Vintage curves explained: why your newest loans look like your best
In brief
Your newest loans look like your best, mostly because they are too young to be 90 days late.A vintage curve follows the loans lent in one quarter as they age, so every quarter can be compared at the same months on book. A quarter running above the older ones shows up while the book's delinquency rate can still look fine.
Every quarter at the same age
Each line is one quarter's loans, read by months on book, not by calendar date. Each line stops at that quarter's age today.
- Q3, the quarter the episode reads
- Older and younger quarters
- Range of Q1 and Q2: history, not a target
The figures behind the chart
| Month | Q1 | Q2 | Q3 | Q4 |
|---|---|---|---|---|
| 0 | 0% | 0% | 0% | 0% |
| 1 | 0% | 0% | 0% | 0% |
| 2 | 0% | 0% | 0% | 0% |
| 3 | 0% | 0% | 0% | 0% |
| 4 | 0.55% | 0.45% | 0.75% | not yet |
| 5 | 1% | 0.85% | 1.4% | not yet |
| 6 | 1.4% | 1.2% | 2% | not yet |
| 7 | 1.75% | 1.55% | not yet | not yet |
| 8 | 2.05% | 1.85% | not yet | not yet |
| 9 | 2.3% | 2.1% | not yet | not yet |
| 10 | 2.5% | not yet | not yet | not yet |
| 11 | 2.65% | not yet | not yet | not yet |
| 12 | 2.75% | not yet | not yet | not yet |
Months to reach the 2% marker
The same marker, read as speed: the month on book at which each quarter's curve first reached it.
- Q1Month 8
- Q2Month 9
- Q3 · the quarter in questionMonth 6
Q3 reached 2% at month 6, sooner than any older quarter. The 2% line is a marker for comparing speed, not a limit. A recent quarter above the older ones is a question for your risk and collections teams, not a verdict. Illustrative numbers.
The episode in four answers
What is a vintage curve?
A vintage curve follows every loan lent in one quarter as it ages and plots a delinquency measure against months on book, so each quarter can be compared at the same age. In this episode the measure is the share of the money lent that quarter that went to loans ever 90 or more days past due. A loan that gets there stays counted even if it pays later, and it counts the full amount lent: money on loans that fell behind, not money lost. So the curve is flat for the first few months, then climbs, levels off and never comes back down.
Why do new loans look better than older ones?
With monthly payments, a loan can't be 90 days late until it is about four months old, so a young quarter looks clean whether it is good or bad. A delinquency rate mixes loans of every age into one number, and lending more improves it on age alone. In the example from episode one, 4,000 new loans, all current, took the rate from 5% to 3.9%, while the original 10,000 loans went from 5% to 5.4%.
How do you compare quarters on a vintage chart?
At the same months on book, not by calendar date. In the episode, Q3 is at 2% at six months, where Q1 was at 1.4% and Q2 at 1.2%. Q3 reached 2% at month six; Q1 took eight months and Q2 nine. The 2% is a marker for comparing speed, not a limit, and the range of the older quarters is history, not a target.
Does a quarter above the older ones mean it is weak?
Not on its own. It is a question, not a verdict. With only two older quarters, a recent one will often sit above both by chance, and small quarters or a few large loans can open a gap. Three clocks move a curve: the loan's age, the quarter it was lent in, and the calendar. Comparing at the same age takes out the first. The chart can't fully separate the other two, but it can test the calendar: in the months Q3 climbed, Q2 rose just as Q1 had at the same age, so the calendar is the less likely cause, though not ruled out.
Full transcript
The video's English captions, as spoken. Select a time to play from that point.
0:00The quarter that looks best
Your newest loans look like your best ones. Hardly any are late. That's mostly age, not proof of quality. With monthly payments, a loan can't be 90 days late until it's about four months old. So a young quarter looks clean, good or bad, and you keep lending the same way. In our example, one quarter is just six months old, and loans that have been 90 days late already account for 2% of the money lent. Is that a warning, or normal for its age? In this video, we'll explain vintage curves, and how to tell whether that 2% is a warning. Here's the plan. One, why age hides risk. Two, building a vintage. Three, reading quarters side by side. Four, the early warning. Five, what moves a curve. And six, what to ask on Monday.
0:52Why age hides risk
A delinquency rate mixes loans of every age into one number. Lend more, and the rate improves on age alone. In episode one, 4,000 new loans joined, all current, and none left the book. The rate fell from 5% to 3.9%. But the original 10,000 went from 5% to 5.4%. The old loans got worse.
1:15Building a vintage
A vintage fixes this. Take every loan lent in one quarter, and follow them together as they age. The clock is months on book: zero at disbursal. This time we count money, not loans. Each month, ask one question: what share of the money lent that quarter went to loans that have ever been 90 or more days past due? Once a loan gets there, it stays counted, even if it pays later. And it counts the full amount lent: money on loans that fell behind, not money lost. Plot it, and you get a curve. Flat at zero for the first few months. Then it climbs, and levels off. It never comes back down. Collections that stop a loan reaching 90 days slow the climb. Money recovered after that won't show here. At the same age, a lower curve is better.
2:06Reading quarters side by side
Now put four quarters on one chart. The bottom axis isn't the calendar. It's months on book, so every quarter is compared at the same age. Q1 has twelve months of history. Q2 has nine. Quarters nine months or older form a range. That range is history, not a target. Q4 has three months, and its curve is still flat at zero. That doesn't make it good. It's too young for this curve. Its first missed payments show sooner, in your early buckets. That leaves Q3.
2:37The early warning
Q3 is the quarter from the opening. At six months, it's at 2%. At the same age, Q1 was at 1.4%, and Q2 at 1.2%. Q3 was above them at months four and five too, and the gap grew every month. Q3 reached 2% at month six. Q1 took eight months. Q2 took nine. 2% is a marker for comparing speed, not a limit. Meanwhile, the book's delinquency rate can still look fine. If Q3 is weak, and Q4 went out on the same terms, you're already two quarters deep, and Q4 still looks clean.
3:16What moves a curve
Treat that as a question, not a verdict. With only two older quarters, a recent one will often sit above both by chance. Small quarters wobble, and so can a few large loans. Ask what sits behind the gap. Three clocks move a curve. The loan's age. The quarter it was lent in: its product, its channel. And the calendar: the months every open loan lives through together. Comparing at the same age takes out the first. The chart can't fully separate the other two, but it can test one. In the months Q3 climbed, Q2 rose just as Q1 had at the same age. So the calendar is the less likely cause here, though it isn't ruled out.
3:57What to ask on Monday
Two things for Monday. Ask your risk team which recent quarter runs above the older ones at the same age, and what was different about it. And ask collections how much of that quarter is one or two months behind today. Those loans can still reach 90 days. Your team can reach them first. We built Lokta's Vitals screen to draw these curves from each loan's installments: when each one fell due, and when it was cleared in full. It shades the range of the older quarters, and flags a recent quarter above it for review. The flag tells you where to look. Your team decides what it means.
4:32What's next
Next on Lokta Lendisms: promise-to-pay fulfillment.
Sources cited in the video
- OCC Comptroller's Handbook, Credit Card Lending, Version 2.0 (April 2021): roll-rate, historical and vintage analysis, Appendix G pages 155-158, and vintage analysis in the glossary, page 175
- Breeden and Thomas, Specification Errors in Stress Testing Models (2009): loan age, origination vintage and calendar time