Vintage Curve
Definition
A vintage curve is a line showing the cumulative losses, or cumulative delinquencies, of all loans originated in the same month, quarter or year, plotted against the number of months since origination. Vintage analysis, also called static pool analysis, lays the curves for successive cohorts on the same chart so that a loan made in 2024 can be compared with one made in 2022 at the same age, which a portfolio-level loss rate cannot do. Consumer and small-business loans lose money on a predictable schedule: almost nothing in the first three months, a rising slope through months 6 to 18 as weak borrowers fail, and a flattening as the survivors pay down, so the curve for a normal cohort is S-shaped and reaches most of its lifetime loss by month 24 for a three-year loan. A cohort whose curve sits above its predecessors at the same age was underwritten worse or into a worse economy, and the gap is visible within six months of origination, long before the cohort's losses reach the income statement. Rating agencies, warehouse lenders and CECL models all build on vintage curves, and a lender that is growing fast has a portfolio-level loss rate dominated by young, low-loss cohorts that can look excellent while every vintage is deteriorating.
Why it matters
Every lending platform that has failed in the last decade reported good portfolio-level loss rates until shortly before it did, and in most cases its vintage curves had been deteriorating for a year or more. The portfolio rate is an average across cohorts of different ages; growth fills the average with young loans; and the young loans' losses arrive after the growth has been financed. Vintage analysis is the corrective, and it is the analysis that warehouse lenders, rating agencies and sophisticated note buyers actually run. For an individual investing in marketplace notes, a fintech-originated ABS or a private credit fund that buys consumer paper, the question to ask is not what the loss rate is but what the last four vintages look like at month 6, month 12 and month 18 against the ones before them.
Common misconceptions
- •A low portfolio loss rate is not evidence of good credit. It can be evidence of fast growth; the two are indistinguishable without vintage curves.
- •Vintages are not comparable at different ages. A 2024 cohort at month 8 must be compared with the 2023 cohort at month 8, not with its lifetime figure.
- •Cumulative losses are not annualized losses. A curve that reaches 8 percent over three years is roughly 2.7 percent a year on original balance and higher on the declining outstanding balance.
- •Prepayments distort curves measured on outstanding balance. Cumulative loss on original balance is the standard because it is unaffected by how fast the survivors pay off.
Technical details
Worked example: three vintages of a three-year personal loan
Cumulative net loss as a percent of original balance, by months since origination.
2022 vintage: month 3 0.2%; month 6 1.4%; month 12 4.1%; month 18 6.0%; month 24 7.1%; month 36 (lifetime) 7.8%.
2023 vintage: month 3 0.3%; month 6 1.9%; month 12 5.3%; month 18 7.6%; month 24 8.9%. On the 2022 shape, lifetime about 9.8%.
2024 vintage: month 3 0.5%; month 6 2.8%; month 12 7.2%. On the 2022 shape, lifetime about 13.7%.
The lender's reported annual charge-off rate over this period rose only from 3.1% to 3.9%, because originations grew 60% a year and the portfolio was two-thirds loans under a year old. The vintage curves show underwriting that deteriorated by 25% a year for two years.
By month 6 of the 2024 vintage, twice the 2022 loss at the same age, the outcome was visible; the portfolio rate would not show it for another year.
How the curves are built
Cohort definition: loans originated in a month, quarter or year; by product, grade or channel where the mix has changed.
Numerator: cumulative gross charge-offs, or net of recoveries (state which); or cumulative 30-plus or 60-plus delinquency for earlier warning.
Denominator: original principal balance of the cohort; never the current balance.
Age axis: months on book from origination or from first payment date.
Extrapolation: young cohorts are projected to lifetime by applying the shape of mature cohorts (the ratio of month-36 loss to month-12 loss, for example), adjusted for macro conditions; this is the core of a CECL vintage-method allowance.
Reading a vintage chart
Stacking order: newer curves above older ones at the same age means deterioration; below means improvement or a tighter mix.
