Sneaky Pegs Found Hidden Orders

Also bank errors in your favor, Uber, Gary Cohn, Dina Powell, Martin Shkreli and bond market liquidity.

Sneaky pegs.

Let me tell you a story that is so long and boring that I find it rather soothing. The New York Stock Exchange is a place where people go to post orders to buy and sell stocks. (This is a figure of speech -- it is actually a bunch of computers in New Jersey where algorithms go to post orders to buy and sell stocks, plus a TV set downtown -- but just go with it.) If you want to buy 100 shares and the stock is trading at $10.00, you might post an order to buy 100 at $9.99 and wait for someone to come in and sell to you at that price.

But if you post that order then people will know it is there, and you may not want that. Orders convey information, and telling people that you want to buy at $9.99 may give away too much information. (I mean, orders are anonymous, but telling people that someone wants to buy at $9.99 may still give away too much information -- specifically, information about how much demand there is for the stock.) So you can put in a hidden order (or "non-displayed" order, or "dark" order) to buy at $9.99. Then if someone wants to sell at $9.99, you will buy, but someone just examining the order book won't know about your demand at $9.99. 

Another thing you can do is "peg" an order. This means that if people have bid $9.99 to buy, and offered to sell at $10.01, you can put in a pegged order to buy at the bid, $9.99. And then if someone sells at $9.99 you will buy. But if instead all the other bidders move their bids down to $9.98, then your bid will automatically move down to $9.98: You have "pegged" it to the prevailing best bid or offer. You will buy at the best available price, but what that actual price will be can fluctuate.

The way the New York Stock Exchange set up the pegged orders allowed you to set a limit. So you might put in a pegged order with a cap of, say, $9.95. If the best bid was $9.90, your order would come in at $9.90. If it was $9.99, though, you wouldn't be on the best bid; that would be above your cap. Your order would price somewhere else. You might think it would price at $9.95, but actually the 2006 NYSE rules for pegged orders didn't work that way. Instead, "if the BBO [best bid or offer] was outside of the range set for the PI ['pegging interest' order], however, the PI would price at the level next closest to the quote, but within the PI’s range, where other interest (i.e., another order) existed." So if the best bid was at $9.99, but there was another bid at $9.95 (in your range), you would price there. If there wasn't another $9.95, but there was a $9.94, your pegged order would be displayed as $9.94. 

But someone very craftily figured out a way to fish for information with this mechanism. Because NYSE would display your order at the level where there was any interest, including non-displayed orders. So if the best displayed bid was $9.99, and the next-best displayed bid was $9.90, and you put in a pegged order with a $9.95 cap, and it displayed at $9.93, then you'd know there was hidden interest at $9.93. And then you could go make use of that knowledge for whatever nefarious purposes people were trying to prevent by putting in the hidden order in the first place.

NYSE did not adequately disclose this, and that was bad, and on Tuesday the Securities and Exchange Commission fined NYSE $14 million for this and other failings. But I wanted to tell that story in boring detail just to point out how boring it is, and how arcane. NYSE's rules allowed people to use pegged orders to detect hidden orders that (1) were not the best bid or offer, (2) were at levels with no displayed interest, and (3) were within the pricing range of the pegged order. (And it's even more arcane than that, because NYSE didn't even tell you where your pegged order was priced: You had to figure it out from context clues.)

When people talk about the markets being rigged in favor of high-frequency traders, this is the stuff that they mean. This is absolutely a problem! Favored traders should not be able to get information about other people's orders using undisclosed tricks. But it is such an arcane problem, such a subtle trick, such a complex way to get such a fleeting advantage. 

Reading about high-frequency trading you sometimes get the impression that the stock exchanges would just send the HFTs a note being like "hey John Smith is trying to buy 100 shares at $9.99, you'd better buy them first and sell them to him at $10.00." But the reality is often more like this. There's nothing you can directly do with this knowledge: If the best bid is $9.99, and you discover a hidden order at $9.93, it's not like you can go sell the stock to that hidden order at $9.93, or like you'd want to. It gives you some information about the shape of the order book and the demand for the stock, but it's not a foolproof cheat to profitability. It's just algorithms fighting each other for little scraps of informational edge.

