Merrill Lynch Sold Some Stocks Too Fast
There are a lot of dumb obvious ways to "fat finger" your way into losing a lot of money in the stock market. You can try to buy 2,000 shares and accidentally buy 2,000,000 instead. You can try to buy 2 million dollars of Alphabet stock and accidentally buy 2 million shares -- more than $1.5 billion worth -- instead. You can try to buy 2,000 shares, not realize that your order has been sent, and then send the order again. (And again and again.) Trading is complicated and computers are fussy and you were out late last night and all those zeros tend to blur together.
Yesterday the Securities and Exchange Commission fined Bank of America Merrill Lynch $12.5 million "for maintaining ineffective trading controls that failed to prevent erroneous orders from being sent to the markets and causing mini-flash crashes," and the SEC's order lists 15 examples of Merrill fat-fingering the stock market. But what's striking about the examples is that none of them are quite that sort of dumb fat finger. Instead, they're like this:
On November 7, 2012, the price of Diageo plc American Depository Receipts (symbol “DEO”) rose by approximately 14% in less than 2 seconds, after a Merrill Lynch financial advisor inadvertently sent a market order on behalf of multiple clients to buy 115,500 shares of DEO, when intending to submit a market order for only a portion of that order—500 shares of the larger 115,500 share order.
On January 18, 2013, the price of PPG Industries, Inc. common stock (symbol “PPG”) fell by approximately 10% in less than 2 seconds, after a Merrill Lynch institutional trader inadvertently entered an order to sell 300,000 shares of PPG at the market (to hedge a customer swap order), when intending to enter the sell order as a market on close order.
On March 8, 2013, the price of Anixter International, Inc. common stock (symbol “AXE”) fell by approximately 12% in less than a second, after a Merrill Lynch institutional trader erroneously entered a market order to sell short 91,524 shares of AXE stock as part of a principal hedge instead of sending the order to a Merrill Lynch algorithm for execution over the last thirty-minutes of the trading day, as had been intended.
In all of these cases, Merrill Lynch actually meant to buy (or sell) all the shares that it bought (or sold). There was no mistake about the goal. The only mistake was one of tactics: Merrill didn't mean to buy (or sell) the shares so fast. Merrill bought that whole slug of Diageo, for instance, in two seconds, instead of reeling it in in 500-share lots over the course of minutes or hours. Merrill sold all those PPG shares in two seconds at around 3 p.m., instead of in a few seconds at around 4 p.m., when the closing auction would have attracted a lot of buyers. Merrill sold all those Anixter shares in one second, instead of dribbling them out over 30 minutes. The clients got exactly the shares they wanted, but not at the prices they wanted or expected. Instead, they got them at ridiculous prices -- prices that were 12 or 14 or 21 or, in two cases, 99 percent away from the market price a few seconds earlier, when the clients put in the orders.
So, sure, that's bad. It's not really bad for the clients, who for the most part didn't really get those ridiculous prices because the trades were torn up. ("Given that erroneous trades were canceled by the relevant exchanges in most instances, we are not aware of any client who was harmed as a result," says a Bank of America spokesman. ) But it's bad for everyone to have trades getting torn up:
“Mini-flash crashes, such as those caused by Merrill Lynch, can undermine investor confidence in the markets,” said Andrew Ceresney, Director of the SEC Enforcement Division. “It is essential that broker-dealers with market access have reasonable controls to prevent erroneous orders that disrupt trading.”
Still, reading this order, it is tempting to say that investor confidence in the markets should be undermined a bit. What's disturbing about this case is precisely that nothing went fundamentally wrong: The clients really did want to buy (or sell) that much stock. They could pretty easily have found sellers (or buyers) at reasonable prices (that is, without moving the price much) within a reasonable period of time (that is, like an hour or so). The quantities were not even particularly large: PPG, for instance, was a $20 billion company that traded more than $400 million a day. It should not have been hard to find buyers for Merrill's sell order, which was for $42.8 million worth of PPG stock, 0.2 percent of the shares outstanding. If Merrill had waited an hour, it would have been easy. But it didn't, by accident, so the stock was briefly down 10 percent.
