Martin Wheatley runs the FCA, so he gets his picture here.

Banks Manipulated Foreign Exchange in Ways You Can't Teach

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 FX market, in which traders are able to buy, sell, exchange and speculate on currencies, is one of the world's largest and most actively traded financial markets," with trading averaging $5.3 trillion a day. And six of the biggest banks in that market just settled charges of manipulating it, paying a total of about $4.3 billion to a bunch of regulatory agencies, with a bunch more regulators (and Barclays!) still to come. So that seems like a big deal. Bigger than Libor, even, though still a speck next to the Everest that is Bank of America's pile of mortgage settlements.

I guess you could ask, well, OK, but how much money did those banks make manipulating that $5.3 trillion foreign exchange market? I don't know! No one seems to care. The U.K. Financial Conduct Authority says "that it is not practicable to quantify the financial benefit" that each bank got from its manipulations; the U.S. Commodity Futures Trading Commission and Office of the Comptroller of the Currency don't even acknowledge that the question might be interesting.

But you can sort of gesture at an answer. For instance you can look at the profits mentioned in the orders. So take Citigroup, the worst offender by sheer amount of fines at $1.02 billion (so far!). Here's its FCA order. It contains exactly one example of an attempt to manipulate a foreign exchange fixing (paragraphs 4.38 to 4.44). "Citi's trading in EUR/USD in this example generated a profit of USD99,000," on a trade with a notional amount of about 542 million euros. The Citi trader made that money by teaming up with his chat-room buddies to manipulate a fix. Here's how they reacted:

Subsequent to the ECB fix, Citi’s trading was variously described by other traders in chat rooms as “impressive”, “lovely” and “cnt teach that”. Citi noted “yeah worked ok”.

Citi did some other bad stuff, but that's the only example in the FCA order with an actual profit figure. The CFTC and OCC orders for Citi have no profit figures at all. So we have one case of Citi making $99,000. The FCA order covers conduct over a six-year period, though that might be a bit generous. Figure 250 trading days a year and you get $150 million of profits for Citi if every single day was as "impressive" and "lovely" as the day singled out by the FCA as, one assumes, a particularly egregious example of manipulation. That's not the case: The CFTC order, for instance, recounts a fixing in which "the Citibank trader reported that he was 'hosed.'" The manipulation was not foolproof, and fools abounded. But take $150 million as an upper bound, and Citi paid about an order of magnitude more in fines than it made in manipulative profits.

So what was this manipulation? Well, there are two big fixes in the G10 foreign exchange markets, the 4 p.m. WM/Reuters fix and the 1:15 p.m. ECB fix. The fix is a benchmark that a lot of customers use, making them relatively price-insensitive: If they can transact at the fix, they don't care that much what the fix is. So each bank gets a bunch of orders from customers to transact at the fix. Some customers want to buy (say) euros at the fix; some want to sell. So the bank nets the buyers with the sellers and ends up with some net position. Say its customers mostly want to buy euros from it at the fix. Then it has to buy those euros from the market, to remain flat. It can do that before or after or during the period at which the fix is set. Its goal is to buy the euros from the market at a lower price than it sells them to the customer.

Generally speaking, if you have a lot of euros to buy in the next five minutes, you might think that the price of euros will go up over those five minutes. If those five minutes are from 1:10 to 1:15, and the fix occurs at the end, then that is a profit opportunity: Your buying will push the price up over those five minutes, say from $1.3216 to $1.3222, but you will buy at the average price over those five minutes (roughly) and then sell at the final price (the fix). So you'll make a couple of hundredths of a cent per euro.

Here are some things to say about that. First of all: It's fine. That's the most important thing. That thing I just described in that last paragraph, it's fine. The FCA acknowledges that explicitly: As part of legitimately risk management of client orders, you're going to trade around the fix, and that's going to affect the fix. That's how it works. If clients want to buy a lot of euros at one time, that's going to push the price of euros up at that time. There's nothing wrong with that.

Second: That thing I just described two paragraphs ago, it's risky and uncertain. I said, "If you have a lot of euros to buy in the next five minutes, you might think that the price of euros will go up." But that's only true if no one else has a lot of euros to sell in those five minutes. If you know that you're buying, and you don't know who's selling, you will feel a bit constrained in your ability to push around the price. You can't be sure that the price at which you sell (the 1:15 fix) will be higher than the average price over the previous five minutes. Maybe someone else's selling will push it down. So you'll be a bit more cautious. You won't buy a ton of euros in the last minute before the fix to really push the price up. You might even save some of your buying until after the fix, since that will be better for you if the price really does go down.

