Ooh a little on the nose.

Photographer: Fabrice Coffrin/AFP/Getty Images.

UBS's Dark Pool Was Shady

Matt Levine is a Bloomberg View columnist. He was an editor of Dealbreaker, an investment banker at Goldman Sachs, a mergers and acquisitions lawyer at Wachtell, Lipton, Rosen & Katz and a clerk for the U.S. Court of Appeals for the Third Circuit.
Read More.
a | A

The Securities and Exchange Commission has had a big week of market-structure enforcement. It started with Direct Edge's $14 million "Hide Not Slide" fine on Monday, and then there was Tuesday's army of spoofers. Today brought a $14.4 million fine against UBS Securities for doing bad stuff in its dark pool. So that's a $28.4 million week, even assuming, as I kind of do, that the SEC will recover nothing from the spoofers.

But it feels a little like cheating to count UBS in this week's tally. UBS's main violation is that, "Between May 2008 and March 2011, UBS violated Rule 612 of Regulation NMS promulgated under the Exchange Act by accepting and ranking hundreds of millions of orders priced in increments smaller than one cent ('sub-penny orders')." Which is bad! UBS also was selective in its disclosure about those orders (called PPP orders, for "PrimaryPegPlus," apparently all one word). Basically, it told its favored high-frequency-trader clients how they worked, but was a bit shifty about them with other clients. Notice, though: "Between March 2008 and March 2011." This ended almost four years ago. Because, "On March 11, 2011, and after a Commission examination team had identified PPP and raised concerns that it might violate Rule 612 of Regulation NMS, UBS decommissioned the PPP order type."

Huh. How do you think that went? I mean, I have no idea, I wasn't there, and it was four years ago, but here's my imaginative reconstruction:

SEC examiner: Hey do you guys realize you're allowing high-frequency traders to post orders with sub-penny pricing?

UBS: Oh, umm, yeah. Yeah. I suppose we are doing that.

SEC examiner: You know that's illegal, right?

UBS: Hmm. Fair point. This is awkward. Uh ... we're really sorry?

SEC examiner: No, it's no big deal, we're not mad or anything. But you should probably cut it out.

UBS: Gotcha, won't happen again. Thanks for the heads-up.

Right? I mean, the other possibility is that the SEC has been investigating this case for the last four years, ever since it first noticed the problem in that 2011 examination. But once it had caught the clear violation of Rule 612, what was left to investigate? And why would it take four years?

Dark Pools

My best guess is, the SEC caught this in 2011, but in 2011 the technicalities of how you operate your dark pool were ... sort of viewed as technicalities. If you got it wrong, the SEC told you to fix it, and then you and the SEC had a good laugh about how you messed up. Now, of course, Catching Bad Dark Pool Stuff is a high-profile (Michael Lewis!), competitive (Eric Schneiderman!) and  lucrative ($14.4 million!) business. And if the SEC went back over its files to find Bad Dark Pool Stuff that it previously caught, in order to re-catch it for profit and profile, well, you could hardly blame it, could you?

The SEC is really good at looking busy on market structure stuff, is what I'm saying here.

The case itself is kind of weird too. The biggest and most interesting allegation is that UBS allowed favored high-frequency traders in its dark pools to post orders that were a little better than the national best bid and offer, in order to get priority in execution.  So if a stock was $50.00 bid, $50.02 offered on Nasdaq, you could post a bid of $50.002 on UBS's dark pool, and any marketable order coming to that dark pool would trade with you first, giving you priority over any chump sitting there with a $50.00 bid.  

That is illegal -- you can only post orders in penny increments -- but why is it bad? Well mainly it's unfair to other people playing by the rules and making their bids and offers in penny increments:

As a result, the PPP order type facilitated the very result that Rule 612 was designed to prevent: it allowed one subscriber to gain execution priority over another in the order queue by offering to pay an economically insignificant sub-penny more per share. Further, because UBS ATS allowed its subscribers to place orders at prices that were unavailable at ATSs and exchanges that complied with Rule 612 of Regulation ATS, UBS ATS obtained an unfair competitive advantage over those venues in its efforts to attract and execute orders from market participants.

The losers were people who were unfairly competed out of the chance to buy and sell stock. But there were winners, too -- not just the favored high-frequency traders, but also the people who got to sell (buy) stock at prices above (below) the market price, albeit by "an economically insignificant" amount.

