There are two questions worth asking about today’s Wall Street Journal story about how Deutsche Bank “made at least €500 million ($654 million) in profit in 2008 from trades pegged to the interest rates under investigation by regulators world-wide”:

  • is that a lot?, and
  • did they do it by manipulating Libor?

The second one is hard, huh? Here’s the Journal:

[A] former employee has told regulators that some employees expressed concerns about the risks of the interest-rate bets, according to documents. He also said that Deutsche Bank officials dismissed those concerns because the bank could influence the rates they were betting on.

A Deutsche Bank spokesman said those allegations were “categorically false.”

So, who knows; Yves Smith says “unless the source can provide some sort of supporting evidence, this is ‘he said, she said,’ and the matter will shake out in the German bank’s favor.” I sort of come at this from the other direction, which is:

  • Every other bank has mountains of emails and IMs to the effect of “hey we’re gonna go mess with Libor don’t tell anyone.”
  • Deutsche isn’t, like, the #1 most-careful-with-emails-and-IMs bank in the history of banks.
  • So, totally possible that supporting evidence will float up, no?

Also totally possible that these were legitimate trades unrelated to a few bad apples at DB who were admittedly manipulating Libor, of course. But where is the fun in that?

So let’s talk about the first question. Per the Journal, Deutsche had €68 million of DV01 to various Libor bases (basises?), making money if USD/EUR/GBP 1-to-6-month curves steepened. And it ended up making some €500 million on those trades in 2008. Is €500 million a lot for an interest rate trading book in 2008?

It sort of depends on your denominator. Deutsche’s total FICC trading revenues in 2008 were €124 million, so these trades are like 4x its total profit, but that was affected by some €8 billion in crisis-related write-downs and credit trading losses. But “[t]hese losses were partially offset by record results in Foreign Exchange, Money Markets and Commodities, where customer activity remained strong.” So rates + currencies + commodities made at least, like, €8.124 billion that year; the €500 million on Libor steepeners is a healthy contribution but not an outsized one.1

But here’s a denominator I like. From Deutsche’s 2008 annual report:

Just to focus on the biggest line, Deutsche had €32 trillion of notional amount of interest rate swaps, with a gross market value of around €1.1 trillion and a net market value of €19 billion. And its one-day 99% interest-rate value-at-risk was, at year-end, €130 million.2

Let’s do some fake math. Pretend that all these swaps are Libor swaps, and that their average duration is 2 years. That means that one basis point of Libor moves would lead to €6.4 billion of changes in market values of those swaps. The trades the Journal is writing about move by €68 million per basis point, or just over 1% of that. Compared to DB’s total interest-rate-derivative book, these trades were tiny.

Is that a stupid denominator? Of course it is. Deutsche Bank’s trading book is not “go buy €32 trillion of Libor swaps”; it is, roughly speaking, “go buy €16 trillion and sell €16 trillion.”3 You can get a sense of this from the market value breakdown: Deutsche has €570 billion of positive-valued positions and €550 billion of negative-valued swaps, meaning that its net position is somewhat under 2% of its gross position. If you just multiply – that is, if you assume that Deutsche Bank’s book is pretty uniform about duration, fair-value-to-notional, etc., but leans about 2% in one direction or another – then you get about €111 billion of “expected” DV01 in Deutsche’s interest-rate book.4 Which makes the trade highlighted by the Journal look pretty big, in that it seems to be taking more than half the interest rate risk you’d naively expect from Deutsche’s whole rates trading book.

But here’s what’s always struck me as weird about Liborgate. That €6.4 billion number that I mentioned earlier isn’t entirely fake: one basis point in moves in (every) Libor really should have cost someone €6.4 billion, and made someone else €6.4 billion, through changes in value of swaps on Deutsche Bank’s books. Of course none of those someones were Deutsche Bank: Deutsche just sat in the middle, exposed to let’s say less than 2% of those moves.

Yves Smith thinks today’s Deutsche allegations are plausible, asking “who would take such a big bet then if they weren’t confident they had some sort of advantage?” Again, though, I come at it the opposite way: if your business is based on manipulating rates, why are you running a matched book? There’s an intuitive plausibility to the Journal‘s basic tale of (1) bank put on big bets, (2) risk managers fretted, (3) they were reassured by traders saying “well we’ll just manipulate Libor so this bet pays off for us.” But if that was really the thinking, why not do it all the time? Why go through the effort of laying off 98% of your interest rate risk, building a mostly balanced book of long and short swaps, instead of just leaning really hard into bets on Libor going up, say, and then working the phones hard to push it up?

That more or less describes the career of UBS’s Tom Hayes, whose trading strategy seems to have been “make directional bets on my ability to manipulate Libor, then do it,” but most of the Libor manipulation elsewhere seems to have been considerably less ambitious: not “let’s put on a lot of directional bets and then manipulate Libor so they come true,” but rather “oh hey I ended up long a bit of Libor in my client trading last week, could you manipulate it a smidge so I make some money?”

I dunno, it’s sort of disappointing, isn’t it? Betrays a lack of drive, or imagination, or confidence in their ability to manipulate Libor. Also, weirdly, it will work out worse for the banks now. If smoking-gun evidence emerges that DB was manipulating Libor systematically, it will rue the fact that it was long and short almost the same amounts of Libor. Every €111mm, say, that DB made in 2008 will give rise to €6.4 billion in damages now, making it not really worth it in hindsight.