That, fundamentally, is what these settlements are about. In simple form, what the banks did here was take the risk out of their risk management. They'd get into chat rooms -- yes, yes, with dumb names like "the 3 musketeers" or "1 team, one dream" -- and just tell each other whether they were buyers or sellers. Once they knew, then all the mystery was gone. If their customers were all buying euros at the fix (i.e. the banks were selling), then they'd want as high a fix as possible, and would buy inefficiently just before the fix so as to push the price up. If they all had customer sell orders, vice versa. And if it was a mixed bag, they'd sort of figure out in advance which side predominated so that that side could make the most money.

You're not supposed to do this, because it's just unsporting. You can trade in advance of the fix -- even if it pushes the price in your favor -- as a matter of legitimate risk management of client orders. But once you know what everyone else is doing, your risk goes away. Now you're just pushing around the price to make money. You're trading inefficiently in a way that moves the price, secure in the knowledge that you can't lose from your inefficiency. That's not how it's supposed to work.

But be careful not to exaggerate this. Here's how the CFTC describes what the chat room traders would do, in a passage that repays careful study:

If traders in the chat room had net orders in the same direction as what they desired rate movement at the fix to be, then the traders would at times either (1) match off these orders with traders outside of the chat room in an attempt to reduce the volume of orders in the opposite direction transacted during the fix period; (2) transfer their orders to a single trader within the chat room who could then execute a single order during the fix period; or (3) transact with traders outside of the chat room to increase the volume traded by chat room members during the fix window in the direction favored by the private chat room traders. At times, traders also increased the volume traded by them at the fix in the direction favored by the chat room traders in excess of the volume necessary to manage the risk associated with their banks’ net buy or sell orders at the fix.

Focus on the numbered alternatives. Let's say that the chat room traders are selling euros to customers at the fix, so they want a high fix. They want to buy a lot of euros in the few minutes right before the fix, to push the price up. They know their buddies are on board with that plan. They don't know if other banks are on board with it though. So they have three alternatives:

  1. Try to net off with outside banks (buy from them at the fix), so that the chat room traders have fewer euros to buy, but those outside banks have fewer to sell.
  2. Just hang out in the chat room and don't deal with outside banks.
  3. Try to increase their positions using outside banks (sell to them at the fix), so that the chat room traders have more euros to buy in those last few minutes, but the outside banks have more to sell.

Notice that 1 and 3 are opposites! Banks could further their manipulation by buying from outside banks, selling to outside banks, or doing neither. This should make you suspicious. All of those things can't work equally well! If buying from other banks would push the price up, or down, then selling to them should push the price down, or up. The fact that the chat room traders sometimes did one, and sometimes the other, means that they hadn't found a reliable cheat, a way to take the risk out of their trading. It means in some sense that their manipulation didn't work. I mean, it worked fine. But there's a reason you "cnt teach that." It makes no sense!

Try to account for the flows here:

  • The chat room traders had a lot of client buy orders at the fix.
  • They decided to manipulate the price up.
  • They did that either by increasing the amount of buy orders they had at the fix (by reducing the buy orders that outside banks had), or by decreasing the amount of buy orders they had at the fix (by increasing the buy orders that outside banks had), or by doing neither and just putting all their buy orders into the hands of one of them.
  • At the fix, the price went up.
  • Sometimes.
  • Except sometimes it didn't, and then the chat room traders were "hosed."

I submit to you that no economic activity happened in these chat rooms. The price went up because the chat room traders had a lot of customers who wanted to buy euros at the fix, and no one else had customers who wanted to sell euros at the fix. When the price didn't go up -- when the chat room traders were "hosed" -- it's because other banks did have customers who wanted to sell euros at the fix. The price went up when there were more buyers than sellers, and not when there weren't. The market worked.

Now this isn't strictly true, not at all. These guys really were cheating. By knowing in advance what their buddies were doing, by concentrating orders in one set of hands, by feeling out the non-chat-room banks, the traders in their chat rooms did get an unfair advantage that allowed them to feel comfortable trading in inefficient ways that did push the price around more than pure supply and demand would dictate. But this isn't Libor manipulation, where they just made stuff up. This is tinkering at the edges of real supply and demand. And it helps to explain why their profits seem to have been relatively modest and inconsistent. There was only so much they could do. And sometimes they just got it wrong.

(Also they took out customer stop orders. I guess that's manipulation? I assume that all dealers everywhere are taking out customer stop orders all the time, but, um, FX dealers definitely were! )

On the other hand, the fines are huge. Partly this is because the market is so big: Making $99,000 by pushing around an FX rate by a few pips could cost lots of other people millions of dollars. Partly I suspect it's just a hangover from Libor: You guys, again? Back so soon?

