I keep coming back to a number that should not sit next to another number and look calm. Roughly one in twelve borrowers inside recent subprime auto deals was at least 60 days behind in July, the worst reading of that kind since 2018, and yet the bonds wrapped around those loans have mostly kept paying. That is not a magic trick. It is a structure. And once you see how the structure is built, the calm stops looking mysterious and starts looking expensive for the person sitting in the driver’s seat.
Picture a used truck financed at a rate north of 20 percent, then sliced into a security that pays investors something closer to 7 percent. The gap between those two figures is the whole story. Missed payments, fee income, a car that can still be sold, and a habit of pushing due dates further down the calendar all fit inside that gap. The household feels the strain. The bond, for a while, does not.
The Machine That Stays Current While the Driver Does Not
America’s subprime car market has turned into a clean lesson in how financial engineering can look healthy while the consumer underneath it gets steadily worse. Analysts who went through close to three million auto loans originated between 2021 and 2023, then packed into publicly traded asset-backed securities, found a system with enough interest, fees, and collateral protection that a borrower can fall behind, restructure, and eventually lose the vehicle without automatically cutting off cash to lenders and bondholders.
I have found that people hear “the bonds are fine” and translate it into “the loans must be fine.” They are not the same sentence. A loan is a promise from a household. A bond is a claim on a pool, a waterfall, and a set of rules about who gets paid first. Those rules were written to survive a certain amount of broken promises. The interesting question is not whether the rules work. It is how much pain they can absorb before the math changes.
A Spread Wide Enough to Hide a Lot of Trouble
Start with the rate. Subprime borrowers in these pools paid interest averaging about 18 percent. The securities built from those same loans were issued at rates reaching roughly 6.7 percent. That is not a rounding error. It is a cushion measured in double digits.
Why does the gap exist? Because the borrower is risky, the car is used, the credit file is thin or damaged, and the lender wants to be paid for all of that. Investors in the senior slices of the deal do not take the same risk. They sit higher in the stack. They accept a lower yield because other people’s money, and other people’s interest, stands in front of losses.
Think of it as a toll road with a very high toll. Most cars get through. Some break down. The toll is set so high that the operator can still cover fuel, staff, and a pile of breakdowns and have something left. In my experience reading these deals, the spread is the first thing sophisticated buyers look at, and the last thing struggling borrowers ever see explained in plain language.
- Borrower rates near 18 percent create excess interest every month the loan stays outstanding.
- Bond coupons near 6.7 percent leave a wide margin for losses, servicing, and reserves.
- Fees keep accruing whether the account is clean, late, or already on a workout plan.
- Recoveries from sold vehicles plug holes that interest alone cannot cover.
None of that requires the borrower to be comfortable. It requires the pool, on average, to throw off more cash than the bonds demand. Averages are kind to structures and cruel to individuals. One borrower can be ruined and the pool can still clear its tests.
Fees That Do Not Care If You Are Current
Servicing is the unglamorous middle of this market, and it is also where a lot of the durability lives. The company that collects payments, fields calls, books extensions, and arranges repossessions usually earns a fee tied to the balance or the activity. That fee arrives in good months and in bad ones.
There is nothing shady, by itself, about being paid to service a loan. Someone has to send the statements. The awkward part is the incentive. A servicer that gets paid for keeping an account alive has a reason to prefer a modification over a quick repo, at least for a while. A servicer judged on loss severity may prefer to take the car, sell it, and stop the bleeding. Both approaches can be rational. Neither one is designed around the borrower’s total cost.
They said they can push the loan back and you will be back current.
A borrower describing a loan extension
Current, in that sentence, is an accounting status. It is not a raise. It is not a cheaper car. It is a date moved. I keep thinking about how soothing that word sounds on a phone call and how little it changes the payment that will still be due.