Bank Made Huge Bet, and Profit, on Libor [WSJ]
Former Deutsche Bank Employee Claims Bank Took Big Libor Bets During Crisis Because It Could Influence Rates [NC]

1. That’s from page 16 of Deutsche’s 2008 annual report. Nor is it that extreme versus other banks. Goldman FICC made $3.7 billion in 2008, with at least $6 billion of credit and mortgage losses (page 87 here), meaning that rates + currencies + commodities made about $10 billion combined.

2. The VaR is page 98 of the annual report. But that reminds me: what is going on with this? From the Journal:

At the time, the German bank didn’t include the interest-rate bets in the main measurement of trading risks, known as value-at-risk, or VAR. People familiar with Deutsche Bank’s thinking said it was common practice to exclude such bets, and including them would have lowered the company’s overall VAR.

In 2009, Deutsche Bank made “significant methodology improvements” to the value-at-risk, adding to the measure the interest-rate risk related to “different money-market instruments and swaps based on them,” according to a securities filing by the company.

See page 83 of the 2009 annual report, which actually lists among the methodological improvements “Inclusion of basis risk between different money market instruments and swaps based on them.”

Can someone explain that to me? Like, were the interest rate swaps that the Journal mentions really not included in VaR calculations at all? Or is it just that the basis nature of the trades wasn’t included? Like, was the VaR calculated only for parallel shifts rather than for the sort of steepening that DB was betting on? This seems very strange to me but perhaps I’m missing something.

3. I hate myself for the “buy/sell” terminology here but whatever.

4. That fake math is in section (2) of my little spreadsheet. Section (3) tries to tie these numbers to VaR; if you think that DB has €111mm of Libor DV01, you assume that’s all to USD3mL, and you make some assumptions about volatility, then you get about €840mm of interest-rate VaR, or 6.5 times the right number of €130mm. So, y’know, Fake Math Is Fake. Though! Remember, DB’s VaR was kind of fake too.

Comments (12)

  1. Posted by Guest | January 10, 2013 at 12:53 PM

    I think you're missing smth in arriving at your conclusion, Matt. Deutsche was actually involved in all the same IBOR shenanigans as everyone else. However, they were doing it properly and with real edge, rather than just punting. Deutsche historically had always had a massive short end franchise, based on a them being a proper European "money center" bank. Sorta like the JP of Europe. As a result, the flows that the short end desk saw were truly ginormous. That, in turn, implied that they were able to get into the trades they wanted (ones where they could "massage" the outcome) by utilizing end-user flow, rather than punting and paying bid-offer in the market. So they would have run their book mostly matched (collecting healthy bid/offer in the process, obviously), but with the ability to get into prop trades where they were confident about the outcome. It's a trader's dream, really. One of the ex-Deutsche names that has been mentioned was one of the biggest, smartest and most successful short end traders ever. Whether he is proven guilty in this IBOR stuff is unclear, of course.

  2. Posted by Fact Checker | January 10, 2013 at 12:53 PM

    then you get about €111 billion (sic) of “expected” DV01 in Deutsche’s interest-rate book.4

  3. Posted by guest | January 10, 2013 at 1:18 PM

    All very erudite, but tell me more about the "massage" part.

    ~J. Epstein

  4. Posted by alebro | January 10, 2013 at 5:00 PM

    swap traders make profits from the credit spread / liquidity premium they charge. in 2008 this was around 20bps.

  5. Posted by curve guy | January 10, 2013 at 5:20 PM

    Matt, re your VaR question in footnote #2:
    these are not simple curve steepener trades but rather bets that the forward curve based on 1M rates and the forward curve based on 3M rates would move apart. Like, a 3M rate would not be equal to compounded 1M forward rates any more. This has nothing to do with what kind of shifts you apply to your yield curve in your VaR model (I'm sure they had all kinds of PCA factor based shocks) but rather that what you thought before was a single Libor curve has fallen apart into many different Libor curves, one for each tenor.
    It's possible that their VaR model did not capture tenor basis risk at that time (even though tenor basis risk was around at least since Aug 07)

  6. Posted by curve guy | January 10, 2013 at 5:41 PM

    also, not much to see here really — tenor basis swaps was the name of the game at that time, every rates desk did exactly the same as DB

  7. Posted by guerst | January 10, 2013 at 10:20 PM

    When did this site get so legitimate? Am I lost?

  8. Posted by guerst | January 10, 2013 at 10:23 PM

    I MEAN LOOK AT THESE COMMENTS EVEN

  9. Posted by D.O.U.G. | January 11, 2013 at 7:48 AM

    Matt. Keep up the good work. Subject to the insane amount of unreadable footnotes and the fact you sometimes write like a 12 year old girl from Orange County, I learn something new every day from your market musings and actually look forward to reading these posts.

    D.O.U.G.
    (Nott Matt's Mom, but a fan of her work)

  10. Posted by Dana Caffrey | January 25, 2013 at 8:43 AM

    Interesting article. I've learned something from you. Thanks for sharing this information. Interest rate swaps play a very important role in the economy.

  11. Posted by sohbet | May 12, 2013 at 6:27 PM

    telling these little stories, here's a good idea: Have a point. It makes it so much more interesting for the listener.

  12. Posted by sohbet | July 8, 2013 at 9:25 PM

    officers to investigate Blackstone's fishy financial transactions, but they wound up giving them