Slope: a steeper early slope means faster defaults, often from a riskier or fraud-affected cohort; a curve that keeps rising past month 24 on a three-year product means losses are coming from borrowers who paid for a long time first, a macro signal.
Level at maturity: the lifetime loss to compare with pricing; a product with a 14 percent coupon and a 10 percent lifetime loss on a two-year average life earns less than it appears.
Gaps and reversals: a cohort that starts high and then flattens below its predecessors usually reflects a change in charge-off policy or modifications, not recovery.
Volume: read the curves with each vintage's origination volume; a bad vintage that was also the largest matters more than the chart implies.
Where vintage data appears
ABS prospectuses: Regulation AB static pool data for the sponsor's prior pools, by vintage and month.
Bank and fintech disclosures: vintage disclosure of the allowance and of gross charge-offs by origination year is required in US GAAP financial statements under CECL for public companies.
Marketplace platforms: some publish vintage loss curves by grade; the ones that do not usually publish a blended return that the curves would explain.
Private credit funds: consumer and specialty finance funds report vintage curves to their own investors; asking for them is standard diligence.
Rating agency presale reports: expected loss ranges derived from the sponsor's vintage performance under base and stress cases.
Diligence questions
Ask for cumulative net loss and 30-plus delinquency curves by monthly or quarterly vintage for at least the last three years, on original balance.
Ask for the same curves by product, grade and channel; a stable blended curve can hide a worsening channel.
Ask how the lender extrapolates young vintages to lifetime, and compare its lifetime assumption with its mature cohorts.
Compare the implied lifetime loss of the newest vintages with the loss assumption in the securitization, the warehouse borrowing base and the allowance.
Check the origination volume of each vintage; the loss dollars are volume times curve.
Frequently asked questions
- What is a vintage curve?
- A chart of how much a group of loans made at the same time has lost, month by month, as it ages. The 2024 first-quarter vintage, say, is all loans originated in those three months; its curve shows cumulative charge-offs as a percentage of original balance at month 1, month 2 and so on. Plotting several vintages together shows whether newer loans are performing better or worse than older ones at the same age.
- What is vintage analysis used for?
- Comparing underwriting quality over time, forecasting lifetime losses for young cohorts by reference to older ones, detecting deterioration early, and separating the effect of portfolio growth from the effect of credit quality. Lenders use it to set reserves and pricing; investors in loan pools, ABS and lending platforms use it to check whether the loss rate they are shown reflects the loans they are buying.
- What does a normal vintage curve look like?
- An S. Losses are near zero for the first three to four months (a borrower has to miss several payments and the loan has to reach charge-off), rise steeply from about month 5 to month 18 as the borrowers who were going to fail do so, and flatten as the remaining balance shrinks and the remaining borrowers have demonstrated they pay. For a three-year unsecured loan, 60 to 70 percent of lifetime losses have usually occurred by month 18 and 85 to 90 percent by month 24.
- Why does a growing lender's loss rate look better than it is?
- Because the loss rate divides charge-offs by the current portfolio, and a fast-growing portfolio is mostly young loans that have not had time to default. A lender doubling its book every year can report a 3 percent annual loss rate while every vintage is on track to lose 8 percent over its life. Vintage curves remove the growth effect by measuring each cohort against its own original balance.
- What is static pool analysis?
- The same thing under the name used in securitization: the pool is fixed (static) at origination and its performance is tracked without new loans being added. Regulation AB requires ABS issuers to present static pool data for prior securitizations, by vintage, showing delinquencies, losses and prepayments by month, so investors can compare the new pool with its predecessors.
- How early can a bad vintage be identified?
- Within three to six months of origination, from early delinquency (30-plus days past due at months 2 to 4) rather than from losses, which arrive later. A cohort whose 30-plus rate at month 3 is 50 percent above the prior cohort's will, on a normal roll-rate chain, show cumulative losses 50 percent higher by month 12. Lenders that publish vintage delinquency curves as well as loss curves give investors that six-month lead.