Issuer Charge-Off Amounts Are Not Comparable in Auto ABS
Correction, September 20, 2026: I withdraw this post's $2.49 billion / 6.3% "missing loss" claim and its aggregate before-and-after loss table. The proposed exposure calculation included paid-off loans carrying small write-offs and incorrectly counted collected principal as gross loss. I also generalized Harley-Davidson's recovery reporting to other net-of-proceeds filers; the follow-up investigation does not support that generalization. The payoff correction is implemented and tested locally. It has not been deployed, and this post provides no replacement June or July aggregate totals.
The filed charge-off amount can represent a full balance, a loss after collateral proceeds, a partial write-down, or a small write-off at payoff. Comparing it with the loan's prior balance helps identify those differences, but a low ratio alone does not establish a terminal charge-off.
The September 20 review found two separate problems: the exposure calculation was overstating payoff losses, and the recovery assumptions did not hold across issuers. The first has a shared calculation fix. The second requires evidence about where each issuer reports the cash.
A recovery rate of 101%
The audit started while building the motorcycle ABS post. The first pass at recovery rates divided cash recoveries by the filed charged-off amount. On Harley-Davidson's completed 2019-2021 charge-off cohorts it returned 101%, prompting a review of the denominator.
I ruled out the obvious explanation first. If recoveries were filed as running totals and re-filed each month, summing them would double count. They are not: the typical charged-off loan shows one or two recovery months, and the ninetieth percentile is three. That is a collateral sale, not a cumulative field.
The mechanism appeared when I split the 936 charged-off loans in those cohorts by the size of their recoveries relative to the filed amount, then reconciled each group against the balance the loan carried entering its charge-off month.
| Recoveries vs filed charge-off amount | Share of loans | Filed amount / entering balance (median) | Filed amount + recoveries / entering balance (median) |
|---|---|---|---|
| At or below 1x | 54.6% | 1.00 | 1.45 |
| 1x to 2x | 19.3% | 0.69 | 1.49 |
| Above 2x | 26.1% | 0.24 | 1.05 |
The top group files the full balance, and recoveries arrive on top of it. The bottom group files roughly a quarter of the balance, and adding recoveries back reconstructs the balance within about 5%. That pattern supports a charge-off filed net of proceeds, with those proceeds also reported as recoveries. Subtracting them again from the netted charge-off understates net loss; dividing them by the netted amount inflates the recovery ratio.
The completed Harley-Davidson cohorts' recovery rates measured against entering balances were 62-66%, as reported in the motorcycle analysis. Those figures describe that sample. They do not establish how another issuer reports proceeds.
What the follow-up investigation found
I examined the event rows, payoff codes, paid-in-full dates, principal cash and recovery fields in the saved September 20 investigation. The small filed amounts had several different explanations.
| Observed pattern | What the loan-level evidence shows | Consequence for the calculation |
|---|---|---|
| Full-balance charge-off | Filed principal broadly reconciles to entering balance | Filed recoveries can be compared with that gross exposure |
| Netted amount with proceeds also in recoveries | The pattern documented in the Harley-Davidson sample; also observed in Nissan's netted bucket | Entering exposure and filed recoveries reconcile differently from the raw charge-off field |
| Netted amount with proceeds in principal cash | Some charge-off rows report the balance difference as principal collected | Gross exposure minus filed recoveries alone can overstate net loss |
| Netted amount with proceeds absent | Ally and BMW's reviewed rows have charge-off codes but do not report the proceeds needed for reconciliation | A gross exposure floor does not supply the missing net-loss information |
| Small write-off at payoff | Payoff code or paid-in-full date, often with collected principal explaining the balance drop | Keep the filed write-off; collected principal is not gross loss |
| Open partial write-down | A positive filed amount while the loan continues reporting a balance | Recognize the filed loss without flooring it at the entire entering balance |
Mercedes-Benz, GM and Ford supplied clear examples of the payoff problem: loans with a small filed write-off, a payoff marker and principal collected that explains the balance leaving the pool. The earlier guard treated a positive write-off plus an exhausted balance as sufficient evidence of a terminal charge-off. It then raised the small write-off to the full entering balance.
That error invalidates the original aggregate "missing loss" interpretation. The saved follow-up impact analysis uses seasoned loans with first charge-off dates through April 30, 2025; it is not a replacement for the original June release audit or a newly rebuilt July publication.
The recovery issue remains separate. CapitalOne reports too little recovery information to support a comparable economic recovery rate or loss-given-default measure in the reviewed population. Ally and BMW's missing proceeds also prevent a complete reconciliation. Absence of a filed recovery cannot establish that no collateral cash was received.
Why the pool-level reports cannot show it
Pool totals in a monthly Form 10-D or a servicer remittance report do not provide the row-level reconciliation used here. A low charge-off amount can have several explanations, and it does not imply that a corresponding amount appears in recoveries.
