Felix Salmon

I’ve written myself into a corner, now, and can’t think of any way to get out of writing the promised blog entry on super-senior tranches. Especially when Kevin Drum asks so nicely. So here it is. Deep breath…

By now, you understand how a synthetic bond can behave very much like a real bond. So consider the situation of a bank, which has made a bunch of loans, to 100 different companies. The companies all value their relationship with the bank, and the bank values its relationship with the companies. At the same time, however, the bank would like to free up some capital. It doesn’t want to sell the loans outright — so instead it creates a synthetic bond referencing those 100 credits, and sells that.

Essentially what the bank is doing is taking the interest payments from the companies it’s lent money to, and using them to make insurance payments against those companies defaulting. If the companies default, the buyers of the synthetic bond end up sending money to the bank, which will offset its loan losses. The bank has brought down its credit exposure to those companies even though it hasn’t sold the actual loans. And because its credit risk has come down, its capital requirements have come down too, and the bank has more free capital to use elsewhere.

Because the loans are still on the bank’s books, it needs to take mark-to-market write-downs on those loans if they fall in value but don’t default. On the other hand, when that happens the value of the bank’s default insurance is almost certain to rise by a very similar amount. So the bank really has managed to construct a pretty good hedge here. Not perfect: no hedge is perfect. But pretty good.

So far so boring. But of course banks are never happy with simple hedges: they want to make money from all this financial high technology. (Incidentally, it’s not clear that they won’t: Alan Kohler had a thought-provoking column a couple of weeks ago saying that once a few more big defaults happen, "a mass transfer of money will take place from unsuspecting investors around the world into the banking system. How much? Nobody knows, but it’s many trillions.")

In any event, let’s go back to the synthetic bond that the bank issued. Let’s say it’s structured so that each of the companies is paying an identical amount in interest every year: call it $1 million each. If the bank bundled up all those loans into a collateralized loan obligation, or CLO, then the CLO would be paying out $100 million a year, unless or until one of the companies defaulted.

But rather than just sell the CLO outright, the bank would most likely split it up into tranches. Companies default, but they don’t all default at once. If the companies in question were all investment grade, you could be sure that at least 80 of them would still be making interest payments at any one time. So if you sell off the right to the first $80 million of interest payments, the ratings agencies will slap a triple-A rating on that income stream, and it can be sold at a tight spread and a pretty high price. Then the next $5 million might have a double-A rating, and the next $5 million a single-A rating, and the next $5 million a triple-B rating, and the last $5 million will either have a junk rating or else just be considered "equity".

Now it’s possible that the bank will contrive to make a small profit here, if the sum of the value of all the tranches is greater than the amount of money that the bank lent out in the first place. But the operative word is small.

How do things change if the bank issues a synthetic bond rather than a cash CLO? Well, it can sell off the equity and the junk and the single-A and the double-A tranche, thereby protecting itself if interest payments fall by $20 million. It can then sell off a bit of the triple-A tranche, protecting itself if payments fall by $25 million. Now remember that the first $80 million of payments are rock-solid, risk-free: that’s why they carry triple-A ratings. So the bank’s remaining risk, after selling off that triple-A-rated synthetic tranche, has been brought down to safer-than-triple-A levels. Some of the banks referred to it as a "quadruple-A" risk, although that’s not a real-world rating. But the banks were so comfortable that defaults at that level could never happen that they didn’t feel any need to hedge themselves against it happening.

Janet Tavakoli, back in 2003, published a nice little table of the difference between a cash CDO and a synthetic one:

There are nuances and differences here that we don’t need to worry about too much. But the main thing to notice, in this example, is that a bank could protect itself against the first 13.5% of a group of bonds defaulting, and then declare that it was fully hedged: even the triple-A tranche had been sold off, and all that remained was a risk-free super-senior tranche.

Clearly, the cost of protecting yourself against 13.5% of a group of bonds defaulting is lower than the cost of protecting yourself against 100% of that group of bonds defaulting. Not a lot lower, since no one really imagined that more than 13.5% of the bonds could ever default. But enough lower that the bank could end up making a nice profit by selling off the credit risk associated with the bonds, and holding on to the excess income.

