The two-month Extension Default Rate, and how to read it

An extension moves a payment. Whether it did any good is a different question, and most of the ways to answer it take a year or more. The two-month Extension Default Rate is the shortest answer I am comfortable putting a number on.

The definition is worth being fussy about, because almost every argument I have had about this metric turned out to be an argument about the definition. The Extension Default Rate is the share of extended loans that are 30 or more days past due (DPD) or charged off roughly two months after the extension, divided by observed extensions.

That is the whole calculation. No adjustment, no model, no attempt to say what the loan would have done otherwise. It counts a status and divides.

Why two months

Two months is not a compromise between rigor and impatience. It is close to the earliest point where the answer means anything.

A loan inside its deferral looks current, because that is what the deferral does. To register as 30+ DPD again, the moved payment has to come due and then a billing cycle has to pass. Measure at one month and you are mostly counting loans that have not yet had the chance to miss anything. Two months is the first window where a status change is informative rather than mechanical.

Longer horizons exist and we track them, but they answer a different question and they read higher. The two numbers are not interchangeable, so the horizon belongs in the label every time. A rate quoted without one is not a rate.

Prime and subprime, on one basis

Pooled over the latest 24 grant months, with data through April 2026, the prime pool sits near 17-18% and subprime near 38%. Prime runs roughly half. Not exactly half, and I would not hang anything on the ratio, but as a sense of the distance between the two books it is fair.

The "pooled over the latest 24 grant months" clause is not throat-clearing. It is the denominator, and in this metric the denominator does more work than the division.

Every extension in that pool went to a borrower who was somewhere specific when the servicer moved the payment. Some were current. Plenty were already behind. Those groups do not read anywhere close to the same. Restrict the pool to one starting bucket, or slide the window, and the headline moves while nothing underneath has changed.

So I want the basis attached whenever someone quotes this rate at me: which extensions are in the denominator, over what window. Without it I cannot line the number up against anything, including a second reading of the same pool.

Starting delinquency does most of the work

What moves this rate most is where the loan was standing when the extension was granted.

Extensions granted to loans that were already 30+ DPD show the highest rates in the data. Extensions granted to current loans show the lowest, and the buckets in between sort in order. The spread is wide enough that a pooled rate is mostly reporting the mix of starting statuses rather than anything about servicing.

This is also why pooled comparisons across issuers or across time go wrong so easily. A pooled rate can move several points without any single starting-status cohort moving at all, because the composition of who got extended changed underneath it. We walked through a live example of exactly that in the post on how high-LTV subprime extensions perform. Hold the starting bucket constant or the number will tell you something that is not there.

Loan-to-value moves it in both segments

The second gradient is origination loan-to-value (LTV), the loan amount relative to the vehicle's value at origination. The Extension Default Rate rises with LTV in both prime and subprime.

That direction is consistent with what we have already shown about who gets extended in the first place: subprime extension usage has been climbing fastest in the highest-LTV buckets. So the high-LTV slice is both the most extended and the slice where post-extension delinquency runs highest. Those are two separate observations from the same files, and they point the same direction.

I want to be careful about what that does and does not establish. It is a description of the observed record. Loans at high advance rates that receive extensions are more often 30+ DPD or charged off two months later. Why is a harder question than this metric can answer.

What the rate does not tell you

This is the section I would most like people to actually read.

It does not tell you eventual loss. A loan that is 30+ DPD two months after an extension has not resolved. Some cure, some do not, and the two-month reading does not sort them. Treating this number as a loss estimate is a misuse of it.

It also does not tell you what the extension did to the loan. Extended loans are not a random sample of the book, and nothing in the calculation corrects for that. Set the extended cohort next to the rest of the pool and you are measuring who gets extended at least as much as anything else.

It is bounded by what issuers report. The denominator is observed extensions, meaning extensions flagged in the SEC ABS-EE loan-level files. Modifications that are not flagged that way are not in the denominator, and reporting practice varies by issuer.

And it is descriptive, full stop. It says what the tape shows two months out. That is a genuinely useful thing to know early, and it is also all it is.

How I would use it

As a level, sparingly. As a comparison, constantly.

Fix the starting delinquency bucket, the segment and the horizon, then let one thing vary. Same bucket, two quarters apart. Or a portfolio against its own history in the same bucket, rather than against a pooled figure carrying everyone else's mix. Those comparisons hold up. A pooled level is fine as a market-wide reading so long as its basis travels with it, but as a read on one servicer it is the version most likely to mislead, and also the version most likely to reach you secondhand.

Then treat the reading as what it is: an early status check on a portfolio action, available a few months after the action, on data that already exists. That beats waiting a year, and it does not require pretending the number knows more than it does.

Methodology

The Extension Default Rate is the share of extended loans that are 30 or more days past due or charged off roughly two months after the extension, over observed extensions. The horizon offsets by the number of payments moved before adding two months, so the outcome is measured after the deferred payments resume rather than while the loan is still inside the deferral.

Extensions are loans flagged with a positive normalized payment_extended_number or an SEC modification_type_code of 4 or 04. Starting delinquency is the borrower's bucket in the month before the extension. Prime is FICO 660 and above at origination, subprime is below 660. LTV is origination loan-to-value.

The prime and subprime figures quoted here are pooled over the latest 24 grant months. Source is SEC ABS-EE loan-level filings parsed into LoanTape's cache, deduped to one record per loan per reporting month by latest filing, with data through April 2026. The COVID relief window is reported but not used as a baseline.

The loan-level view behind this is the ABS-EE dataset. Related reads: auto extensions as a credit signal and extension rates running above their 2018-19 average. The subscriber version, with issuer, trust, and post-extension outcome cuts, is in the Modification Effectiveness Deep Dive. For the loan-level files, see LoanTape pricing.