Every financial contract is subject to a bunch of risks, and in some sense each of those risks affects its value. There’s some chance that an asteroid will crash into the earth next year, rendering your 30-year interest rate swap considerably less valuable, and if you’re so inclined you can discount its value for that possibility.
One nice thing to imagine is that your financial contract is, like, one contract, and all the risks are spelled out in that contract, and you can figure out the value of the contract based on real or market-implied probabilities of all the risks happening etc., and you add them all up and you conclude “the market value of this contract today is 12!” or whatever and you go on your merry way. But that doesn’t need to be true. Some of your risks live in the contract and are part of the contract; some live in the counterparty and have to do with the counterparty’s riskiness; some live in whatever collateral arrangements you have with the counterparty and have to do with the mechanics of your collateral; some are asteroids.1
Anyway, remember the Deutsche Bank whistleblower story? I said last week that the question of whether DB’s actions constituted accounting fraud was not a particularly interesting question, but that is all relative and you’d be surprised what I find interesting. One thing I find interesting: those Deutsche Bank trades! And umm their accounting.
So, some background. As far as I can tell, DB sold a bunch of credit protection in sort of normal ways, CDX and stuff. And it bought a bunch of protection in leveraged super senior tranches. A super senior tranche, classically, is:
- You have a pool of reference assets,
- You pay some spread to a protection writer,
- If defaults wipe out more than some unlikely-seeming percentage – 15%, say – of those assets, then the protection writer gives you money, more or less 1% of notional for every 1% of losses over that threshold,
- So for instance if there are 40% losses you get paid 25%.
- The protection writer is like a big bank or monoline or whatever and, in 2005, is either AAA/AA or is posting mark-to-market collateral or both.
So there’s your trade. A leveraged super senior is the same thing, except replace that last bullet point with:
- The protection writer posts a bunch of collateral – 10% of max exposure, say – day one.
- The protection writer is a Canadian asset-backed commercial paper conduit or some other non-credit party.2
- If certain bad things happen that make you worry that you don’t have enough collateral, you can ask the protection writer to post more collateral, but (1) they don’t have to, (2) they don’t want to, and (3) they can’t.3
So the same thing only different. The basic claim here is that Deutsche Bank started out treating its leveraged super seniors as almost like regular super seniors (although with a 15% haircut to roughly account for the fact that they were non-recourse beyond the initial collateral), and ended up treating them as exactly like regular super seniors (with no haircut at all), with the result being that their assets, capital, etc. were all inflated compared to how they’d look if DB had treated these trades properly. This claim, to be fair, comes from whistleblowers some of whom weren’t at DB at the time of the trades, and is of course disputed – Deutsche disagrees, and their valuations seem to have been signed off on by their regulators at BaFin.
Still, if it’s true: that seems bad, no? Here I drew you a picture:
And then I drew you a more worrying picture:
The claim is that DB pretended they’d bought a put (the first picture – if credit gets worse than X, the contract covers their losses), but actually bought a put spread (the second picture – if credit gets worse than X, the contract covers their losses only up to X + 10%).4 Those pictures, intuitively, should have very different values.5
So how bad is that? It strikes me that there are about two ways to think about this.
First, maybe this trade is a super senior tranche – i.e. a put, the first picture – plus some wonkery in the credit support documents that might make it look in practice in some states of the world like a put spread, the second picture. In this theory the trade is a super senior tranche – the protection writer owes DB the full amount of losses, in some abstract sense of the word “owes” – but it might be hard for DB to collect. Here is a DBRS note on leveraged super seniors that says “although they feature leverage, the leverage is located outside the structure, as opposed to being an integral feature of the structure (such as in CDO squared transactions).” That is sort of a weird thing to say! To say it you need to imagine that:
- There is The Thing – the stuff “inside the structure” – which is a regular super senior tranche, and
- There is Outside The Thing, such as the non-recourse collateral arrangements, which basically say “oh never mind you don’t actually have to come up with all that money.”
If you believe that then it’s just about imaginable to value The Thing as a super senior tranche and then put on a surprised look when someone asks you “well what about the Outside The Thing that says the counterparty doesn’t actually need to pay?” “I didn’t think about that,” you say, “I was just valuing The Thing.”
This sounds stupid, I know, but it has an analogy in credit valuation adjustments in regular derivatives. If you buy a derivative, you can calculate the value of the cash flows in the contract called the “derivative.” You can also worry about, well, will my counterparty actually pay me if it turns out he owes me a ton of money under the “derivative”? And so there is a whole apparatus around (1) trying to make sure the counterparty will pay you (collateral! credit support annexes!) and (2) figuring out what the odds are that he won’t, and how much that costs (CVA!). The derivative is The Thing; the CVA is Outside The Thing. The CVA is enough of A Thing of its own that there are all sorts of rules about how to account for it, but it’s worth noting that those rules are not just “the CVA is part of the value of a derivative and changes in CVA are in all respects identical to changes in market prices, etc.”
Perhaps of interest for our purposes: the way CVA works, roughly speaking, is that if your derivative receivable is less than your collateral, you have no CVA. If the mark-to-market today is $10bn and you have $11bn of collateral, you can show $10bn of assets with no CVA adjustment, no matter how shady your counterparty is. If the mark-to-market doubles to $20bn and you can’t get more collateral, then the extra $9bn needs to be CVA adjusted: you can’t just pretend that you’ll get all of that $9bn.