Two Playbooks, One Stressed Household
Not every lender handles trouble the same way. One large originator in these pools modified nearly two thirds of the loans that ended up in its securitizations, often by sliding missed payments toward the end of the contract. Close to a quarter of those loans were modified at least four times. Another major player modified far fewer accounts and moved faster to repossess and sell the vehicle.
If you only watch bond performance, both styles can look competent. Extensions keep cash moving and delay the moment a loss is recognized. Faster repossession crystallizes the loss, recovers whatever the car will fetch, and stops a borrower from driving a depreciating asset while the balance barely moves. The bondholder may prefer one path or the other depending on where they sit in the capital stack. The driver usually prefers neither.
Perhaps the most interesting aspect is how often the softer path fails anyway. About a quarter of the modified loans examined still ended in repossession. Another 15 percent slipped back into delinquency. Among borrowers at the more aggressive modifier, almost one in three modified loans still finished with the vehicle taken back. An extension is a pause, not a cure.
| Servicing style | What it delays | What it does not fix |
| Repeated extensions | The date a loss is booked | The payment the household cannot afford |
| Faster repossession | Further balance growth | The borrower’s need for a car |
| High coupon cushion | Stress on senior bonds | The principal the borrower still owes |
| Collateral sale | Open-ended servicing cost | Depreciation already suffered |
A Truck, Five Modifications, and Almost No Principal
One Virginia borrower financed a Chevrolet Silverado for about $32,000 at 21.5 percent. After five modifications and more than $10,500 in payments, the truck was repossessed. The principal had fallen by less than $50.
Read that again, slowly. More than ten thousand dollars left the household. The balance barely noticed. That is what a high rate plus repeated deferrals can do. Interest and fees eat the payment. The extension resets the delinquency clock. The car keeps aging. When the repo finally happens, the borrower has financed a long, expensive lesson in how amortization works when the rate is hostile and the term keeps stretching.
I am not interested in scolding the person who needed the truck. Plenty of jobs in this country are impossible without a vehicle, and public transit is a theory in a lot of zip codes. The hard part is that the product sold as a solution becomes, after a few bad months, a machine for extracting payments without retiring debt. That is a design outcome, not a personal failing.
The Small Car That Outlasted the Budget
Another borrower took $12,000 at nearly 20 percent for a used Chevrolet Sonic, then fell behind. The lender modified the loan four times and pushed the schedule out by nine months. The car broke down. Repair money was borrowed. The account slipped again. She later described the pressure to accept extensions as something close to badgering. Being told you can be current again is a powerful sentence when the alternative is losing the car you use for work and for kids.
Her loan landed in a $1.2 billion securitization holding more than 53,000 auto loans. Four years and nearly $13,000 in payments later, she still owed $9,230 on a car that had been worth about $8,500 on the day she bought it. Payments exceeded the original price of the vehicle. The balance did not.
They got us between a rock and a hard place.
She also understood the trap with unusual clarity. Giving the car back earlier might have saved thousands. Keeping it was how she got to work and moved her children. That tradeoff does not show up in a bond prospectus. It shows up at 6 a.m. in a driveway.
How a Securitization Actually Gets Paid
A subprime auto ABS is a stack of claims on the same pool of loans. Monthly collections, recoveries, and, in some deals, reserve releases move through a waterfall. Senior notes are paid first. Mezzanine notes come next. Equity or residual interests take what is left, if anything is left. Credit enhancement, overcollateralization, excess spread, and sometimes a reserve account sit between borrower trouble and investor losses.
Excess spread is the star of the current moment. When loans yield 18 percent and bonds cost something near 7 percent, the difference, after servicing and other expenses, can absorb a surprising number of charge-offs before a rated tranche feels it. That is why a pool can look sick at the loan level and orderly at the bond level. The sickness is being prepaid by the borrowers who are still current, via a rate they would never have accepted if they had better options.