The ABS-EE data lets me compare each loan's observed balance history with its charge-off amount, terminal code, payoff date, principal cash and recovery rows. Those fields distinguish a payoff write-off from a terminal loss and reveal whether proceeds are separately reported. They cannot reconstruct proceeds that are absent from the tape.
The local payoff correction
The corrected shared calculation preserves the filed Silver values and changes when the exposure floor applies:
- Find the first charge-off event: a positive filed charge-off amount, or zero-balance code 4.
- Take the balance entering that month, from the prior month's ending balance. Where no prior month exists, use current balance plus the first filed amount.
- Keep positive filed write-offs at filed value until a terminal charge-off is observed. Code 4 remains terminal. A positive write-off with an observed exhausted balance is terminal only when neither a payoff marker nor sufficient principal cash explains it. A missing balance is unknown.
- When the terminal guard applies, the governed cumulative gross loss becomes the greater of cumulative positive filed charge-offs and the first-event entering exposure. Each month records the non-negative increase in that cumulative amount.
For example, a $100 loan paid off with $99.90 of principal cash and a $0.10 write-off records $0.10 of gross loss. An explicit code-4 charge-off still receives the exposure floor. Positive filed write-offs continue to count as loss events under the existing frequency definition; this correction changes their monetary treatment.
The calculation retains its cutoff behavior: later terminal evidence contributes from that observation forward. Adding a later month does not itself move the loss into an earlier month.
The current net-loss formula remains gross loss minus filed recovered_amount. Applying an additional offset for proceeds reported as principal cash, and excluding issuers with insufficient recovery reporting from comparable net-loss metrics, remain separate unfinished work. Correcting the payoff floor alone does not resolve those questions.
Release status
On August 22 I put a correction notice on the four public posts whose figures were built from the filed field: the subprime loss-to-liquidation and prime loss-severity posts from July 26, the recovery-by-make post from August 4, and the loss-to-liquidation-by-FICO post from August 7. The notice reads the same on each: the loss values used issuer-filed charge-off principal, which is not economically comparable across filers, and every figure and chart is being recomputed on the first-charge-off exposure basis. The posts stay readable. Their loss charts are replaced with a placeholder until the corrected versions are rebuilt.
The premium decks were withdrawn rather than annotated, since a deck cannot carry a per-chart caveat. That covers the five State of Auto monthly decks from January through May 2026, the vehicle collateral loan-to-value (LTV) deep dive, and the payment extensions deep dive. Their URLs return 410 Gone. Five gallery entries with monetary loss content, the public performance snapshot CSV, one product-post screenshot, and two screenshots in the June newsletter were also withdrawn. Two charge-off research drafts that were queued for publication are held.
Delinquency and extension figures on the public trackers are unaffected and remain published; only the monetary loss observation is withheld. Delinquency, measured as days past due (DPD) against active balance, does not use the charge-off field.
As of this correction, the payoff fix and its downstream query wiring have been tested locally. The prepared code advances the charge-off basis to v3, invalidates prior certification and caches, and requires a full Gold rebuild followed by fresh validation and regenerated published artifacts. That rebuild and publication have not been performed for this correction.
The auto dashboard and withdrawn analytical assets cannot be described as corrected merely because the local helper changed. Release review still requires measured issuer movements on the corrected basis. The earlier expected-movement totals are historical evidence, not approval for the new calculation.
Reading issuer loss numbers in the meantime
The investigation changes how I interpret the raw fields.
A recovery ratio near or above 100% warrants a denominator review. Harley-Davidson provides a documented example of proceeds already netted from the filed charge-off amount, but that finding does not apply automatically to another issuer.
A small write-off paired with a zero ending balance warrants a payoff check. Payoff codes, paid-in-full dates and principal cash can explain the balance reduction without turning the collected balance into a loss.
Finally, comparable net-loss and recovery measures depend on where the proceeds are filed. Correct gross exposure is necessary, but insufficient when cash appears in a different field or is not reported.
The reconciliation that produced this, every charged-off loan's balance history against its filed amount and recoveries, is the same loan-level extraction we build for enterprise clients on any issuer or segment.
Methodology note
Source is loan-level ABS-EE data filed with the SEC, parsed monthly and restricted to retail loan trusts; leases stay on a separate basis. The Harley-Davidson table preserves the August 22 analysis of 936 charged-off loans in completed 2019-2021 charge-off cohorts. The September 20 follow-up used saved loan-level probes of seasoned loans with first charge-off dates through April 30, 2025 and positive pre-charge-off balances, drawn from the July 2026 warehouse state. Those populations differ from the withdrawn June cross-issuer comparison. No replacement all-issuer aggregate is asserted here. The local correction preserves filed recovery amounts and existing loss-event definitions while changing the payoff treatment in monetary gross loss.