Deutsche Bank Ignored Its CDS Problems Until They Went Away
Today the Securities and Exchange Commission fined Deutsche Bank $55 million for accounting misstatements relating to some crisis-era credit derivatives, and reading the SEC order makes me feel like a peasant in the Dark Ages discovering a classical Greek statue in a pile of rubble. They had Art, back then, in 2005 or whatever. Their credit derivatives had a beauty, a symmetry, a purity of line and a depth of feeling that we're unlikely to see again in our lifetimes.
- Some people (basically Canadian pension funds; we'll call them "investors") put some money in a pot (called a "conduit").
- The pot basically just put the money in the bank.
- Then the pot also sold credit default swaps to Deutsche Bank: The pot agreed to use its money to pay Deutsche Bank if there were defaults on certain corporate bonds.
- These swaps were "super senior," meaning that they would only pay off if just enormous numbers of investment-grade companies defaulted on their debts.
- The swaps were also "leveraged," meaning that if there was $1 billion in the pot, the pot might write credit protection on $11 billion worth of bonds.
- The investors could get their money out of the pot on short notice: The pot was funded with commercial paper, so every 30 or 60 or 90 days or whatever, the pot's debt came due and investors could choose to take their money out.
- But the pot had already pledged all its money to Deutsche Bank as collateral for the credit default swaps, so if investors wanted their money back, where would the pot get the money from?
- Once answer is: The pot could find new investors to put in money to replace the old investors (by remarketing the investors' maturing commercial paper).
- If the pot couldn't find new investors, though, it could go to Deutsche Bank for the money: The conduit paid for a liquidity put from Deutsche Bank, in which Deutsche Bank was obligated under certain circumstances to buy commercial paper that the conduit wasn't able to remarket.
This is a magical trade, and Deutsche Bank ended up doing almost $100 billion worth of it. (That is: About $8.5 billion worth of investor money, in about 30 pots, providing credit protection on $98 billion of notional. ) Part of the magic is just the general magic of securitization, which manufactures value out of nothing more than bankers' ingenuity and ratings agencies' mistakes. Deutsche Bank could buy cheap credit protection from the conduit, which it could then re-sell to hedge funds and make a profit. The investors in the conduit got short-term highly-rated commercial paper that paid a higher interest rate than similar paper that wasn't, you know, constructed with such a high degree of artistry. And there was still money left over, between what Deutsche Bank paid for protection and what the pot paid to investors, for the people running the pot to make money for themselves.
But there is also a specific magic to this trade, which is: Deutsche Bank was writing guarantees to the thing that was writing it guarantees. That is, it was buying credit protection from this conduit, which would pay out if things went very wrong in the credit markets. But it was also guaranteeing liquidity to this conduit, meaning that if things went wrong for the conduit, Deutsche Bank would have to give the investors in the conduit their money back. Meaning that if things went wrong all around, Deutsche Bank might be getting its credit protection from a pot full of Deutsche Bank's own money.
That's not even what the SEC case is about. It's just a beautiful fact. To me, the highest form of derivatives artistry is achieved when a bank sells derivatives on its derivatives to itself, and this trade comes alarmingly close to that. But in fact, Deutsche Bank's liquidity guarantee was more or less an illusion, and when the Canadian asset-backed commercial paper market froze up, Deutsche Bank (and other banks in similar positions) refused to buy back the commercial paper. Ultimately there was a big settlement, the investors' short-dated commercial paper was poofed into long-term bonds in 2009, conditions returned to normal, and people who bought the bonds in the depths of the crisis made a lot of money. As you'd expect: These conduits were basically writing mega-disaster insurance to Deutsche Bank, and the global credit crisis, though a pretty big disaster, turned out to be not all that bad for investment-grade corporate credit defaults. The super-senior mega-disaster insurance never paid off.
Anyway though the SEC case! Notice that these trades were leveraged super senior tranches: The conduits wrote $98 billion worth of credit protection to Deutsche Bank, but only put up $8.5 billion of collateral. If the losses on the $98 billion worth of bonds were greater than $8.5 billion, Deutsche Bank would have to go back to the conduits and ask for more money. One assumes that the conduits would, at that point, say no, as they had the right to do.