Rough monthly cushion, simplified: Borrower yield ~18% Bond coupon ~6.7% Gap before expenses ~11 points What fills the gap defaults, fees, servicing, residual
This is a sketch, not a deal model. Real waterfalls have triggers, target overcollateralization, and fees that move. The direction of the math is what matters. High borrower rates subsidize the structure. Extensions keep more loans inside the subsidy for longer. Repossessions convert metal into cash that can be used to pay notes. Put those together and you get a security that can hum while a large minority of its borrowers cannot.
Collateral That Ages While the Debt Gets Deferred
A car is a wasting asset. That sounds obvious until you set it next to a modification that adds months to the term. Every extension gives the vehicle more time to depreciate, pick up miles, and need repairs the borrower may not be able to fund. The debt, meanwhile, has not been forgiven. It has been relocated.
Used-car prices were unusually strong in the years right after the pandemic, which flattered recovery values and made a lot of 2021 and 2022 deals look safer than the credit files underneath them. Prices have cooled. A recovery rate that worked when wholesale values were inflated does not automatically work when they are not. Rating agencies have already lifted loss projections on some 2022 deals from one active subprime issuer to as high as 31 percent, citing delinquencies and extensions. That is a pool-level number, not a headline about a single missed payment.
Here is the part I think investors under-discuss. An extension can improve a delinquency ratio this month and worsen a loss severity two years from now. The car is older. The balance is higher than a normal amortization schedule would have allowed. The borrower who needed four modifications is not, on average, a stronger credit than the borrower who needed none. You have changed the timing of the problem. You have not changed the income that has to solve it.
Delinquency Is Already Telling on the Pool
The share of borrowers in securitized subprime auto deals who were at least 60 days delinquent reached 8 percent in July, the highest since 2018. That is the kind of figure that used to end arguments. It does not end this one, because the bonds have largely held together. Performance and deterioration are living in different rooms of the same house.
Serious delinquency is a lagging signal of household cash flow and a leading signal of losses, with a messy middle filled by modifications. If a servicer is quick to extend, the 60-day bucket can look better than the underlying stress. If a servicer is quick to charge off, the loss number shows up sooner and the delinquency number may look cleaner. Comparing issuers without comparing their modification policy is how smart people talk themselves into the wrong conclusion.
- Watch 60-day delinquency, but pair it with the modification rate.
- Watch extensions, but pair them with remaining term and vehicle age.
- Watch recovery rates, but pair them with wholesale price trends.
- Watch bond spreads, but remember senior tranches are built to look fine until they are not.
A single metric will flatter somebody. A set of metrics is harder to charm.
Why the Bonds Can Keep Performing Anyway
Four protections keep showing up when you pull these deals apart.
Excess interest is the big one. It turns a painful default rate into a manageable one for anyone not holding the first-loss piece. Servicing fees keep the operation funded so collections do not stop when borrowers struggle. Resale value on repossessed cars, even depreciated cars, returns cash to the trust. Extensions keep some payments flowing and postpone the recognition of losses, which can matter a great deal for triggers and for the optics of a deal still in its revolving or early amortization life.
Together, those features have let subprime auto ABS remain surprisingly durable. Durable is not the same as cheap, and it is not the same as safe at every attachment point. Junior notes and residuals are where this story gets less comfortable. They are paid to take the volatility that senior notes are structured to avoid. If loss projections in the low thirties show up in cash rather than in a model, someone in the stack feels it. Often that someone is not the buyer of the AAA slice.
I’ve found that retail commentary jumps straight from “defaults are rising” to “the whole market breaks.” Institutional commentary sometimes jumps the other way, from “senior bonds are covered” to “nothing is wrong.” Both jumps skip the capital structure. The stress is real. The transmission to each note is a design choice made years ago, when the loans were originated and the deal was sold.
The Household Budget Is Not a Waterfall
A trust can rank its claims. A family cannot. Rent, food, a car payment, a repair, childcare, and a medical bill do not arrive with a prospectus. When income is tight, something slips. Auto loans have a special status in that scramble because the collateral can be taken, and because losing the car can cost the job that was supposed to pay for the car. That circular pressure is why extensions get accepted even when the borrower half-knows the math is bad.
