The headline delinquency number isn't your risk. The risk is the forward roll-rate on everything else these borrowers owe — and your origination snapshot can't see it yet.

The number everyone is quoting is "nearly 25% of student loan borrowers are behind." It's a real figure, it's alarming, and if you use it the way most coverage uses it, you'll reach for the wrong conclusion. The interesting risk for a lender isn't in the student-loan book at all — most of you don't hold federal student loans. It's in what a student-loan delinquency does to the rest of a borrower's wallet, and in the fact that your origination scorecard is looking at a photograph when the story is in the motion.

First, get the number right

"Nearly 25%" comes from a Century Foundation / Protect Borrowers analysis cited by CNBC (Feb 20, 2026). It's an advocacy-group estimate, and — this matters — it's a share of borrowers with a payment due, not a balance-level delinquency rate. The federal balance-level figure is far lower: roughly 9.6% of student-loan balances were 90+ days past due or in default as of the NY Fed's Q4 2025 Household Debt and Credit report. The NY Fed's own borrower-level work lands in the same neighborhood as the headline — about 23.7% of borrowers required to pay were behind in early 2025 — so "one in four or five borrowers behind" is defensible. "25% delinquency" stated as a portfolio rate is not. Use the borrower framing, attribute it, and don't let a reader think a quarter of student-loan dollars are delinquent.

The timeline is the part that should be on your wall. Forbearance began March 2020. Payments resumed October 2023 under a 12-month "on-ramp" that suppressed negative reporting. That on-ramp ended September 30, 2024, and the Department of Education resumed reporting missed payments to the bureaus. Because federal loans report delinquent only at 90 days past due — not the 30-day mark most products use — the first wave of fresh 90+ DPD marks didn't hit credit files until around February 2025. There's a structural ~90-day lag baked into when this even becomes visible.

Where the contagion actually lives

Here's the mechanism that makes this a credit-risk story rather than a policy story. Payment history is roughly 35% of a FICO score, and a 90+ DPD is a severe derogatory. Counterintuitively, the borrowers who fell hardest were the clean ones: the NY Fed found starting scores of 760+ dropped about 171 points on a new student-loan delinquency, versus ~87 points for sub-620 borrowers who already had derogatories. Roughly 2.2 million borrowers lost 100+ points; FICO's own data put the average drop for newly-delinquent borrowers around 62 points and the national average score at 714–715.

Now the part your snapshot can't see. A borrower who is current on your card or auto loan today but whose score just fell 100 points is, as of this moment, mispriced relative to their refreshed risk. Your point-in-time bureau pull captures the lower score and the student-loan tradeline. What it does not capture is the forward deterioration on their other accounts that hasn't happened yet. The contagion is in the roll-rate, not the current state.

And it is rolling. TransUnion tracked seriously-delinquent student-loan borrowers from December 2024 to June 2025 and measured how their other products performed. The growth in serious delinquency:

• Credit cards: +479%
• Unsecured personal loans: +186%
• Auto: +67%
• Mortgage: +20%

That hierarchy is the whole insight. Borrowers under strain protect their collateral and let unsecured balances bleed first — TransUnion's survey found roughly a third say they're consciously prioritizing other bills over student loans. (This is one vendor's portfolio study; treat the exact multipliers as directional, not gospel. But the shape — unsecured first, secured last — is what you'd predict and what you should act on.)

What to actually do

Three concrete moves, none of which require waiting for the next national report:

1. Add a score-trajectory attribute, not just a score. A current-score snapshot hides direction. Compute the change in bureau score over a trailing 6–12 months on your existing book and flag the segment holding a student-loan tradeline with a recent delinquency. That flag is your early-warning population.
2. Run an adverse-selection check on new originations. Compare the score distribution of your approved population before and after Q1 2025. If you're seeing inflow of applicants who are current-but-recently-impaired, or whose scores partially recovered, your cutoffs are operating on a population your scorecard wasn't fit for. Watch population-stability-index drift on the student-loan-holding segment specifically.
3. Re-weight your roll-rate monitoring toward unsecured. The +479% card signal means your card and personal-loan books are where this shows up first. Secured products lag and will under-signal — don't take a calm auto book as evidence the contagion missed you.

The honest take

Don't oversell this. Three things cut against the alarmist version:

• It's concentrated, not universal. Mortgage delinquency among these borrowers rose only ~20%. If your book is secured and prime, your real exposure is smaller than the headline implies.
• The flow is already decelerating. The NY Fed's transition rate into serious delinquency actually fell from 16.2% (Q4 2025) to 10.9% (Q1 2026). This may be a one-time repricing of a stock of borrowers who got marked all at once when reporting resumed — not an accelerating cascade. (Note: that transition rate is a flow measure, not the same thing as the delinquency level; don't conflate them.)
• Much of the distress is already priced. TransUnion found 51% of subprime borrowers delinquent — but those are populations you already underwrite as high-risk. The genuinely "invisible" risk is narrower: the formerly-clean borrower whose score just dropped and whose other accounts haven't turned yet.

Takeaway

The student-loan number is a headline. The contagion is an operational signal, and it points at one specific population: borrowers who are current with you, hold a recently-delinquent student loan, and just took a score hit your origination snapshot already absorbed but your loss forecast hasn't. Build the score-trajectory flag, check your new-origination distribution for adverse selection, and watch unsecured roll rates in the overlap segment. If the cascade fizzles, you've lost nothing. If it doesn't, you saw it a quarter before the aggregate did.

‍