Prime and subprime extensions differ in level, not shape
Nearly everything we have published on payment extensions has been about subprime. That is where the volume is, and it is where the usage rate sits well above its 2018-19 average. The easy assumption is that prime extensions are a different instrument entirely, a rare administrative courtesy for someone who changed banks and missed a statement.
Half of that assumption holds up. Prime extension usage is far lower than subprime, and the loans are in better shape when the extension lands. 77% of prime extensions go to loans under 30 days past due. Of that 77, 61 points are fully current and 16 points are already 11 to 29 days late. Subprime runs the other way, with the mass of its extensions landing on loans further behind than that. So who gets extended is genuinely different between the two books.
What is not different is how the outcomes sort afterward.
The level gap is roughly two to one
We measure post-extension outcomes with the Extension Default Rate: the share of extended loans that are 30 or more days past due (DPD) or charged off about two months later, after the deferred payments resume. Measuring after the deferral matters, because a loan sitting inside a deferral reports current whether or not anything improved.
Prime runs 17 to 18%. Subprime runs about 38%. Call it half.
That ratio is the headline, but the absolute prime number is the part I keep going back to. One in six prime extensions is 30+ DPD or charged off two months on, and 61% of them started from a fully current loan. Whatever a prime extension is, it is not paperwork.
| Measure | Prime | Subprime |
|---|---|---|
| Extension usage | far lower | higher |
| Delinquency when the extension is granted | 61% current, 16% at 11-29 DPD, 23% at 30+ | minority current, mass already delinquent |
| Extension Default Rate (30+ DPD or charged off, ~2 months out) | 17-18% | ~38% |
| Extension Default Rate vs origination loan-to-value | rises with LTV | rises with LTV |
| Extension Default Rate vs delinquency at the extension | rises with delinquency | rises with delinquency |
Read the bottom two rows against the top three. The levels and the mix differ. The directions do not.
Prime extensions sort by term and by vehicle age
Take usage first. Split the prime book by original term and the extension rate climbs at every step.
| Original term | Prime monthly extension rate |
|---|---|
| 60 months or less | 0.08% |
| 61-72 months | 0.26% |
| 73-75 months | 0.47% |
| 76 months and up | 0.76% |
Bottom to top, that is 9.7 times. No crossovers, no flat stretch in the middle. A prime borrower on a 76-month contract shows up in the extension data almost ten times as often as a prime borrower on a 60-month or shorter one.
Vehicle age does the same thing with a shallower slope. Extension rates run 0.28% on vehicles zero to one year old and reach 0.57% at both the 8-9 and 10+ year buckets, about 2.1 times. That is a gentler curve than term, and it lines up with the older-collateral delinquency pattern we have shown on the same files.
Both are the same species of curve we found in subprime, where extension usage climbs with origination loan-to-value. High advance rates, long amortization, older collateral. The loan characteristics that carry heavier extension use in the segment everyone watches carry it in the segment nobody does.
The outcome gradients line up too
Usage is the easier half. The harder question is whether extensions behave the same way once granted, and here the two books track each other closely.
In both segments, the Extension Default Rate rises with origination LTV, and in both it rises with the borrower's delinquency bucket at the time of the extension. We showed the subprime version of that grid in high-LTV subprime extensions cure the least, where the highest-LTV group posts the worst outcome in every starting delinquency bucket. Prime produces the same ordering. Lower, but the same.
That is what I mean by shape versus level. Two grids, same sort order in both dimensions, one of them scaled down.
About that 77%
The 77% is two populations stacked in one bar. 61% of prime extensions go to loans that are fully current, zero to ten days. Another 16% go to loans already 11 to 29 days late. The rest, 23%, are 30 days or more past due when the servicer moves the payment.
That split decides whether anyone can see this activity at all. A 30+ DPD series will never register the 61%. Those loans report current the month before the extension and current the month after, with a deferred payment in between. The 16% at 11 to 29 days do report late, so they are visible, but only to someone who breaks out the early bucket, and most public delinquency reporting starts at 30 and never shows it.
So the honest version of the invisibility point is narrower than "prime extensions never show up in delinquency data," and more useful. Most of the sub-30 population is invisible on a standard tape. The 16-point remainder is not invisible at all, it is sitting in a bucket that most reporting has already folded into current before anyone looks at it. Whether you see prime extension activity is mostly a question of which bucket you kept.
One measure runs against that framing. The mix does differ. Prime extensions are drawn from a healthier starting pool and subprime extensions are not, and current loans post lower Extension Default Rates in both books. Some unknown share of prime's 17 to 18% is composition rather than credit.
These cuts cannot separate the two. Whether prime's lower Extension Default Rate is the mix, or something about prime borrowers and prime collateral on top of the mix, is not answerable from a usage table and an outcome table. Anyone who tells you which one it is from this data is telling you more than the filings support. What the data does say is narrower and still worth having: the gradients hold in both books, in the same direction, on the same cuts.
One model, scaled
If you hold paper across both books, you don't need two frameworks for this. The cuts that sort subprime extension risk are the cuts that sort prime extension risk, and the ordering does not flip anywhere we have looked. Take the subprime mental model, halve it on the outcome side, cut it much further on the usage side, and you have a workable read on the prime book.
The error I would guard against runs the other way. A 0.08% extension rate on short-term prime contracts looks like a rounding error, and it is tempting to conclude the gradient does not exist up the credit box. It exists, on a very small base. Seventeen to 18% only looks small when you set it next to 38%.
The other thing this changes is what counts as a surprise. If prime extension usage rises inside the 76-month or older-vehicle buckets, that is the part of the prime book already sitting at the top of its own curve. It does not need a new explanation.
What I would watch
The 76-month and up bucket, since it is the top of the prime usage curve and the place where a shift shows up first at the loan level.
The Extension Default Rate on the 23% granted at 30+ DPD. The pooled 17 to 18% blends them with the 61% that were current and the 16% that were barely late, and that 30+ slice is where the prime and subprime numbers should land closest. The 11 to 29 day group is worth pulling out on its own too, since a standard delinquency screen misses it by a single bucket.
Whether the vehicle-age gradient holds its 2.1x spread as the older buckets fill in with 2021-and-later collateral.
Methodology
Source is SEC ABS-EE loan-level filings for public auto ABS, parsed into LoanTape's loan-level cache and deduped to one record per loan per reporting month by latest filing. Data runs through April 2026.
Prime is FICO 660 and above at origination, subprime is below 660. An extension month is a loan with a positive normalized payment_extended_number or an SEC modification_type_code of 4 or 04. The monthly extension rate is extension events over eligible loans, where eligible means active, non-terminal reported loans with balance above $1 and the required prior-delinquency and FICO fields present.
The Extension Default Rate is the share of observed extensions that are 30+ DPD or charged off about two months after the inferred extension ends. The horizon offsets by the number of payments moved before adding the two months, so the outcome is measured after the deferred payments resume rather than while the loan is still inside the deferral. Starting delinquency is the borrower's bucket the month before the extension: current is zero to ten days, then 11-29, then 30 and above. LTV is origination loan-to-value. Original term and vehicle age are as reported at origination.
Cohorts are pooled over recent reporting months. The COVID relief window is not used as a baseline.
This runs on the same loan-level infrastructure behind LoanTape's ABS-EE dataset, Form 10-D data, and ABS remittance data. It sits alongside the earlier work on extensions as a credit signal. The subscriber version, with issuer, trust, and post-extension outcome cuts, is in the Modification Effectiveness Deep Dive. For the loan-level files behind it, see LoanTape pricing.