Loan modifications can change the accounting timeline. They cannot manufacture wages. Moving a missed payment to the end of the contract does not make the monthly number affordable, and it does not stop the odometer. If the original problem was a payment set too high relative to income, a longer term at a similar rate is a slower version of the same problem.
There is a temptation, in market writing, to treat the borrower as the raw material of the bond. I try not to. The raw material talks back. She needed the Sonic. He needed the Silverado. The structure needed their interest rate. Those needs are not aligned, and no amount of clean reporting will make them aligned.
What Extensions Quietly Cost
An extension feels like relief because the immediate threat recedes. The invoice for that relief is spread across later months, and it is easy to underestimate.
- Interest keeps accruing on a balance that should have been smaller.
- Fees can be capitalized or collected at the moment of the workout.
- The car ages into a lower recovery value if a repo still comes.
- The borrower may take on other debt, including repair loans, to keep the modified account alive.
- Credit reporting may look better for a spell, which can encourage more borrowing elsewhere.
None of those lines requires bad faith. A call-center script that offers to “make you current” can be accurate and still leave the person poorer. Accuracy about status is not the same as honesty about total cost. If I were rewriting those scripts, the second sentence would be the one about principal, not the one about the delinquency flag.
Investors Are Watching the Wrong Floor
The deterioration is already visible at the bottom of the structure, among the people making the payments. Bond spreads and rating actions tell you how far that pressure has traveled upward. They are not a substitute for looking at the loans.
A practical watchlist, if you cover this market or own a slice of it, is shorter than the marketing decks suggest. Modification frequency. Re-default after modification. Sixty-day delinquency with and without extensions. Cumulative loss versus original projection. Recovery lag, meaning how long a car sits before it is sold. Remaining term versus vehicle age. And the excess spread actually released, not the excess spread promised at issuance.
Senior bonds can pass every one of those tests for a long time and still be a poor relative value if the spread you are paid no longer compensates for the option that losses migrate up the stack. Junior bonds can fail those tests and still be interesting if the price already assumes a grim recovery. The error is treating the sector as a single trade. It is a set of attachment points with very different jobs.
A Short History of This Kind of Calm
This is not the first time a consumer credit product has looked stable in security form while looking ragged in household form. The pattern rhymes with other asset-backed markets: high coupons, structural credit enhancement, and a servicer with discretion. The rhyme is not a forecast. Auto loans have a advantage that unsecured consumer debt does not. There is a car. You can take it. That recovery right is a large part of why these bonds have a reputation for hanging on.
The disadvantage is just as simple. Cars depreciate, break, and are worth less in a soft wholesale market. A house can sit. A car cannot, not without cost. Recovery values are also local and seasonal. A truck in one region is not a sedan in another. Pool-level averages hide that, which is fine until a particular origination vintage was heavy on a particular kind of vehicle bought at a particular peak in price.
Vintages from 2021 through 2023 carry that risk in a specific way. Vehicles were expensive. Borrowers stretched. Terms were long. Rates for subprime buyers were already high before the broader rate cycle made everything else more expensive too. Inflation in insurance, repairs, and fuel then landed on the same budgets. A loan that was tight in month one became implausible in month eighteen, and the modification desk became the pressure valve.
Insurance, Repairs, and the Payment Nobody Modeled
Underwriting models are good at the contractual payment. They are worse at the payment that shows up when a water pump fails or a premium renews 30 percent higher. Subprime borrowers do not have a side account labeled “maintenance reserve.” When the car breaks, the choice is often a repair loan, a missed auto payment, or both.
That is how a borrower can be “helped” four times and still end upside down. The securitization sees an extension. The household sees a new bill stacked on an old one. I suspect this layer, more than any single macro print, explains why modified loans re-default so often. The original affordability problem was never only the auto note. It was the auto note plus the life required to keep the car moving.
