Structured Credit Explained: ABS, CMBS and CLOs

In two thousand six, one credit rating agency stamped its highest grade, triple-A, on nine thousand and twenty-nine mortgage securities. More than thirty a day, every working day.

Thirty a day? That's not analysis. That's a conveyor belt.

A Senate investigator used almost that word. He called Moody's a triple-A factory. Nearly forty-three thousand mortgage bonds got the grade between two thousand and two thousand seven. Eighty-three percent of the two thousand six batch were later downgraded.

So the machine was broken. Case closed.

That's the thing. The SAME machine, pointed at different loans, built bonds that lost almost nothing. Point one percent. Same lawyers, same agencies, same slicing. One blew up the world. The other barely flinched.

Okay, hold that. Before you tell me why... what IS the machine? Like, actually.

Simpler than it sounds. You take thousands of loans. Mortgages, car loans, credit card balances. You sell them into a shell company that exists only to hold them. Then you sell claims on the money those borrowers pay each month.

So I'm not buying loans. I'm buying a seat at the payments.

And the seats are ranked. That's the whole trick. Picture a fountain. Cash pours in at the top, fills the senior bondholders, overflows to the next level, then the next. At the bottom sits the equity tranche. They get whatever's left.

And when borrowers stop paying, the bottom goes dry first.

Losses eat upward. The equity is wiped out before the triple-A loses a cent. Which is how you take a pile of loans nobody would call safe and manufacture something that looks safe on top.

Manufacture. You chose that word on purpose.

It's the honest one. Pool home loans, that's RMBS. Office towers and shopping centres, CMBS. Car loans, credit cards, student debt, that's ABS.

All deeply boring.

Which is why the frontier is weirder. Investors now buy bonds backed by music royalties. Aircraft leases. Timeshares. Leases on data centre space.

Somebody's pension is funded by streaming royalties.

Somebody's insurance policy, more likely. And then the last one. CLO. Same fountain, but the pool is corporate loans. Debt to mid-sized companies, below investment grade.

Junk, basically. And THAT'S the one that didn't blow up?

July tenth, two thousand seven. Moody's does something it had never done. It downgrades three hundred and ninety-nine subprime mortgage bonds issued the year before. Three months later, another two thousand five hundred and six tranches.

In three months.

And here's the part that still makes me wince. Wall Street took the leftover slices, the risky middles nobody wanted, pooled THOSE, and re-sliced them into fresh triple-A bonds. A CDO. Sometimes a CDO of CDOs.

No. You cannot make something safe by collecting more of the unsafe thing.

You can on paper, if you assume the loans fail independently. One family in Nevada defaults, bad luck. The models assumed Nevada and Florida and Ohio weren't really connected.

Except they were all connected to one thing.

National house prices. When those turned, everything turned at once. Triple-A CDOs issued before two thousand eight recorded roughly three hundred and twenty-five billion dollars of losses. Long-run loss rate on investment grade CDO notes... thirty-four percent.

Thirty-four. On something sold as nearly risk-free.

Now the CLOs. About five hundred billion dollars of American CLOs issued between ninety-four and two thousand nine. Point eight eight percent of tranches defaulted. On Moody's rated deals, triple-A and double-A... zero.

Zero? Through the worst crisis in eighty years?

Zero. Loss rate around point one percent. Versus thirty-four.

Then what's different? From where I'm sitting it's the same fountain.

It's what you pour into it. A CLO holds real loans to real companies, senior, secured, spread across industries. A subprime CDO held slices of other securitisations, every one of them riding the same housing market. One was diversified. The other only looked it.

So the structure didn't fail. The food did.

That's the lesson nobody puts on a poster. Slicing doesn't create safety. It moves it around. CLOs also have a manager who can trade, and coverage tests that shove cash upward when the pool starts rotting.

Still. Zero feels too clean. Nothing is zero.

Good instinct, and the sceptics say exactly that. In early two thousand nine, research shows senior CLO coverage ratios fell from around one point one six to zero point eight five.

In English?

Even the most senior notes were briefly under-collateralised. They survived on recoveries, and on the fact that the structure can't be forced to sell into a panic. Survival. Not invincibility.

And how big is the thing now?

It bottomed out at two hundred and sixty-three billion after the crisis. As of April twenty twenty-five, Bank of America puts it near one point four trillion. CLOs own roughly sixty percent of the entire American bank loan market.

Sixty percent? So CLOs basically ARE the lending market.

For those companies, close to it. And the mix is drifting. Middle-market CLOs, backed by smaller private loans, used to be ten to twelve percent of issuance. Recently as high as twenty-four.

Smaller companies, less public information, barely traded. That's where I'd start sweating.

That's where the whole argument lives. The structures have been tested, mostly in good weather. The record is real... it's also short, and the collateral keeps getting stranger. Data centres. Royalties. Fibre.

So, the opening question. Why one and not the other.

Because a triple-A was never a property of the bond. It's a bet on what's underneath, and whether those things fail together. In two thousand six that bet was placed thirty times a day, by people paid by the person asking for the answer.

And my mortgage, your car loan. They're sitting inside one of these right now.

Almost certainly. You're not just a borrower. You're a line in somebody's waterfall... and the only question that ever matters is who's standing below you when the water runs out.