Anyway, NYSE was fined for some other things too, of which the most notable is that it messed up reopening of stocks during volatility on Aug. 24, 2015. Some stocks moved so much that they triggered automatic halts, and when NYSE's Arca exchange reopened trading in those stocks, it applied price collars -- excluding bids or offers more than 1, 2 or 5 percent away from the last price -- to the reopening auction. But its rules didn't say it could do that:

Arca’s application of price collars to reopening auctions was a material operation of the exchange that should have been the subject of an effective exchange rule. On August 24, 2015, however, Arca described opening and closing auction collars in its rules, but did not include reopening auction collars in its rules.

The fix for this omission was to (1) apply the opening and closing auction collars to the reopening anyway and then (2) write a rule about it: "Arca promptly engaged in discussions with Commission staff concerning a proposed rule amendment to reflect the exchange’s use of collars in reopening auctions, and filed a rule proposal on December 7, 2015." But you're supposed to do that in the other order. Arca's rules had a gap, and it filled in the gap on the fly, and the SEC fined it for doing that.

In a sense: quite rightly! But you could imagine a different vision of exchange operation. For instance you might say that exchanges should be run by smart people with good judgment and that, when a situation arises that was not anticipated in their rules, they should respond to that situation in good faith using their best judgment. This is not that approach. This is: You'd better have the rules in advance, because you can't just make up rules when you need them, even in an emergency.

That's a good approach too -- in some contexts it is called "the rule of law" -- but it is especially appropriate when your customer base is mostly computers. If you deal with humans on a human time scale, you can improvise a solution and call them up and talk it through and try to get buy-in that your approach is fair. If you deal with algorithms on a microsecond time scale, your rules have to be algorithmic themselves. It is a vision of a stock exchange as a knowable computer architecture, not as a consortium of reasonable brokers who meet together to figure stuff out in their common interest. It is undoubtedly the correct vision, here in the 21st century, but it is a bit removed from the NYSE's early days of buttonwood trees and gentlemanly trading.

Bank errors.

A theme we've discussed a bit recently is the uncanny tendency of banks, when they make dumb computer errors, to do so in ways that make them money at the expense of their customers. "You rarely hear of a scandal like 'Wells Fargo accidentally lost track of thousands of credit-card fees and just didn't collect them,'" I wrote last month, "or 'MetLife accidentally created thousands of fake accounts and sent them pensions.'" Of course there are exceptions. Here is an astonishing one:

A botched systems upgrade by BNP Paribas prevented the French bank from properly booking structured-products trades in Germany for a week, potentially affecting thousands of trades.

BNP's lost week started on Dec. 2, 2015, and "it was not until December 9 that the bank noticed its long-planned migration to a different pricing platform had caused the new system’s execution engine to be disconnected from the rest of the bank’s systems." Obviously this is bad and "raises questions over how well the bank was hedged if it was not fully aware of its market-risk position," as well as questions about ... what ... exactly ... they are doing over there. ("BNP’s risk management is considered exemplary," says BNP.)

But the immediate issue is that they missed one particular trade, and are now being sued over it:

The lawsuit, first filed in 2017, turns on an allegation from the trader that he is owed as much as €163m from a December 2015 trade in so-called certificates, which are retail structured products that are particularly popular in Germany. The certificate in question dropped overnight from €54,000 to €108.80 because of an inputting mistake by a BNP employee. The trader ordered 3,000 of the certificates off-exchange.

While banks can cancel fat-finger trades, they typically only have a day or so to do this. BNP’s systems-migration issue meant the trade was accepted but not logged and therefore not recognised as a mistake.

