Explainers  /  3 of 20

Censoring

Why young vintages stop early on a chart and are left blank, not shown as zero.

A young vintage has not lived long enough to be read at every age. Its loans made last have not reached month 24 yet, so nobody knows how many of them will charge off by then. Censoring is how we handle that gap: those months are left out, and the vintage's line stops early. It is never filled with zeros.

Eight loans

One made-up vintage, eight car loans made in 2024, seen in September 2026. Month on book 1 is the month a loan is made.

Loan Made Amount Months on book now Charged off at month
Ana Jan 2024 $20,000 33 14
Bo Mar 2024 $25,000 31 23
Cy May 2024 $18,000 29 -
Dee Jul 2024 $22,000 27 -
Ed Sep 2024 $15,000 25 -
Flo Nov 2024 $20,000 23 -
Gus Dec 2024 $24,000 22 -
Hal Dec 2024 $16,000 22 -

To read the vintage at month 24, every loan needs 24 months of history. Flo, Gus and Hal are younger than that. Only $100,000 of the $160,000 lent, 62.5%, has reached month 24.

Counting them as zero

Count the young loans as if they will not charge off: $45,000 / $160,000 = 28.1% charged off by month 24.

Two months later, in November 2026, all eight have reached month 24, and Gus charged off at month 23. The real number is $69,000 / $160,000 = 43.1%. Counting the young loans as zero was 15.0 points too low.

Leaving them out: the 99% rule

We show a month only when at least 99% of the vintage's money could have reached it. In September 2026, every loan has reached month 22, so the line runs to month 22 (12.5%) and stops. Month 23 holds 75.0% and month 24 62.5%: both stay blank until the young loans get there.

The real numbers

Subprime car loans (FICO 300-659 at origination) by the quarter they were made, through September 2026: the share of each quarter's balance in loans that have charged off. Each young quarter stops at the last month its youngest loans have reached.

Quarter Last month shown Share then
2023 Q3 36 21.6%
2023 Q4 34 19.9%
2024 Q1 31 19.2%
2024 Q2 28 17.1%
2024 Q3 25 14.4%
2024 Q4 22 11.6%
2025 Q1 19 10.3%
2025 Q2 16 7.1%
2025 Q3 13 5.3%

2024 Q4 was made from October to December 2024. Its December loans are 22 months old, so the line stops at month 22: 11.6%. Month 23 holds 70.8% of its balance and month 24 35.8%, short of 99%; month 24 arrives with the November 2026 tapes.

Counting the younger loans as zero from here on, with no more charge-offs, would draw 2024 Q4 at 12.5% by month 24, up only 0.9 points from month 22, as if charge-offs had nearly stopped. Over the same two months, 2024 Q3 rose 1.6 points, from 12.1% to 13.6%.

Common pitfalls

  • A line that stops is not a line that flattens. Blank means not known yet. A chart that fills the gap with zeros, or with the loans' last known state, makes young vintages look better than they are.
  • Ranking a young vintage on a month it has not reached. Compare vintages only at months every one of them has reached.
  • Why 99% and not 100%. The 1% leeway keeps a sliver of very late loans from holding back a whole vintage.
  • Loans that join or leave the data. A loan first reported after month k, or that stops reporting before month k without a known outcome, is left out of month k's numerator and denominator too. It is never counted as a loan that paid.

Where to see it

Source: LoanTape, SEC Form ABS-EE loan-level data, through September 2026. Quarterly vintages, FICO 300-659 at origination, all issuers.