The AAA rating on a CLO tranche says almost nothing about the quality of the loans underneath it. It says that the loans are unlikely to fail together, and that the people who recover money from failed loans will get enough back. Those are bets on correlation and recovery, and the second bet has been quietly getting worse for a decade, partly because CLOs themselves keep buying.
The winter of 2008, when the loans crashed and the senior notes held
In December 2008 the broad US leveraged loan market traded at roughly 60 cents on the dollar. These loans are floating-rate debt owed by companies rated below investment grade, usually private-equity buyouts. Default rates climbed through 2009 to around 10% of issuers. By any price-based measure, the asset class had collapsed.
The collateralized loan obligations holding those loans did not collapse with it. A collateralized loan obligation (CLO) is a fund that buys a portfolio of 150 to 300 leveraged loans and pays for them by issuing a ladder of bonds, called tranches, plus a slice of equity. The senior tranches of CLOs from that era traded down hard on fear. Their realized principal losses, measured once the deals paid off, were minimal. Mortgage CDOs from the same period were wiped out. CLOs owned corporate loans that also fell sharply in price, and they came through largely intact.
The common explanation is that corporate loans are "better" than subprime mortgages. That misses the point. The loans were marked at 60 cents. What saved the CLO was how it was built: what it was forced to sell, what it was never forced to sell, and how losses were allowed to reach each layer. For a reader trying to judge CLO risk today, the lesson from 2008 is to study the structure first and the loan-quality story second.
How subordination turns B-rated loans into AAA notes
Subordination means some investors agree to absorb losses before others. In a CLO, cash coming in from the loan portfolio flows down a waterfall. The AAA notes are paid interest first, then AA, then A, and so on down to the equity, which gets whatever is left. Losses travel the other direction. They hit the equity first, then the lowest-rated debt, and reach the AAA only after everything beneath it is gone.
Here is an illustrative $500 million CLO with a capital structure typical of recent deals:
- AAA notes: $310 million (62%)
- AA notes: $55 million (11%)
- A notes: $30 million (6%)
- BBB notes: $30 million (6%)
- BB notes: $25 million (5%)
- Equity: $50 million (10%)
The AAA holder sits on top of $190 million of junior capital, or 38% of the pool. That figure is the attachment point, the share of the portfolio that must be lost before the AAA loses a dollar.
Now convert losses into defaults. A loss is the defaulted amount times the share you fail to recover. First-lien leveraged loans have historically recovered something like 60% to 70% of par. At 65% recovery, each defaulted dollar costs 35 cents.
Defaults needed to impair the AAA = Attachment point / (1 - Recovery rate)
At 65% recovery: 38% / 0.35 = 108.6%. More than the whole pool would have to default. At 40% recovery: 38% / 0.60 = 63.3%. Nearly two-thirds of the portfolio, cumulatively, before the top tranche is touched.
The equity sees it differently. Its cushion is zero. At 35% loss severity, cumulative defaults of 28.6% erase the $50 million, ignoring the excess interest the deal earns along the way. That is why equity investors are paid like equity investors.
The practical reading: the AAA rating is a statement about the attachment point divided by loss severity. If you want to know how safe a tranche is, track recovery assumptions at least as closely as default forecasts.
Why the structure survived a price crash: locked funding and par-based tests
Subordination explains who loses money first. It does not explain why 2008 failed to force CLOs into the fire sale that destroyed so many leveraged vehicles. The answer is the funding.
A CLO's notes are term funding. They are issued once, carry a stated final maturity of roughly 12 years, and cannot be called away by the noteholders when prices fall. There is no repo lender asking for more margin and no fund investor redeeming shares. A hedge fund holding the same loans on borrowed money in late 2008 got margin calls and had to sell at 60 cents. The CLO kept the loans, kept collecting interest, and waited.
The second piece is how the deal measures its own health. CLOs run coverage tests. The overcollateralization (OC) test compares the par value of the loans to the par value of the notes at each level. The interest coverage (IC) test compares interest collected to interest owed. Crucially, OC tests count most loans at par, their face value, not their market price. A performing loan trading at 75 cents still counts as 100.
Market prices enter only at the edges. Loans rated CCC above a limit, commonly 7.5% of the portfolio, are carried at market value instead of par, and defaulted loans are carried at an assumed recovery. When a test fails, the waterfall diverts cash that would have gone to equity and uses it to pay down the senior notes. The structure deleverages itself from the top, using its own income, without selling anything at distressed prices.