More to the point, though, these trades had a mark-to-market value: Similar credit default swaps traded in the market, and so you could get a sense, from market prices, of the expected value of the protection that the conduits had written to Deutsche Bank. And if the market value of the credit protection got high enough, Deutsche Bank could go to the conduits and ask for more money. Roughly speaking, the trigger was about 50 percent of the collateral, so if the market value of the credit protection got to more than about $4.25 billion, Deutsche Bank could ask the conduits to raise more money (by selling more commercial paper to investors). If a conduit said no, Deutsche would terminate its trade at the market value and take some or all of the money in the pot.
In fact, market prices anticipated very high levels of defaults, the mark-to-market value of this protection got very high, Deutsche Bank did call for more collateral, and as part of the big settlement investors did put a lot more money in the pots, roughly doubling the total to $16.6 billion. This was a rational decision for them: Although there were big mark-to-market losses on these positions -- at their peak, the expected value of the pots' liabilities to Deutsche Bank exceeded $10 billion -- in the end, the actual liability was zero. In late 2008 and early 2009, market prices implied that there would be tons and tons and tons of investment-grade corporate defaults. There were not.
But the point is that the Canadian investors in the conduits didn't have to put up more money. There were circumstances in which it would be rational for them to do so -- like if they thought that the mark-to-market losses would reverse themselves and they ultimately wouldn't have to bear huge losses. There were circumstances in which it would not be rational for them to do so -- like if credit conditions actually became really horrible and they would just be throwing good money after bad. In fact, the circumstances in which Deutsche Bank would really need the insurance provided by these contracts -- where credit conditions were really bad -- are exactly the circumstances in which the conduit investors would probably walk away. So while Deutsche Bank had bought insurance on $98 billion worth of debt, it didn't get the full value of that insurance. If it had $1 or $2 or $5 billion of losses, the insurance would cover that. If it had $10 billion of losses, the insurance might cover it, or might fall short, depending on the investors' decisions. If it had $30 billion of losses, no way.
This is called "gap risk." And here's how Deutsche Bank accounted for it:
From the end of 2007 through 2008, in a time frame slightly over one year, Deutsche Bank employed five different approaches to measuring the Gap Risk in valuing the LSS Positions. These methodologies included taking a percentage haircut on the value of the LSS Positions, applying a static valuation adjustment, and using a model. However, each change in methodology had the effect of decreasing the amount that Deutsche Bank assigned to the Gap Risk in its LSS Positions during this time period.
Now that happens to have been a period of substantially increasing credit risk. Logically, the worse credit conditions got, the bigger the gap risk was. But on Deutsche Bank's accounting statements, the gap risk kept going down. Bad work! E.g.:
Initially when Deutsche Bank entered into the trades, Deutsche Bank measured the Gap Risk by taking a 15% haircut from the value of the LSS Positions. For the first quarter of 2008, Deutsche Bank froze the haircut at its December 31, 2007 amount and used a static valuation adjustment of $200 million for the Gap Risk.
A fixed percentage of the value makes a rough kind of sense: The more money Deutsche expected to get back from this insurance, the more it should have haircut that value for the gap risk. A fixed dollar amount does not make sense. But in fact it was even worse than that: As the value of Deutsche Bank's insurance kept going up, the haircut for gap risk mostly went down, from $200 million to $20 million, then back up to $78 million, then frozen at $78 million until the fourth quarter of 2008, when it went to zero. Deutsche decided to just forget about it.
The SEC finds this ... fishy? And, right? "Deutsche said in its statement on Tuesday that it did not account for the risk because it 'did not believe there was a reliable method for measuring the gap risk in light of the existing market conditions,'" and while that may have been true it's not a great reason to mark it at zero. The SEC notes that Deutsche Bank did value the gap risk for other purposes, including for negotiating leverage in the big Canadian commercial-paper settlement in December 2008. It used two different models and came up with values of $1.5 billion and $3.3 billion, suggesting that (1) the measurement was indeed very fuzzy, but (2) zero was probably the wrong answer.
Deutsche also "maintained that it did not suffer any losses": The realized gap risk was zero, so who is the SEC to say that Deutsche's valuation of zero was wrong? That is only a very imperfect defense. One problem with it is, come on, lots of derivatives end up with zero realized value; that doesn't mean that their value was zero throughout their lives. An even bigger problem, though, is that the realized value of Deutsche Bank's purchased credit protection was also zero. It's not just that the leveraged super senior tranches never paid out more than their collateral (in which case Deutsche would have had losses from the gap risk). The leveraged super senior tranches never paid out at all. But Deutsche Bank's financial statements showed a positive market value of over $10 billion for those tranches at their peak during the credit crisis. Now they're running off toward zero.