Partly it's that these are basically morals offenses, and they are, you know, offensive. There are the traders who discuss bringing another trader into their chat room, asking whether he would "add huge value to this cartell" or would instead "mess this up and sleep with one eye open at night." There are the traders who "dont want other numpty's in mkt to know" about what they get up to in their chat room. There are the traders at Citi who used "informal and sometimes derogatory code words to communicate details of clients' activities" to other banks. Sadly the FCA gives no examples of the derogatory code words, though one presumes that they were worse than "numpty."

I guess that's how bank settlements work now? These guys in these chat rooms sure sound bad. And they sure intended to manipulate the FX markets. It's much less clear how much damage they actually did, or even could have done. But that's a secondary question. After Libor, terrible chat room conversations are enough in themselves to justify billion-dollar fines.

  1. That's from page 3 of the CFTC settlements, e.g. this one, as well as page 6 of the FCA settlements, e.g. this one.

  2. Specifically, the current settlements are with the U.K. Financial Conduct Authority, the U.S. Commodity Futures Trading Commission, the U.S. Office of the Comptroller of the Currency and the Swiss Financial Market Supervisory Authority. When I mentioned the settlements this morning, the Office of Comptroller of the Currency had not yet weighed in. Now it has, to the tune of $950 million, and it's picked up Bank of America too, which seems to have been unscathed by the CFTC and FCA but which paid $250 million to the OCC.

    Bank of America coughed up $250 million to the OCC, for doing the same sort of stuff as the other five banks. But it seems not to have settled with the FCA or CFTC. But we know that it set aside at least $400 million to settle the FX cases due to discussions over the last few weeks. So that gives you at least some sense of how much is still to come. Bloomberg News lists the U.S. Justice Department, the Federal Reserve, the U.K. Serious Fraud Office and the New York Department of Financial Services as still interested. And of course plaintiffs' lawyers.

  3. When I quote, I'm mostly quoting from a form that is applicable to all the banks. It's e.g. page 34 of the JPMorgan one.

  4. So call it like one and a half basis points. I guess if you do that every day it's pretty good.

  5. Assume a lot of sics in all the quotes. These are from broker chat rooms. You know what to expect from broker chat rooms.

  6. There's no real discussion of how long the misconduct lasted. The FCA uses a "Relevant Period" from "1 January 2008 to 15 October 2013." The OCC says "2008 through 2013." The CFTC says "2009 through 2012," which I suspect was an effort to be accurate.

  7. To be fair, other banks had better days than that. RBS made $615,000 on one fix (paragraph 4.44 here). One bank in a UBS chat room made $220,000 on a fix (page 7 here). On the other hand, some did worse; JPMorgan's most manipulative example made it $33,000, and like Citi's, its CFTC order suggests it was more hosed than hosing.

  8. There are better explanations of the fix mechanics in Annex B of the FCA orders; if you're lost I recommend reading those.

  9. Those numbers are based on paragraph 4.41 of the Citi FCA order, though I am stylizing everything else.

  10. FCA paragraph 3.4:

    A firm legitimately managing the risk arising from its net client orders at the fix rate may make a profit or a loss from its associated trading in the market. Such trading can potentially influence the fix rate. For example, a firm buying a large volume of currency in the market just before or during the fix may cause the fix rate to move higher. This gives rise to a potential conflict of interest between a firm and its clients.

    We've talked about this a bit before.

  11. The FCA's explanation of the alternatives is in some ways more dynamic:

    (a) Net off these orders with third parties outside the chat room, thereby reducing the volume of orders held by third parties that might otherwise be transacted at the fix in the opposite direction. Traders within the market have referred to this process as “taking out the filth” or “clearing the decks” or similar;

    (b) Transfer these orders to a single trader in the chat room, thereby consolidating these orders in the hands of one trader. This potentially increased the likelihood of successfully manipulating the fix rate since that trader could exercise greater control over his trading strategy during the fix than a number of traders acting separately. Traders within the market have referred to this as “giving you the ammo” or similar; and/or

    (c) Transact with third parties outside the chat room in order to increase the volume of orders held by them in the desired direction. This potentially increased the influence of the trader(s) at the fix by allowing them to control a larger proportion of the overall volume traded at the fix than they would otherwise have and/or to adopt particular trading strategies, such as trading a large volume of a currency pair aggressively. This process was known as “building".

  12. The Citi FCA order has a good explanation of how that worked at paragraphs 4.11 to 4.13, 4.45 and 4.46. It ends:

    This behaviour was reflected in language used by G10 spot FX traders at Citi in chat rooms. For example, a Citi trader referred in a chat room to the fact he “had to launch into the 50 offer to get me stop done”. On another occasion, a trader at Citi described in a chat room how he “went for a stop”.

    I was out of the office this morning and read that FCA order in a hard copy, on which I highlighted that paragraph and wrote "haha obvs."

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:
Matthew S Levine at mlevine51@bloomberg.net

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