Whoops! That's a 162-million-euro bank error in his favor. Good for him. BNP is, rather unsportingly, fighting the lawsuit hard, and "argues that the trader had a duty as a counterparty to report any obvious errors to the bank." But it seems to me that if you are a bank dealing in retail structured products, you really should not be relying on your retail customers to notice your obvious errors. If they're so obvious, you should notice them yourself! You're the bank! It is not a good look for a big sophisticated bank to do a dumb trade, forget about it for a week, and then blame its customer for taking advantage of its naïveté. 


I like to say that Uber Technologies Inc. is a large public company that happens to be private, but it's actually stranger than that. It is also a mature company that happens to be immature, a big stable company that happens to be hugely cash-flow-negative and locked in an existential battle for its future. Here's a story about how Uber is raising a $1.25 billion leveraged loan despite its negative $2.2 billion of earnings before interest, tax, depreciation and amortization last year:

And given Uber’s cash burn and annual loss, investors will probably be asked to assess the company by other metrics. One might be its blended valuation of $54 billion by a SoftBank Group Corp.-led investor group. That made it the biggest venture-backed technology enterprise without a stock listing. Management may also tout the $4.5 billion of cash that company holds on its balance sheet as of December 2017, according to documents seen by Bloomberg.

Now I am not a leveraged-loan investor but none of that sounds all that appealing? The $4.5 billion of cash is a fraction of what Uber has raised, and it covers about two years of negative Ebitda, meaning that if you lend money to Uber and it keeps going at this rate, it will run out of cash before your loan comes due. Equity investors at least get an indefinite-lived claim with, potentially, a lot of upside; lenders, in the best case, just get their money back with interest. The $54 billion equity valuation underneath the loans is not ... you know ... easily monetizable. But this is all just old-fashioned thinking; the reason that Uber is probably an attractive credit is not that it has positive cash flow or valuable assets, but that it is Uber. In the financial markets, Uber just creates its own reality.

By the way: "Another twist to Uber’s plan is that it’s pitching the debt directly to investors, bypassing the traditional route of using banks to fan the loan out to the lender group." You don't need banks to syndicate your loan if you are Uber. (Though "it does have Morgan Stanley acting as an adviser.") Spotify Technologies SA is doing a non-IPO IPO, and now Uber is doing a non-syndicated syndicated loan. I am not convinced that the big private tech companies are going to fully disintermediate the banks any time soon, but it is a bit of a trend.

In other news from the global consortium of car-hailing companies:

Uber Technologies Inc. has reached an agreement in principle to sell most of its Southeast Asia operations to local rival Grab Inc., ending a costly fight for market share in the fast-growing region, according to people familiar with the matter.

In exchange for its operations in Southeast Asia, Uber would gain a roughly 30% stake in Grab, these people said. 

Why compete when you can all own each other and divide up markets to maximize profitability? Or minimize unprofitability I guess.

Goldman comings and goings.

Here are some ideas for what Gary Cohn should do after leaving the White House, mostly random suggestions like "run SpaceX" or "chair and mentor a promising fintech company" or "Baker’s Bay in the Bahamas is the new in-place." I like James Tisch's suggestion that he "could head a financial institution, be the front man for a venture firm, or own a local restaurant." Three basically equivalent options! He should open a restaurant on 44th Street next to Anthony Scaramucci's Hunt & Fish Club. Eventually that street could be a whole row of ex-Trump-official financiers' restaurants.

Elsewhere people at Goldman Sachs Group Inc. are mad that Dina Powell, another Goldman alum who briefly worked for President Trump, is not only coming back but is being put on the management committee:

The decision to create a new seat on the committee for an executive whose prior experience at Goldman Sachs centered on philanthropy and related initiatives has become a topic of debate inside its hyper-competitive headquarters in lower Manhattan. Before joining the White House last year, Powell’s projects included helping small businesses and charities. Adding that background to a board stocked mostly with profit producers serving elite global clients is frustrating some top executives, people close to the situation said, asking not to be identified describing private conversations.