The economic actors explain why this works. The CLO manager wants to keep collecting management fees, which are paid from the waterfall and partly subordinated. Equity holders want cash flow to resume, so they prefer a temporary diversion to a liquidation. Senior noteholders get repaid faster when tests fail. Every party with a vote is incentivized to hold rather than sell. This is the one place intuition gets it wrong: people assume a vehicle full of junk loans is fragile in a crash, when the CLO's design is specifically a refusal to be a forced seller.
What this means for you: a CLO tranche's mark-to-market price and its credit risk are two different things. The first reflects liquidity and fear; the second reflects whether losses can actually reach it through the waterfall.
The permanent bid and the erosion of covenants
Now turn the machine around and look at what it eats.
CLOs are the single largest buyer of US leveraged loans, with commonly cited estimates putting their share near two-thirds of the market. A CLO is built to stay invested. During its reinvestment period, typically the first four to five years, the manager must recycle every repayment into new loans. New CLOs launch whenever the gap between loan spreads and liability costs makes the equity return attractive. The result is a buyer that shows up regardless of what the loan documents say.
Look at what the CLO's rules actually constrain. The tests and portfolio limits care about ratings (through a weighted average rating factor), diversity across industries, the CCC bucket, single-name concentration, and spread. Covenants are absent from the list. A maintenance covenant is a promise by the borrower to keep a financial ratio, such as debt to earnings, within a limit every quarter; breach it and lenders get to renegotiate early. A CLO's compliance tests give no credit for that protection and impose no penalty for its absence.
So when private-equity sponsors and their banks propose loans without maintenance covenants, the largest buyer has no structural reason to refuse. Competition among managers for scarce loans, especially when new CLO issuance is strong, pushes them toward accepting whatever terms clear. Covenant-lite loans went from a niche product before 2008 to the overwhelming majority of new institutional issuance, commonly reported near 90%. The demand for the output, AAA notes that pension funds, insurers, and banks want to own, loosened the standards on the input.
This was not a moral failure by managers. It is a pricing mechanism. The arbitrage that creates a CLO rewards spread and rating, so those are the variables that get defended in negotiation. The variables nobody is paid to defend erode first.
Where the machine breaks: correlated defaults and falling recoveries
Return to the formula. The AAA is impaired only when defaults, multiplied by loss severity, exceed the attachment point. Both inputs can break.
Correlation is the first risk. The 108% figure above assumes defaults are spread out over time, with interest income replenishing the cushion between them. If a single shock pushes a large share of borrowers into trouble at once, the subordination is consumed faster than excess spread can rebuild it. Diversity tests help, but they diversify across industries, and a recession, a rate shock that hits every floating-rate borrower, or a sector collapse concentrated in a few large industries can make diversified portfolios behave like concentrated ones. In spring 2020, rating agencies downgraded loans so quickly that CCC buckets overflowed their limits, junior OC tests failed across a large share of deals, and equity distributions were diverted. The senior notes were fine. The equity learned what correlation feels like.
Recovery is the second risk, and it connects directly to covenant erosion. Covenant-lite loans let a struggling company deteriorate longer before lenders gain leverage, so by the time a default happens, less value is left. Loose documents also enable liability management exercises, transactions where a borrower moves collateral or reorders creditor priority. J.Crew's transfer of intellectual property to an unrestricted subsidiary in 2016 and Serta Simmons's 2020 uptier, which elevated a group of lenders above others in the same loan, became templates. A CLO that ends up on the wrong side of such a deal recovers far less than the 65% its history suggested.
Rework the AAA math at 30% recovery: 38% / 0.70 = 54.3%. The threshold is still extreme, and AAA notes remain heavily protected. The cushion beneath the mezzanine is a different story. A BB tranche attaching at 10% now fails at 14.3% cumulative defaults instead of 28.6%. Lower recoveries do not threaten the top of the structure first; they compress the middle and the bottom, where most of the investors chasing yield sit.
The chart compares the two ends of the machine. BKLN holds leveraged loans directly; JAAA holds AAA CLO notes. During a stress episode like the April 2025 tariff selloff, the loan fund absorbs the price shock while the AAA fund barely registers it. That gap is subordination and term funding showing up in prices.
What to do with the mechanism
A CLO converts weak loans into strong notes by stacking investors in order of loss and by refusing to sell in a panic. The machine fails when defaults cluster or recoveries collapse, and its own steady demand has weakened the covenants that protect recoveries.
For a reader evaluating CLO exposure, three practical habits follow. Judge any tranche by its attachment point divided by loss severity, and stress the severity, not just the default rate. Treat AAA notes and BB notes as different asset classes that happen to share a portfolio: the first is a bet that correlation stays moderate, the second is a bet that recoveries hold up. And when you read that loan documentation has loosened, recognize that the bill arrives through recoveries, years later, and lands on the lower tranches first.