You can imagine a similar thought process on the leveraged super seniors: DB’s counterparties were shady, perhaps, but they’d collateralized the full initial exposure and hten some. And some rough math suggests that the collateral actually did more than cover Deutsche Bank’s actual exposure even at the height of the crisis, meaning that Deutsche was never owed more than it could cover out of the collateral.6 On a CVA-ish view of the world, you could almost get comfortable treating these trades as though they were regular super senior trades. (And, cynically, you could imagine traders convincing auditors, used to a CVA-ish view of the world, to get comfortable with that treatment even if the traders thought it was dubious.)
The second way to think about it is, you could say: that’s ridiculous. The trade is intended and designed and built to be a put spread, the second diagram. No one could possibly have thought that if 100% of the reference names defaulted, some ABCP conduits would stump up another $120bn to make Deutsche Bank whole.7 Therefore you couldn’t value the trade the same was as you’d value a super senior tranche, and if you did, especially as credit spreads widened and the world seemed closer to ending, you were clearly up to no good and possibly up to fraud. That’s what Eric Ben-Artzi’s allegations sound like: if Deutsche people were in fact running around making its accounting for these tranches more aggressive as the crisis got worse, that suggests that they were doing something other than having an intellectually honest consideration of how to value features of the tranches and related credit documentation.
Anyway! Let us again leave for another day (1) what happened and (2) whether it was fraud. Let us also leave aside the fact that the realized losses were extremely low, well below the collateral amounts,8 meaning that there was no difference in outcome between the leveraged super senior that DB bought and the regular super senior that it maybe pretended it had. The point is just that one of these views of the world is “right,” from a like theoretical-finance mark-everything-to-market perspective; the other is intellectually displeasing and stupid and messy, but also not all that surprising in comparison with all the legal and operational and regulatory gaps and differences and weirdnesses that actually exist in the world. I don’t know who’s right here, but you could see why the disagreement might get pretty heated.
Earlier: Deutsche Bank Ignored Some “Losses” Until They Went Away
1. And other features of physics, law, the world, etc. Like: your swap doesn’t have an “Asteroids” section, but either it has a “force majeure” or “impossibility” or similar section, or some rules for dealing with such a situation are implied by ISDA documents and/or background principles of contract law, etc. etc. Anyway the result is, your counterparty will stop paying you after the asteroid hits. Also there are other results.
2. The big FT article said they were Canadian pension funds but this seems unlikely; the pension funds were big investors in the conduits’ paper but probably not direct credit counterparties to Deutsche.
3. Let’s unpack those because they’re fun:
- “Certain bad things”: the trade has a trigger such that if the trigger is hit, the protection buyer (DB) can call for more collateral, and the seller (Canadian conduit) can either post more collateral or unwind the trade at fair value. The bad things can be actual realized losses (a minority of trades), actual mark-to-market on the tranche, or some sort of spread mark-to-market on the index. Here is a very useful primer, via Lisa Pollack.
- They don’t have to: it’s non-recourse! The contract allows them to just unwind instead of posting more. If the unwind burns through more than their collateral, they’re not on the hook.
- They don’t want to: this is debatable, I guess, but the theory is that when the trade moves against you you should unwind it (at fair value – you don’t automatically lose your collateral) and enter a new trade at a higher spread, rather than just put more at risk in this trade. So you should never post. This is a very Black-Scholes way of looking at the world – in the actual world where the trigger is hit, (1) an unwind is likely to be expensive and blow through your collateral and (2) you may not get to enter a more-attractive trade elsewhere – but is a useful rough cut.
- They can’t: they’re Canadian ABCP conduits and so if they wanted to post more collateral they’d have to go out and raise it in the ABCP market. Which, when they needed to post more collateral, was frozen.
4. Possibly worse than that, because of the difference between mark to market and actual payout on defaults. So here you can read about a lawsuit between Barclays and a Canadian conduit called Devonshire. There’s a lot going on here, but for our purposes, Devonshire put up $600mm of collateral on a leveraged super senior, Barclays claimed at trigger was hit and that the mark to market was $1.2bn (so Devonshire would have to put up another $600mm to keep the trade on), and the judge took a look, saw that realized defaults were $12,000 (twelve thousand!) rather than $1.2 billion, and so made Barclays give Devonshire its $600 million back. So Barclays may have thought it had a $600mm put spread but in fact it had mostly a nothing.
5. Depending on your inputs. If the world can never be worse than so-so to great, then the trades don’t really have different values. That, obviously, is the thinking behind putting these trades on in the first place.
6. Rough math goes like:
- ~7-8 years duration
- Spread moves in DB’s favor by ~100-125bps
- = DB’s mark-to-market is 7-10% of notional
- DB has collateral of 10% of notional
Note that in that Barclays case above the collateral was only ~50% of the mark to market, so this rough math could well be off.
7. This is somewhat academic as no one thought that 100% of the reference names would default.
8. Again I can’t resist that Barclays deal where they had a $1.2 billion mark to market and twelve thousand dollars of actual losses.