Now that $10 billion wasn't, like, free profit. These trades were not just a way for Deutsche Bank to get massively short investment-grade credit. Deutsche Bank bought tons of credit protection from these conduits, but Deutsche Bank is just an intermediary. It presumably sold credit protection to offset what it bought from the conduits: Hedge funds who wanted to bet against corporate credit could buy credit default swaps from Deutsche Bank, which were hedged using the leveraged super senior protection from these conduits. The problem was that Deutsche Bank was fully liable on the CDS that it sold: If there were 100 percent defaults, Deutsche Bank would have to pay out 100 percent of that CDS. But it wasn't fully protected on the CDS that it bought: If there were 100 percent defaults, it would probably only get back 9 or 10 or 17 percent or whatever of that CDS.
There weren't 100 percent defaults, of course, but as conditions got worse, the expected value of the credit protection that Deutsche Bank bought and sold diverged. That is not a problem that anyone would want to deal with in late 2008, so Deutsche Bank just didn't. It ignored the divergence; it pretended that the skimpy credit protection that it bought was just as good as the real credit protection that it sold.
Which is kind of the story of securitization! The pots of money at the heart of this trade were created to give investors highly-rated, short-term, super-safe securities with a higher yield than you'd get elsewhere. They did that by paying the investors for a few weird mismatches: The one-sentence description of the security is that it matures in less than a year and can't lose money unless there's a massive wave of investment-grade corporate defaults, but the description of why that sentence isn't quite right runs to 385 pages. On the other side of the trade, Deutsche Bank fell into a similar mismatch: The one-sentence summary of this trade was that it was a cheap way for Deutsche Bank to get $98 billion of super-senior credit protection. The reality wasn't quite so simple, even if Deutsche Bank kept pretending that it was.
This description draws on this restructuring proposal from the Pan-Canadian Investors Committee for Third-Party Structured Asset-Backed Commercial Paper, and the attached JPMorgan report. Pages 16-26 of the proposal ("Background on Securitizations and Asset-Backed Commercial Paper," etc., pages 34-44 of the PDF) are particularly helpful.
Here I am (somewhat loosely!) describing synthetic Canadian ABCP conduits, which kept their money in things like bank medium-term notes, bankers' acceptances, cash, etc. Those things are supposed to be cash-like enough that I'm just calling them cash, though obviously in 2008 one might worry. Pages 51-71 of the JPMorgan report attached as Exhibit D to the restructuring proposal lists the actual collateral for the various deals. Some of the deals are cash deals, rather than synthetic, which is not relevant to the Deutsche Bank situation.
Again that JPMorgan report describes the trades. So in for instance the "Aurora" deal on page 54 (page 335 of the PDF), "Trade ID" 4 is a $125 million leveraged super-senior tranche on the CDX index, attaching at 30 percent losses and detaching at 70 percent losses. So that trade would pay off for Deutsche Bank only if more than 30 percent of the investment-grade companies in that index defaulted. Other trades (e.g. Trade ID 2) were on bespoke baskets of investment-grade corporate borrowers, again with high attachment points.
The SEC says:
Initially, the trades were leveraged approximately eleven times which meant that the Canadian counterparties posted collateral of approximately 9% of the notional value of the trades or approximately $8.5 billion.
But this is an average over lots of trades; casual inspection of the JPMorgan report turns up leveraged super seniors levered as little as 6 times and as much as 40 times.
The SEC says:
The LSS trades at issue had a notional value of C$120 billion, or approximately $98 billion, reflecting credit protection Deutsche Bank purchased from Canadian counterparties. Initially, the trades were leveraged approximately eleven times which meant that the Canadian counterparties posted collateral of approximately 9% of the notional value of the trades or approximately $8.5 billion.
Deutsche Bank purchased approximately $98 billion of leveraged super senior tranches of over thirty CDOs (collectively, the “LSS Positions”), representing over 50% of the Canadian LSS market.
