High-Frequency Firm Found Clever Way to Save on Capital

By just rewriting the net capital rules, I mean. They had a better system, so they used it. The SEC did not approve.
I don't think Latour Trading and Chateau Latour have much to do with each other, but isn't this pretty?

I've talked before about how much I'm enjoying the Securities and Exchange Commission's series of high-frequency trading crackdowns. A lot of people think that "the market is rigged," so the SEC has to do something about it. But the market mostly isn't rigged, so there's not that much it can do. But! A lot of high-frequency traders and exchanges and dark pool operators and other potential market-rigging culprits also, separately, do a lot of dumb stuff. So the SEC can just crack down on the dumb stuff, improve the market and look like it's doing something about market rigging, without doing anything disastrous just for the sake of doing something.

The latest crackdown in that vein is this $16 million penalty against Latour Trading, which is a delightful firm to crack down on, insofar as it is (1) huge ("Latour’s trading at times accounted for as much as 9 percent of the trading volume in equity securities for the entire U.S. market," says the SEC) and yet (2) not exactly a household name. High-frequency trading is a secretive conspiracy of opaque firms that you've never heard of but that use complex trading algorithms to control the markets from the shadows, is a series of words that I just wrote, but that someone else has probably written and meant.

So what did Latour do? It violated net capital rules. If you're a broker-dealer, as Latour is, you need to end every day with sufficient net capital. Roughly the way that works is:

  • you look at the value of each of your positions;
  • you apply a haircut to account for the risk of each position; and
  • you add up the post-haircut values and make sure that the sum is positive.

In 2010 and 2011, Latour reported its net capital at the end of every month. It had net capital before haircuts -- that is, net value of its positions -- ranging from about $33 million to about $86 million during that period. It applied haircuts ranging from $29 million to $65 million, leaving remaining net capital of as little as $1.8 million and as much as $35 million. 1 Which was plenty of net capital.

That, by the way, is a fun fact to pause on. Latour Trading ended February 2010 with $1.8 million in net capital. It traded "as much as 9 percent of the trading volume in equity securities for the entire U.S. market." That's like $21 billion of trading a day. 2 Latour calculated -- wrongly, we'll get to that -- that it had capital equal to a little under 0.01 percent of the amount of stock it traded in a day. Obviously trading $21 billion of stock in a day is not -- not at all -- the same as actually holding $21 billion of stock at any particular point in that day, and Latour was probably well capitalized compared to, say, how much money it made in a day. Still! That is a small firm trading a big chunk of stock.

But that's not the point; the point is, Latour's math was wrong. Its haircuts should have been much bigger, says the SEC, so much bigger that it normally had negative regulatory net capital:

From at least January 2010 through at least December 2011 (the “relevant period”), Latour consistently conducted a securities business while miscalculating the amount of net capital it had and thereby failing to maintain the required minimum net capital by millions of dollars. The firm operated without maintaining its required minimum net capital on 19 of 24 reporting dates during the relevant period. At end of day on those 19 reporting dates, the firm maintained net capital deficiencies in amounts ranging from approximately $2 million to as much as $28 million.


Latour messed up the math in several ways, some more interesting than others. The irresistible one is that, "Latour incorrectly used hypothetical positions to capitalize qualified stock baskets." Remember in my description of net capital I said that you value and haircut "each of your positions" and then sum them up. The key question is: What is a position? Latour's business is largely about making markets in exchange-traded funds. So it will buy shares of an ETF, say the SPY S&P 500 ETF, and sell short the underlying stocks. If you treated the ETF and the stocks as separate positions, then you'd apply about a 15 percent haircut to each side of that trade, 3 and end up with a huge haircut on what is actually a perfectly hedged position.

To prevent this problem, the SEC, in its wisdom, lets you offset your ETF with a "qualified stock basket." A qualified stock basket is just a basket of stocks that is close enough to the ETF. The stocks in the basket need to make up 50 (sometimes 95) percent of the market cap of the ETF's index. If they do, you can get a much lower haircut on the combined hedged position.

Latour had software to go through its ETFs at the end of the day, compare them to its stock holdings, and offset the ETFs with qualified stock baskets. But it did a crucial tricky thing, in that it "allocated hypothetical long and short positions that the firm did not actually hold to create qualified stock baskets."

That ... sounds ... terrible, no? Like, if you are long an ETF, and short the underlying stocks, then you are hedged, and you shouldn't have to hold that much capital, and that is what the rule says, and it is a good rule. But if you are long an ETF, and short the underlying stocks only hypothetically, then you are only hedged in your imagination, and really you should hold some extra capital, because why are you hedging real positions with imaginary ones?

But it's not actually that bad. Here's the way to think about Latour's systems. Latour can get long $1,000 worth of exposure to Apple by buying $1,000 worth of Apple stock. Or it can get long $34 worth of Apple by buying $1,000 worth of an S&P 500 ETF, since Apple is about 3.4 percent of that index. Or it can buy $1,000 of a Nasdaq 100 ETF and get long $134 of Apple. Or it can buy $1,000 of a Russell 1000 index and get long $30 of Apple. Et cetera. In fact, Latour is probably doing all of those things, frequently, throughout the day. And at the end of the day it will have $X worth of Apple exposure, some of it in the form of Apple stock and some of it in the form of various ETFs.