Ehh. Notions of "merit," in the financial industry, are a little slippery. The sad fact about banking is that you don't necessarily become a top profit producer by working your way up from being a middling profit producer. You become a top profit producer by attracting profitable clients. Sometimes -- often, even -- that comes from hard work and honed skills and lots of practice attracting clients. Sometimes it comes from having a high-profile role and connections in government. (Powell's "White House experience could help Goldman do more business with sovereign clients -- who will no doubt ask Powell to explain the unpredictable Trump.") "Important people like to deal with important people," runs one of John Whitehead's famous commandments for Goldman rainmakers. "Are you one?" If you are, all the committees are open to you. 

(Disclosure: I used to work at Goldman, but was not important.)

How's Martin Shkreli doing?

I don't know, but I wanted to flag here that he is being sentenced today, so keep an eye on that. Here's a story about his lawyer's sentencing letter, which ... I feel like ... is not ... entirely accurate?

Martin Shkreli’s lawyer made a last-minute pitch to the judge who’s going to sentence him Friday, saying his client didn’t try to create "a circus-like atmosphere" at his trial, which he blamed on a “rabid media presence" that “did its best" to disrupt the case.

I can think of few people alive who have worked harder to create a circus-like atmosphere in their lives, or to attract a rabid media presence, than Martin Shkreli. He absolutely should be blamed for bringing a circus with him wherever he goes. But that is not a crime, and I'm still sticking to my belief that two years in prison is about the right sentence for the fairly low-impact fraud that he was convicted of. Though as a betting man I would take the over on that. 

People are worried about bond market liquidity.

Here is a claim from Martin Fridson that high-yield bonds aren't volatile enough:

Price swings should fall along a normal distribution. But it turns out the bell curve in high-yield bonds is messed up.

Data for the investment-grade index fell almost perfectly along the expected curve. Government bonds had just a bit more periods of greater volatility than expected. But in high yield, swings of more than one standard deviation—a measure of how far a data set is from its mean—occurred about 30% less often than they should have.

“High-yield bonds, unlike governments or investment-grade corporates, exhibit fewer extreme price moves than would be expected in a properly functioning market,” Fridson, the chief investment officer of Lehmann Livian Fridson Advisors, wrote this past week. “If a dice-rolling experiment fails to produce the predicted distribution of outcomes…we know that the dice aren’t fair.”

One possible response here is that bonds aren't dice and no one said they were fair: If reality does not correspond to your model, then the problem is probably with your model, not with reality.

But there is a tempting other possibility, in that bond index levels aren't exactly reality. You could imagine a situation in which a bond trades at 100, and then the world gets worse, and people try to offload it at 80 but no one is willing to buy it for more than 70. So it doesn't trade, and the last trade on the tape is still at 100. Viewed in a certain light, its price has fallen by 30 percent. Viewed in another light -- the light of the tape -- its price has moved by zero percent. This could be a story about a kind of bond market illiquidity: Perhaps when bond prices should move a lot, they instead stop trading, and so the move in value never shows up in prices.

Things happen.

Cigna to Draw Antitrust Scrutiny Amid Wave of Health-Care Deals. Cigna Deal Shows Being a Health Insurer Isn’t Enough Anymore. Wynn Resorts to Pay $2.6 Billion in Settlement That Removes a Barrier to Steve Wynn Stake Sale. Toys ‘R’ Us Is Prepping to Liquidate Its U.S. Operations. UBS Boosts Compensation for CEO and Staff After Profit Rises. Hess Plans Another Buyback, Winning Peace with Activist Investor. Mortgage Rates at a Four-Year High Threaten to Roil Housing. "The judge suggested Trump could mute accounts he doesn’t want to see, rather than blocking followers." "An Illinois federal judge ruled that a man’s death during an act of autoerotic asphyxiation was not a self-inflicted injury and thus an insurer wrongly denied accidental death benefits to his widow." 

If you'd like to get Money Stuff in handy email form, right in your inbox, please subscribe at this link. Thanks! 

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

    To contact the editor responsible for this story:
    James Greiff at

    Before it's here, it's on the Bloomberg Terminal.