Of course, the bank’s loan position is not actually fully hedged. But so long as more than 86.5% of the interest payments get paid, the bank is fine. And the advantage of leaving the rest of the loan portfolio unhedged is that you don’t need to use all the income from the loans to buy protection on them. There’s money left over — which can be considered interest on the quadruple-A, or super-senior, tranche that the bank retains.

The invention of the super-senior tranche, then, was a way of letting banks have their cake and eat it too. They could take a bunch of debt onto their balance sheets, "fully" hedge it (with only that it-could-never-default tranche left over) and book all the remaining cashflow as pure profit with no credit risk.

Now so long as you’re dealing with a hundred different investment-grade corporate loans, this actually works: such things really don’t all default at the same time. Banks even found a (limited) market for these super-senior tranches, by allowing their hedge-fund clients to take very leveraged bets on them — they felt that the leverage was safe, since there was no default risk. But a huge proportion of the super-senior tranches was never sold off to hedge funds, and instead remained on the banks’ balance sheets.

And of course it wasn’t long until the banks started doing the same thing with mortgage bonds. They would take a bunch of subprime-backed CDOs, and treat the interest payments from the CDOs much as they treated the interest payments from corporations.

Oops.

In the case of CDOs, as we’ve all seen, the models said that there was geographical diversification in the housing market; they said that national housing prices, in aggregate, never went down. (Which was true, until it wasn’t.) And somehow they said that thanks to the magic of securitization and overcollateralization, you could create not only triple-A securities from triple-B-rated subprime assets, but even quadruple-A super-senior tranches, too.

And so the banks took billions of dollars of subprime-backed mortgage securities onto their books, and "fully" hedged them while not really hedging most of them at all. When the income from those CDOs went (or was expected to go) towards zero, the banks had to write off assets which they never really considered risk assets at all. The banks thought that they had hedged all the credit risk associated with the bonds, and were left with a modest and super-safe income stream. Instead, they were left with a time bomb.

It’s worth noting here that although the bank used CDS technology in the process which ended up with these super-senior tranches, it’s not the credit default swaps themselves which blew up. Remember that the bank had a lot of CDOs on its books, and the credit default swaps helped protect the bank from a lot of the losses associated with those CDOs. Just not all of them. And if you underwrite hundreds of billions of dollars’ worth of CDOs, the "unfunded" portion of those CDOs can become enormous on an absolute level — and if it gets wiped out, you can get wiped out.

As ever, credit default swaps, like any derivative, were and are a zero-sum game: they don’t cause big losses themselves. The super-senior losses, ultimately, come from the subprime mortgage market, not from the CDS market. But without the technology of credit default swaps, banks would never have been able to retain those exposures while thinking that they had divested themselves of all the associated risk.

Still, no amount of regulation of the CDS market would have solved the underlying problem, which really had nothing to do with the credit default swaps themselves, and everything to do with the banks’ risk models. Those models said that if you take on this risk and sell that risk, you’re fully hedged. They were wrong. The CDS, so far, have worked. The investors who bought the higher-risk tranches of the synthetic CDOs have been wiped out — the banks, essentially, have taken nearly all their money. If the CDS contracts hadn’t worked (if, for example, the government decided "to simply annul the credit default swaps as void", as Ben Stein has proposed), then the banks would have lost even more.

I’m not saying that the CDS market protected the banks from losses: without it, the banks would never have taken all those mortgage-backed CDOs onto their books in the first place. They never thought of themselves as being in the storage business; they thought they were in the moving business. But without senior management ever really realizing it, banks like Citi ended up storing hundreds of billions of dollars’ worth of subprime bonds on their balance sheet, and failed to properly account for them because they erroneously thought those bonds were risk-free.

Ultimately, then, the error was one of management, not of financial technology. The banks’ balance sheets — and those of their off-balance-sheet vehicles — were expanding faster than the banks’ executives and risk managers could really keep a handle on. And rather than call a halt to that which they didn’t fully understand, they handed down edicts instructing the CDO desks to keep on dancing for as long as the music was playing. Most of the executives probably never even heard the term "super-senior" until those tranches started getting written down. It was their own incuriousness, rather than any CDS technology, which was really their undoing.