For the high-frequency-trader customers, the appeal of paying a bit more than the market price seems to have been that they wanted to improve their chances of executing against information-insensitive orders. The name of the game, in HFT market-making, is trading a lot while avoiding adverse selection. If you can trade mostly with uninformed orders, you can work for a tighter spread than public markets do (1.6 cents instead of 2 cents, say), because your trading is unlikely to move the price against you. Retail orders -- you know, moms and pops and so forth -- tend to be uninformed, so they are everyone's favorite orders to trade with. And in fact the high-frequency-traders using these orders seem to have been trying to interact with retail order flow:

In March 2011, UBS sent a spreadsheet containing certain trading information to a firm that engaged in high-frequency trading and market making (“Subscriber C”) in an effort to encourage its expanded usage of PPP. For each of the thousands of orders Subscriber C had executed in UBS ATS on two prior trading days, the spreadsheet showed, without providing any customer-identifying information, whether the firm’s order had executed against a retail order or a non-retail order. (Such information was not data that UBS typically disclosed to other UBS ATS subscribers and was not information that subscribers could readily ascertain through other means.) The UBS employee that proposed sending the spreadsheet to Subscriber C understood that Subscriber C wanted to use the spreadsheet’s data to adjust its algorithmic trading strategies in ways that would increase the likelihood of its PPP orders executing against orders from retail broker-dealers.

It's unclear what that likelihood was. But what a weird fact! First of all, because, how does retail order flow even get to a dark pool? The market-structure stereotype is that retail brokerages sell their order flow to "internalizers" or "wholesalers" like KCG, Citadel, Citigroup or ... hmm, UBS, whose wholesaling deals include one with Charles Schwab. I always assumed that wholesalers mostly trade with these orders directly, rather than routing them to dark pools to trade with high-frequency market-makers. But apparently a good chunk of UBS's retail flow ended up trading with clients of its dark pool. (Overall, 16-18 percent of UBS's non-directed orders in listed stocks end up in its dark pool.)

But second, because UBS is allowed to do this. I mean: It's not allowed to do what it did. It's not allowed to let its HFT customers post sub-penny bids in its dark pool. But it is allowed to use sub-penny price improvement to trade with retail order flow that it internalizes. That is: It's allowed to do for itself what it (illegally) let its customers do.

This is a controversial, but very common, practice: Wholesalers like UBS pay retail brokers to send their orders to UBS, and then UBS trades with those orders by offering sub-penny price improvement, buying at a fraction of a penny above the national best bid, or selling at a fraction of a penny below the national best offer. Because retail order flow is delicious and uninformed, and so UBS can trade at tighter spreads with retail orders than the public market requires.

You get the sense that that's why this had a no-harm-no-foul feel in 2011. Sure, UBS broke the rules. But all it did was let high-frequency traders do in its dark pool what it was already, perfectly legally, doing with retail orders in its own wholesaling business. If you think that's bad -- a more popular view now than it was four years ago! -- then what UBS did was even worse. But if you're OK with sub-penny price improvement for retail orders -- and the SEC basically is! -- then UBS's efforts to spread that improvement a bit more broadly might not seem all that bad. 

  1. Under Rule 612 of Regulation NMS, it's illegal to "display, rank or accept" a bid or offer "priced in an increment smaller than $0.01," with exceptions not relevant here.

  2. There's a bunch of other allegations too, including one in which UBS allowed customers using its algorithms to avoid crossing with "non-natural" orders, meaning basically orders from evil high-frequency traders. But it didn't tell everyone about this functionality, or tell the evil HFTs that it had categorized them as evil, "until approximately 30 months after it was launched." In July 2012. So, again, sort of an old problem.

    Or, again, UBS gave access to its full dark-pool order book to UBS employees who weren't supposed to have it. But that's "primarily IT personnel"; there's no claim that UBS was illicitly trading against the dark order book.

  3. And avoiding the need to route to Nasdaq if there's no one in the dark pool with a $50.00 bid. I say "Nasdaq" to mean just "the national best bid and offer on a public exchange," basically.

    The mechanics of PrimaryPegPlus are described in paragraphs 15 to 19 of the SEC order. You could peg your order to the NBBO, plus or minus an amount expressed as a percentage of the spread. So in my example, which is also the SEC's example, and which in turn is also the example that UBS used in marketing PPP to its customers, the spread is $0.02, and 10 percent is $0.002. PPP initially worked in one-percent increments -- $0.0002 per share in that example -- but eventually UBS realized that was sort of silly and moved it to 10-percent increments.

  4. ATS = "alternative trading system," i.e., dark pool.

This column does not necessarily reflect the opinion of Bloomberg View's editorial board or Bloomberg LP, its owners and investors.

To contact the author on this story:
Matt Levine at mlevine51@bloomberg.net

To contact the editor on this story:
Zara Kessler at zkessler@bloomberg.net