What went wrong here isn't that the clients wanted to do extraordinary trades, or that they wanted to do ordinary trades that Merrill accidentally keyed in as extraordinary. They wanted to do utterly routine trades, but Merrill showed their hand a bit too early, and they were sunk. If you want to sell 0.2 percent of PPG's outstanding stock in the New York Stock Exchange closing auction, that's no problem. If you want to sell 0.3 percent of Anixter's outstanding stock over half an hour, that's no problem. But if you accidentally ask to sell those quantities immediately, the market will gleefully oblige, and will charge you a 10 percent or 12 percent convenience fee.
There are various ways to analyze this from the perspective of The Market, that is, the array of algorithms lined up against Merrill Lynch on these trades. You could say that the market was being rationally cautious: If you come to the market demanding to dump too much stock all at once, then you must know something that it doesn't, and it should be very hesitant to take your stock off your hands without a big discount. Or you could say that the market was being an opportunistic jerk: If you come to the market demanding to dump too much stock all at once, then it will recognize your desperation and charge you an arm and a leg, because charging you whatever you'll pay for immediacy is how the market makes its money. Or you could say that the market was being unnecessarily skittish: The prices generally rebounded almost instantly, and almost fully, meaning that a rational human market-maker could have assessed the situation and agreed to buy Merrill's stock at a discount of, say, 3 percent instead of 10 percent, making a mutually beneficial trade -- if she had only had the time. It is all more or less the same to the algorithms, whose job is to trade quickly, and who are not paid to stick out their algorithmic necks.
But from the perspective of Merrill and its clients, the lesson is simpler: If you want to buy or sell stock, and you accidentally let slip how much you want to buy or sell, then you will be punished ruthlessly. This is of course no surprise to people who trade stocks in today's highly computerized market: They would probably share the SEC's incredulity that Merrill Lynch's systems didn't prevent it from exposing a whole half-hour's worth of orders all at once. It wouldn't even really be a surprise to people who traded stocks in the good old days of human traders: Of course if you tell everyone that you have a lot of stock to sell, they're going to take advantage of you. That's just what traders do; the whole art in trading is trying to find people to buy your stock without letting on that you're selling.
Still, it isn't hard to perceive the modern market's ruthless speed and efficiency in responding to Merrill's mistakes as a form of instability. A broker's tactical error becomes not just an inconvenience for its client but a "mini-flash crash" for the market as a whole, not just bad customer service but also illegal. And a poor choice about how to go about executing a customer's order can cost the customer, not a little bit of slippage on the final price, but 99 percent of the stock's value. And remember, these customers had the help of a big brokerage with clever algorithms to protect their orders, albeit a brokerage that sometimes hit the wrong buttons for the wrong algorithms. Imagine a customer trying to sell 0.2 percent of PPG's stock without a broker's advice. If you expect equity markets to be simple and transparent and accessible to non-specialists, the fact that a broker's execution mistakes can be this disastrous has to be alarming. But of course there's no reason to expect that.
Also: "The Financial Industry Regulatory Authority fined Merrill Lynch on behalf of the three largest U.S. exchange operators an additional $3 million for the infractions."
One or two sort of are. For instance:
On May 17, 2013, the price of Anadarko Petroleum Corporation common stock (symbol “APC”) dropped by more than 99% in less than a second, due to erroneous orders that Merrill Lynch sent to the market. Specifically, on May 17, 2013, a Merrill Lynch institutional trader entered a principal market order to sell 400,000 shares of APC in the last minute of regular market trading. Within seconds, the trader became concerned that a portion of the order was not being executed and entered a new market order to sell an additional 150,000 shares of APC without canceling the first order. Consequently, the trader sent more shares than intended to the market; and the price of APC stock then plunged, falling dramatically before rebounding.
The basic problem there seems to be the timing, but the duplicate order can't have helped.
