Subprime auto ABS are judged, deal by deal, on the losses in their monthly servicer reports. For the largest issuer, about half of SEC-registered subprime auto ABS outstanding in 2024-25, those numbers leave something out. Santander's trusts default early no more often than Exeter's, but under a clause CCM found in no other public program, Santander buys a third to half of those loans back at par. Its deals' first-year losses would be a quarter to a third higher if they were counted. Bondholders are protected by it; anyone comparing underlying loan performance has to account for it. And as Santander folds its consumer lender into its bank, the promise has moved with it, though the loans behind it are leaving public view.

CCM looked at every buyback reported by 475 SEC-registered auto loan trusts in 22 programs from January 2017 to September 2026: $2.5 billion in all.^[1] Twelve of the 22 programs buy back less than 0.01% of their balance a year. Santander's two subprime programs, Santander Drive (SDART) and Drive Auto Receivables (DRIVE), buy back 0.84% and 1.68%: together 76% of all buyback dollars, on 18% of the pool balance (Exhibit 1).

Exhibit 1

What gets bought back. Elsewhere a buyback is mostly an administrative matter: a current loan the servicer modified beyond what the deal allows, which most deals require it to purchase. At Santander nearly three-quarters of the dollars are early defaults, loans 30 or more days past due that had made at most one payment in the trust. Credit alone doesn't explain the gap. AmeriCredit, GM Financial's subprime program, has the same median credit score as Exeter and buys back 0.03% a year.

Who keeps the early defaults. Santander's pools don't default early more often than its peers'. Across all deals with at least a year of history, 3.3% of SDART's opening pool and 4.1% of Exeter's reached 60 days past due in the first year having made at most one payment. Exeter charges almost all of them off in the trust. Santander buys back about a third of SDART's and close to half of DRIVE's (Exhibit 2).^[2]

Exhibit 2

Smaller than it was. Santander's buybacks peaked early: its two programs bought back $448 million in 2018, 3.7% of their balance, and since 2020 the rate has run at 0.4% to 0.9% a year. Better borrowers explain little of it. Among loans sold within two months of origination, the share defaulting early in SDART's 2022-25 deals (2.9% to 4.0%) is no lower than in its 2017-19 deals; DRIVE's has eased, from about 6% to about 4% to 5%. Most of the drop came from the end of a second stream of modified, current loans in late 2019, and from pools sold more seasoned, past the first two payments the clause covers. The pandemic briefly emptied the pipeline. In April 2020 alone, more than one in six of the loans in Santander's trusts were granted a payment extension, and a third by June.^[3] Buybacks fell by about 90% that spring and summer, consistent with payments deferred rather than missed, before recovering in 2021.

Exhibit 3

One clause. Every sponsor makes representations about a loan's payments when it sells it into a trust. All but one describe the loan as it stood at the cutoff date: not more than 30 days past due, or, at CarMax, that the borrower "has made at least one payment". Santander represents that the borrower "has made, or will make, the first two monthly payments" (Exhibit 3). If the borrower misses either payment, the representation is breached, and Santander must cure the breach or repurchase the loan at its balance plus interest. On EDGAR the wording appears in 351 filings, all Santander's, starting in May 2017.^[4]

The clause arrived two months after Santander settled with the Massachusetts and Delaware attorneys general, who had alleged that it funded loans from dealers that inflated borrowers' incomes; it did not admit wrongdoing.^[5] Early payment default is the classic sign of a misrepresented application, and the clause moved that risk from bondholders to Santander.

Does it make a difference? For bondholders, yes, in their favor: the buyback is at par, so the trust bears none of those losses. For anyone reading the deals' loss numbers, it matters too. Count the early-default buybacks as charge-offs, net of recoveries at each deal's own loss severity, and SDART's first-year cumulative net loss, across all its deals, rises from 3.0% to 3.7%, and DRIVE's from 4.5% to 5.9%, a quarter to a third higher; other programs barely move (Exhibit 2). The filings disclose the repurchases, but not their effect on reported losses. The SEC's loan-level field for the reason a loan was repurchased, which even has a code for early payment default, is blank at every issuer. And the repurchase-disclosure rule counts repurchase demands, not repurchases a sponsor makes on its own, so Santander reported "no activity" for 2024, a year in which it bought back $155 million.^[6]

Only as good as the promisor. The promise is the sponsor's own; a smaller lender's would be worth less. Santander's now carries a bank's credit. SDART 2026-1, in February, is its only SEC-registered deal this year. Its latest, the $1.3 billion SDART 2026-A in October, came from Santander Bank, N.A. (rated Baa1/A-/A by Moody's/S&P/Fitch) as a private placement, and according to Fitch the bank itself makes the same representation, word for word: the borrower "has made, or will make, the first two monthly payments."^[7] What changes is visibility. A private deal files no loan-level data, so the early defaults this article measures will no longer be visible for Santander's newest pools.

CCM read: Santander's reported losses understate how its loans perform. That isn't concealment: the repurchases are disclosed, the clause protects bondholders, and the loss falls on Santander rather than the trust. But set SDART or DRIVE against Exeter or AmeriCredit on reported net loss, and Santander's first year looks up to a third better than its loans performed.