business's liabilities, it became the dominant form of business enterprise. Incorporated businesses attracted large amounts of capital and grew quickly. Their reach became regional and then national. Companies in one state sold products to customers in other states 2,000 or even 3,000 miles away.
Some of these products were shoddy or defective. When customers tried to sue, they were hindered by a variety of legal doctrines, some of which dated back to medieval English law. Many state legislatures and some state courts took steps to modernize the law, resulting in the evolution of today's law of products liability. This body of law is based on principles that lawyers call "tort law," which hold that a person who is negligent can have civil monetary liability for the foreseeable consequences of his or her acts, even if the person didn't intend to cause injury. For example, early in the 20th century, courts began to hold auto manufacturers liable for defects in cars, even when they were thousands of miles away and didn't directly sell the car to the injured person. This was an outcome that the courts of the Civil War era would have considered unspeakable. But it quickly became the law of the land when commerce grew to be national in scope.
The financial markets have evolved way beyond the current legal structure. And the snail's pace of legislative reform, with the Dodd-Frank Act coming two years after the financial crisis of 2008, offers little hope that the top-down government regulation of the financial markets from Washington will keep pace. Maybe it's time to think about applying the principles of tort law to players in the financial markets. The mutual fund complex that evidently triggered the flash crash chose a forceful way of executing a massive quantity of sell orders in the E-Mini that it perhaps should have foreseen would cause disruption, chaos and losses to innocent investors. The threat of financial liability for losses and damage might well make market players pause and think before using potentially injurious trading tools.
A major advantage of applying tort law is that it doesn't try to regulate conduct. It creates liability that leads people to regulate their own conduct. Tort law applies to drivers on the roads. While many drivers don't seem to believe in scrupulous adherence to the rules of the road, almost all drivers try in their own way to be careful because an accident that injures others can lead to a jump in their insurance premiums. In other words, negligence costs them money so they exercise care.
Many on Wall Street would be aghast at the idea of tort law being applied to financial market players. The liabilities, they might proclaim, would destroy the financial system. But the same arguments could have been made about products liability law being applied to auto companies and all manner of other manufacturers. In general, that hasn't happened. And when it threatened, many companies ducked into bankruptcy court and worked out ways to compensate injured persons while continuing as businesses. The courts applied tort law in measured ways that allowed injured persons to obtain recompense without destroying American commerce.
The threat of tort liability to actors in the financial market could lead them to monitor and moderate their behavior. No government would tell them how to trade. They would decide for themselves how to trade. But they couldn't think only about themselves. They'd have to be concerned about smacking the corn flakes out of other market participants. Tort liability would motivate them to design and use trading algorithms and other trading tools in kinder and gentler ways, making the markets a better and safer experience for all.
Sunday, July 11, 2010
ETFs After the Flash Crash
The SEC's recent proposal to apply circuit breakers to certain ETFs highlights the strengths and weaknesses of ETFs. Conceived as instruments you can trade all day long while the markets are open, they actually don't serve this purpose all that well. Because the price of an ETF is derived from the prices of the underlying securities, it will always lag behind underlying price changes. When underlying prices are moving rapidly, as they were during the flash crash, the pricing lag of an ETF may create an arbitrage opportunity for the high speed trading firms whose massive computing power lets them spot and exploit price anomalies faster than the typical retail investor can blink. As the big boys pile in, jarring price movements may scare away market makers, who ordinarily provide liquidity, faster than disaffected voters are abandoning incumbents. Ordinary individual investors trying to make a buck or two can (and sometimes did) lose their shirts.
Experience shows that even though ETFs were conceived and marketed as trading instruments, they're really better for long term investment. They often have low management fees and many are quite tax efficient. Used as a long term instruments, ETFs can be competitive with the cheapest traditional index funds. (See http://blogger.uncleleosden.com/2007/06/exchange-traded-funds-for-beginners.html.) Granted, faith in long term investment has fallen as the Dow has fallen. But it's probably still a better idea than short term, in and out trading where you pay commission costs, bid-ask spreads, and sometimes unpredictable prices, only to have your head handed to you.
Experience shows that even though ETFs were conceived and marketed as trading instruments, they're really better for long term investment. They often have low management fees and many are quite tax efficient. Used as a long term instruments, ETFs can be competitive with the cheapest traditional index funds. (See http://blogger.uncleleosden.com/2007/06/exchange-traded-funds-for-beginners.html.) Granted, faith in long term investment has fallen as the Dow has fallen. But it's probably still a better idea than short term, in and out trading where you pay commission costs, bid-ask spreads, and sometimes unpredictable prices, only to have your head handed to you.
Thursday, May 6, 2010
The Complexities of Computerized Stock Trading: We Told You So
Life gives you very few chances to say "we told you so." So you take the opportunities you get. It appears that today's abrupt 998 intraday point plunge in the Dow (which partially recovered to close down 347) was the result of computerized stock trading programs selling so fast they tripped over each other, and then interpreted the resulting mess as a signal to sell more. Things snowballed. See the explanation from the New York Stock Exchange. http://www.bloomberg.com/apps/news?pid=20601087&sid=aETiygQQ8Y3g&pos=1.
This is a problem we predicted last fall. http://blogger.uncleleosden.com/2009/10/computerized-stock-trading-grave-new.html. Nothing's been done since then, by the exchanges and other markets, big banks and other securities firms, FINRA, or the SEC. The NYSE announced it will cancel trades that were more than 60% above or below the 2:40 p.m. Eastern Time price. It appears that if you make a truly big mistake in the stock markets, you get a second chance. And when you buy stocks at a really cheap price, you can be punished for being astute. When computer programs are inadequate to spot a pile of dog doo on the sidewalk and rush to buy it, they (and the well-capitalized hedge funds and broker-dealers that use them) shouldn't be let off the hook. Make them pay the price of their mistakes and they'll fix their computer programs. Give them a bailout, and what incentive would they have to prevent this from happening in the future?
Too bad all those millions of folks whose 401(k) accounts got clobbered in 2008 didn't get a second chance.
This is a problem we predicted last fall. http://blogger.uncleleosden.com/2009/10/computerized-stock-trading-grave-new.html. Nothing's been done since then, by the exchanges and other markets, big banks and other securities firms, FINRA, or the SEC. The NYSE announced it will cancel trades that were more than 60% above or below the 2:40 p.m. Eastern Time price. It appears that if you make a truly big mistake in the stock markets, you get a second chance. And when you buy stocks at a really cheap price, you can be punished for being astute. When computer programs are inadequate to spot a pile of dog doo on the sidewalk and rush to buy it, they (and the well-capitalized hedge funds and broker-dealers that use them) shouldn't be let off the hook. Make them pay the price of their mistakes and they'll fix their computer programs. Give them a bailout, and what incentive would they have to prevent this from happening in the future?
Too bad all those millions of folks whose 401(k) accounts got clobbered in 2008 didn't get a second chance.
Subscribe to:
Posts (Atom)
Posts