And, because Latour is a high-frequency trader that isn't really interested in holding on to lots of overnight equity risk, $X will be close to zero. But that doesn't mean that the Apple stock and ETF positions will be close to zero. It might be long a lot of S&P 500, and short some Nasdaq 100, and short some stock too, and those will add up to zero-ish Apple exposure. And so on down the line for every other stock: That long S&P 500 exposure will create some long Goldman Sachs exposure too, which isn't offset by the Nasdaq 100 exposure (Goldman's not on Nasdaq), so Latour will be short more Goldman stock, or maybe it'll be short a financials ETF like XLF.

Basically Latour is a giant system of equations: For each stock, it writes down all the ways that it can get exposure to that stock, sums them up and tries to solve for zero-ish for each exposure while still making money. 4

The SEC's rule says: If you are long the S&P 500 ETF, and short Apple and Goldman and most of the other stocks underlying it, you can use those stocks to offset that ETF for net capital purposes. But it says nothing about using the Nasdaq 100 ETF to hedge the S&P 500 ETF. It has a simple static rule: You can hedge an ETF with a qualified basket of actual stocks. How boring! Latour, being cleverer than the rule, said: Well, OK, I'll hedge my long S&P 500 with a hypothetical basket, and I'll hedge my short Nasdaq 100 with another hypothetical basket, and I'll be short 100 shares of Apple in one hypothetical basket and long 100 shares of Apple in another hypothetical basket, so net-net-net-net it's fine that I only own zero shares of Apple.

In other words, Latour's hypothetical positions were probably created by actual positions, not just by imagination. 5

That, to me, is the most interesting bit of naughtiness here. But there are others, some of which are similar to this, like offsetting S&P 500 futures with nearly identical S&P 500 futures, which makes sense in Latour's hedging system but not to the SEC. Others are just dumb though. 6

It's sort of charming, no? The SEC's rules are simple and blunt, and not nearly as sophisticated as Latour's own hedging and risk-management systems. Of course it's not supposed to be as sophisticated: It's supposed to be a blunt instrument for making sure that broker-dealers are well-capitalized. But Latour, like a good high-frequency trading firm, was looking for efficiency wherever it could find it. Even if that meant rewriting the rules to make them more efficient.

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

  1. The table is on page 11 of the SEC order.

  2. Source is Bloomberg MVALUSE <Index> HP <Go>. The average for February 2010 was around $236.7 billion of shares traded per day; 9 percent of that is $21.3 billion.

  3. I think that's how to read Rule 15c3-1(c)(2)(vi)(J) but don't go, like, running your brokerage based on this reading.

  4. I suspect that this oversimplifies and that Latour is more interested in statistical relationships than individual stocks -- like, you can partly hedge GS with MS and so forth -- but you get the idea.

  5. There may be some imaginary ones though. You need to have 50 percent of the underlying market cap in your qualified basket. So if you have 49 percent, you might create a hypothetical stock position to get to 50 percent and get the qualified-basket benefit, and an offsetting hypothetical stock position which would be subject to the regular haircut. That is, the discontinuity of the 50 percent rule might mean that adding 100 shares of long Apple and 100 shares of short Apple would leave you with a lower total capital requirement. So just add 100 shares of long hypothetical Apple and 100 shares of short hypothetical Apple! It's all hypothetical anyway.

  6. So:

    • Latour offset futures with other futures. So in February 2010 it was long $18.4 billion of S&P 500 exposure using E-mini S&P futures, and short $18.3 billion of S&P exposure using "big" S&P futures, which have different contract sizes. To Latour, it was net long $63 million of futures, for a haircut of $14,325. (Seems sorta low, no?) To the SEC, it had a total of $36.7 billion of futures positions -- 580 times as much as Latour calculated -- and so should have had $8.3 million of capital. Latour seems ... more right than the SEC? Economically? But still not, like, entirely right?
    • Latour "used incorrect index composition data for certain international ETFs," which just sounds like something you're not supposed to do? So for instance an international ETF might have 40 components, but only 9 of them were easy for Latour to trade. So it would hedge by trading those nine components. But since it hedged that way it just assumed that that was the hedge. And then it assumed that hedge was a qualified basket for SEC capital purposes. But it wasn't! At all! Latour just substituted its judgment for the SEC's -- and the ETF provider's, for that matter.
    • "Latour failed to take any haircut on some proprietary positions as a result of a computer programming error," oops.
    • Latour took credit for creations and redemptions of ETFs before they happened. So if you are long ETF shares and short the entire basket of underlying stocks, you have a qualified stock basket and a lower haircut. But not a zero haircut. But if you go to an ETF's authorized participant and redeem your ETF shares, it will give you back the underlying stock, and you can use it to close out your short, and then you have no position at all and thus no haircut. (See footnote 8 on page 9 of the SEC order for the mechanics.) So sometimes Latour would, like, e-mail its broker on a Saturday to say, "Hey we want to redeem this ETF first thing Monday morning," and then pretend for net capital purposes that it had already redeemed it, so the ETF would create no more capital requirement.

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

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

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