Auto Subprime Loss Index

Use cases

Hedging

An auto lender or ABS buyer holding subprime exposure is long credit — they own the loans and are paid to carry the risk. That makes them short losses: rising charge-offs erode the book. Going long the Net-Loss Future pays out as losses rise, offsetting those markdowns — a direct hedge on the loss leg. It's capital-light (post margin, not the full notional) and marks to market monthly, so the hedge tracks the book between prints.

Speculating

Take a directional view on the loss rate. Go long if you expect losses to climb faster than the market does; short to get synthetically long credit — if you think losses will shrink, or rise by less than the market expects, shorting the future profits as the realized print settles below your entry. Because each contract settles to a dated print, the strip prices the market's forward loss curve, and your edge is the gap between your view and that curve.

Composition

Loan-level SEC ABS-EE filings from subprime auto securitizations: deals whose balance-weighted issuance FICO is below 640 and whose original pool topped $200M.

Constituents are drawn from staggered 2017–2024+ vintages across the major subprime shelves (Santander Drive, Exeter, AmeriCredit, Bridgecrest, Carvana), so calendar coverage stays continuous as deals season and roll off — the live pool runs up to ~12 deals at a time, with history back to December 2016.

Seasoning filter

Subprime losses follow a hump as a deal ages, which would otherwise dominate the signal. We address it two ways.

First, the index pools loans across many deals of staggered vintages and balance-weights them, so the blended pool age stays roughly flat and the level reflects credit, not the loss curve of any one deal.

Second, a deal exits the pool once its balance falls below 10% of issuance, dropping the surviving-bad-loan tail of a nearly paid-down deal.

How the mark is calculated

Each month we measure the balance newly entering 90+ DPD across the pooled loans and scale it by the empirical ~0.81 roll-to-charge-off — from ~33M loan-months, ~81% of loans reaching 90+ DPD eventually charge off (~94% by 120+).

That recognizes loss ~a year before it's booked (a loan sits at 120+ DPD ~13 months before the charge-off posts), so the index leads realized losses. Recoveries are excluded — they arrive with a long, uncertain lag.

Annualizing the monthly flow gives the print: 100 × 12 × 0.81 × (balance entering 90+ DPD) / (beginning pool balance). A dated future entered earlier then settles in arrears to that print, paying 1% of notional per point.

Contract terms

The Net-Loss Future is a dated, cash-settled monthly future on the index. Each contract settles in arrears to the realized print of its reference month, once the lagged ABS-EE data posts (~the 22nd of the following month) — the settlement date is fixed to that print date.

Payoff is linear in the rate: PnL = side × notional × (rate_settle − rate_entry) / 100, so one point (1% on the rate) = 1% of notional (DV01-style).

Collateralized in USDC (testnet MockUSDC): 20% initial margin, 10% maintenance; below maintenance a holder tops up or is liquidated (0.5% of notional to the liquidator). Listed as a monthly strip, the series form a forward loss curve. Unaudited testnet proof-of-concept — not for real funds.