Bloomberg shows the big drop as being at just after 3:00 p.m. The New York Stock Exchange Closing Auction starts at 4:00 p.m., though it usually prints a minute or so after that.
The other thing that sometimes happened is that Merrill put in market orders when it meant to put in limit orders. That is loosely speaking a variant on the theme of doing it too fast, though conceivably some limit orders were never meant to execute at all.
One is the Anadarko example in footnote 2; the other is Qualys:
On April 25, 2013, the price of Qualys, Inc. common stock (symbol “QLYS”) fell by more than 99% in less than 2 seconds, after a Merrill Lynch trader erroneously entered an institutional customer order to sell 381,020 shares of QLYS using a Merrill Lynch algorithm intended for a different order, which, in turn, routed the order using an algorithm that executes orders immediately and exhausts liquidity at each price level until an order is complete.
Incidentally many of these trades seem to be hedges to derivatives trades that Merrill did as principal. We don't know how, for instance, the PPG swap was priced: Maybe Merrill priced the swap to the customer off of where it executed its hedge (in which case the customer would effectively suffer from Merrill's poor execution), or maybe the swap price was firm and the poor execution was Merrill's problem. It seems like most of the time the trades were just canceled, so they ended up being the stock counterparty's problem, rather than either the derivative customer's or Merrill's.
The proof of the insufficiency is basically that Merrill kept submitting erroneous orders, but you can be more legalistic about it if you want. The SEC locates the problem in the fact that Merrill had a very high threshold for rejecting orders:
The limits that Merrill Lynch established to prevent erroneous orders in many of its trading channels were set at such high levels that they failed to prevent many erroneous orders from reaching the market. These unreasonably high limits, along with the lack of other controls in many of Merrill Lynch’s trading channels, resulted in Merrill Lynch sending erroneous orders to the market.
"For example," says the SEC, "Merrill Lynch applied a limit of 5 million shares per order for one stock that only traded around 79,000 shares per day." That does seem too high?
The stock closed at $142.73 on January 18, 2013 (it has since split 2-for-1), and there were about 142.6 million shares outstanding, for a market capitalization of about $20.4 billion. The SEC says: "The order was approximately 11% of the average daily trading volume in PPG over the trailing 20-day period"; Bloomberg shows the average volume in that time was about $425 million a day.
That Diageo trade was for about $13.3 million worth of stock, on a $79 billion market capitalization. (That is, 2.75 million ordinary shares outstanding, at four shares per ADR, is about 689 million ADR equivalents -- or about $79.3 billion at the $115.20 per ADR closing price on the trade date.) Trading 0.02 percent of its stock caused a 14 percent drop. That's not quite fair, though since Diageo mostly trades in London.
Finally, Anixter: $6.4 million of stock in a $2.3 billion company that traded $36 million a day. (The closing price on March 8, 2013, was $69.51; there were about 32.5 million shares outstanding.)
You could characterize either the rationality or the skittishness in order-book terms. The orders resting on the stock exchanges to buy PPG at various prices don't reflect all of the real demand to buy PPG at those prices. Presumably a lot of investors would have very happily bought PPG down 1 or 2 or 5 percent on that day; if the order book had reflected all of their demand -- or if there was some way to have a constantly informed auction among every market participant -- the price would barely have budged. But on the other hand it isn't particularly rational to leave that order on the order book: Sometimes, if you bid to buy PPG down 5 percent, you'll get to buy it at a bargain because there's an idiosyncratically desperate seller. Other times, you'll end up catching a falling knife because there's an informed desperate seller. Placing an order on the book is giving away a free option.
Another weird legal aspect: The SEC is, sort of, punishing Merrill for failing to disguise its customer's intent. What got Merrill in trouble is that it accurately disclosed to the market how many shares its customers wanted to buy (or sell), and the market reacted strongly to that news. It's a precedent worth keeping in mind when you think about, say, spoofing, or about arguments that Carl Icahn should have to disclose his trading intentions before he trades.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
To contact the author of this story:
Matt Levine at firstname.lastname@example.org
To contact the editor responsible for this story:
James Greiff at email@example.com