How it actually works
A lender writes a few thousand loans, sells them into a , and the vehicle issues bonds backed by the loan payments. Those bonds are not identical. They are stacked, and the stack is the entire product.
Cash coming in from borrowers goes down the stack from the top: the most senior bondholders are paid first, then the next class, then the next. Losses go up the stack from the bottom: the most junior piece absorbs them until nothing is left of it, then the class above starts absorbing, and so on. Everything below you is your cushion. The thickness of that cushion, expressed as a percentage of the pool, is your subordination.
Take an illustrative deal — invented numbers, round for clarity. A $1bn pool of car loans is financed with $780m of Class A notes, $70m of Class B, $70m of Class C, and an $80m residual piece that the originating lender keeps. The Class A holder has 22% subordination beneath them. If cumulative losses on the pool come in at 6%, three-quarters of is gone and Class C has not been touched. At 10%, the residual is wiped out and Class C has lost nearly a third of its money. Class A is still whole and still, on paper, AAA. The loans did not get better; the loss was simply routed elsewhere by contract.
Two things usually sit alongside subordination and do the same job. Overcollateralisation means the pool is bigger than the bonds it backs, so there is more loan than debt from day one. Excess spread is the gap between what borrowers pay (high, because subprime) and what the bonds pay (lower), which absorbs losses monthly before any principal is impaired. Together these are called credit enhancement, and the ratings agencies will tell you exactly how much of it a given rating requires.
The word also travels outside securitisation. Subordinated bank debt ranks behind deposits. A holding company's lenders are structurally subordinated to the operating company's, because the operating company's creditors get paid before anything reaches the parent. And in a , subordination becomes a weapon: a group of lenders amends the documents so that everyone outside the group is pushed down the stack. Same mechanic, worse manners.
Why it matters to this crash
Subordination is the answer to the question we posed on 19 August 2026: how are sixty-day delinquencies on US loans running at 6.9%, above the 5.0% peak of the 2008 crisis, while the lenders' shares sit near record highs?
Because the equity of a subprime lender is not a claim on the loans. Most of the credit risk was sold into securitisations. What the shareholder owns is the origination and servicing fee stream, plus the retained residual — and the residual only takes losses after the ABS buyers' subordination is exhausted. Rising delinquencies eat the residual before they touch the stock. Meanwhile high delinquencies come with high yields, and a lender that can still originate is earning them.
That is not irrational pricing. It is the market correctly reading a waterfall. The relevant question stops being "are the loans bad" and becomes "whose tranche is bad, and can they take it". On generally, this is the pattern we keep finding: the borrower's distress is visible and the loss-bearer is not.
What would make this dangerous
Cumulative pool losses running through the thickness of subordination on recent vintages — not delinquency rates, which are an input, but realised losses versus the enhancement disclosed at issue. That is the point at which mezzanine tranches, sold as investment grade, stop being investment grade.
Downgrades of outstanding subprime auto ABS classes, which force ratings-constrained buyers — insurers, some funds — to sell into a market that already knows why.
The new-issue market repricing, so that a deal needs materially more subordination to clear. This is the transmission channel back to the equity. If the originator has to retain a fatter residual to sell the senior notes, the trade gets less profitable per loan, and if the ABS market shuts entirely the origination and servicing fee stream — the actual asset behind the share price — stops growing.
And the thing we cannot yet see: who holds the mezzanine. We have not covered where the junior tranches of US subprime auto deals have ended up. Until someone does, the honest position is that the losses are landing somewhere identifiable in principle and unidentified in practice.