What a Fairer Version of This Market Would Admit
I am not arguing that subprime auto lending should vanish. People with bruised credit buy cars, and someone will finance them. The question is whether the product is priced and serviced in a way that has a plausible path to a paid-off vehicle, or only a plausible path to a paid coupon.
A more honest market would show, at origination, the share of the first year’s payments likely to go to interest. It would cap the number of extensions before a different remedy is required. It would report modification rates next to delinquency rates so investors cannot celebrate one without the other. It would treat a re-default after a fourth extension as information, not as a surprise.
None of that is radical. It is basic. The fact that a borrower can pay more than the purchase price and still owe more than the car is worth should be a design flaw worth fixing, not a footnote in a surveillance report.
How Far the Pressure Can Travel
So far, the pressure has mostly stayed downstairs. Loss projections have been revised up on certain deals. Delinquency is at a multi-year high. Extensions are doing a lot of quiet work. Senior bonds, by and large, continue to function. That divergence is the reason this market is worth watching rather than dismissing.
The trip upstairs happens if several things arrive together. Excess spread compresses because more loans stop paying and fewer extensions succeed. Recoveries disappoint because used values slip or repos take longer. Triggers redirect cash away from junior holders and, in a harsher case, change the pace at which seniors are repaid. A servicer under strain cuts corners, and cure rates fall. None of those is guaranteed. All of them are visible in the data if you are willing to look below the bond price.
Could the machinery keep humming longer than the skeptics expect? Yes. High coupons are a powerful shock absorber, and cars can still be sold. Could the absorber be smaller than the holders of mezzanine and equity pieces assume? Also yes. Projected losses near 31 percent on some pools are not a rounding difference against an original base case. They are a different deal.
A Practical Read for Anyone Allocating Here
If you are allocating, or simply trying not to be surprised, separate three questions that get mashed together in headlines.
First, are borrowers weaker? The delinquency data says yes. Second, are senior bonds immediately impaired? The structure and the spread say usually not yet. Third, is the residual value of the equity in these deals as robust as it looked at issuance? The modification and loss-revision data say be careful.
That third question is where I would spend the time. Equity and thin credit enhancement live on the assumption that excess spread and recoveries will be there. Extensions borrow against that assumption. They can be the right servicing choice for a borrower who had a one-off shock. Used four times, they start to look like a way to keep a loan inside a pool that would rather not recognize it.
Simple check: payment applied to principal after a mod should not be near zero. If it is, the extension is renting time, not retiring debt.You do not need a terminal to run that check on the anecdotes already public. Ten thousand dollars in, less than fifty dollars of principal retired, then a repo. Thirteen thousand dollars in, balance still above the original value of the car. Those are not edge cases you can wave away as storytelling. They are what the rate and the term produce when income does not cooperate.
The Human Constraint the Model Keeps Missing
Markets are good at pricing a pool. They are bad at pricing a Tuesday. On Tuesday the car does not start, the shift starts at seven, and the extension offer is the only sentence that keeps the week intact. People take the offer. I would probably take the offer. The bond does not have to live that Tuesday, which is why it can remain performing while the borrower cannot.
There is a version of this article that ends with a forecast of imminent breakage. I do not have that forecast. Structures with this much excess interest can absorb more than a casual reader expects, and collateral recovers something, which unsecured credit does not. There is also a version that ends with a shrug because the AAA tranche is covered. That shrug ignores who is paying for the coverage, and how little principal their payments retire.
The more useful ending is narrower. The consumer is financially exhausted in a growing share of these pools. The security built on top of that consumer is still functioning because interest, fees, resale, and extensions were designed to allow exactly that split. The split can persist. It cannot persist for free. Every month of performance at the top is being purchased with time, depreciation, and household cash at the bottom.
Watch the modifications. Watch the re-defaults. Watch whether principal actually falls. When those three stop cooperating, the calm in the bonds will have fewer places left to hide, and the pressure that has lived with the drivers will finally have somewhere else to go.