The pot also paid a fee for Deutsche Bank's liquidity put, which sort of conceptually reduces the amount Deutsche Bank paid for credit protection. The restructuring proposal (page 19) says that "These fees commonly range from 2 to 13 basis points." Deutsche's 2007 annual report says that it had "earned fees for the liquidity facilities and puts of € 9 million during 2007," or call it about $12.5 million of fees on $8.5 billion of liquidity provision, or about 15 basis points.
From the restructuring proposal (page 19):
In the U.S. and many other jurisdictions, funding under liquidity agreements ("global-style" liquidity agreements) is not predicated upon the satisfaction of "market disruption" conditions, although funding is generally not required in respect of defaulted assets. By contrast, Canadian liquidity arrangements in the ABCP market have traditionally been structured such that funding is made available to issuers only in the case of a general disruption in the Canadian ABCP market. In a typical "Canadian-style" Third Party ABCP liquidity facility, there are a number of conditions that must be satisfied before the facility provider is obligated to advance. Two key conditions are (a) the rating agency must affirm the original credit rating of the outstanding ABCP, and (b) the issuer's inability to sell new ABCP must result from a general disruption in the Canadian commercial paper market. In combination, these conditions are designed to ensure that the inability to issue new ABCP relates to a general disruption in the Canadian commercial paper market rather than the creditworthiness of the issuer, its assets or its ABCP.
This sensibly breaks the circularity that I refer to in the text: Deutsche Bank shouldn't have to buy the conduit's commercial paper in conditions in which the credit protection that the conduit sold to Deutsche Bank would likely pay off. On the other hand, it somewhat insensibly makes the liquidity protection kind of worthless.
Standard & Poor's reports investment-grade default rates for 2007, 2008, 2009 and 2010 of: zero percent, 0.41 percent, 0.32 percent, zero percent. Obviously some issuers were downgraded after going into indexes and before defaulting, but still. Even the high-yield default rate stayed under 10 percent. Deutsche Bank's investment-grade leveraged super seniors almost all had attachment points of at least 15 percent.
For simplicity I'm just using aggregate numbers, though of course there were separate triggers for each conduit, and different conduits referenced different indexes, tranches, etc., so it's not like an all-or-nothing $4.25 billion thing. Here's the SEC:
Deutsche Bank had the ability to request that additional collateral be posted at certain specified “trigger” points. The original triggers at which Deutsche Bank could call for additional collateral occurred when the value of the LSS Position increased to a set percentage of the posted collateral, typically around 50% (“Original Triggers”). For example, if one LSS Position had a notional amount of $1 billion, and was 10 times leveraged with $100 million in collateral, Deutsche Bank could call for additional collateral to be provided by the protection sellers when the value of Deutsche Bank’s LSS Position reached $50 million. At that point the protection sellers could choose to post additional collateral or not. If the protection sellers posted additional collateral, which would allow them to avoid an unwind, the LSS Position would remain in effect. If the protection sellers decided not to post additional collateral then the transactions would unwind, and Deutsche Bank would keep the lesser of the posted collateral or the fair value equivalent of a fully-collateralized LSS Position.
From the SEC:
The Montreal Accord benefitted Deutsche Bank by making available additional collateral. Specifically, the total collateral for Deutsche Bank’s LSS Positions increased from approximately $8.5 billion to approximately $16.6 billion, plus additional margin funding from Deutsche Bank of approximately $2.0 billion.
In just one year, from December 31, 2007 to December 31, 2008, the value of Deutsche Bank’s LSS Positions, driven by the disruption in the markets, had quadrupled in value from $2.63 billion to $10.65 billion. By the end of the first quarter of 2009, the LSS Positions were valued at $9.72 billion.
Note that these high values seem to have occurred after the collateral restructuring was at least in process, if not finalized. (It was finalized in early 2009.) It's a bit odd to value some credit-default swaps written by conduits at $10.65 billion if there's only $8.5 billion of money in the conduits, and no one's obligated to put up any more. I mean, a bit odd, not totally out of line, but a bit odd.
This interacts with the fact that (1) the conduits did put up some more collateral, reducing the gap risk, but (2) the triggers were renegotiated to be less favorable to Deutsche Bank, increasing the gap risk. I ignore that here but the SEC has a lot to say about it.
Though one that I was somewhat seduced by years ago. I don't necessarily stand by all of that post, though I have raided it for its headline.
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 firstname.lastname@example.org
To contact the editor on this story:
Zara Kessler at email@example.com