Somebody guessed what your car would be worth

I did the math.

A lease was ending and there was a decision to make. One number is printed in the contract, the price you can buy the car for at the end. The other number is what the car is worth that week. If the second one is bigger, you buy it. If it is smaller, you hand back the keys and walk away.

The number in the contract is called the residual. Somebody guessed it three years earlier.

A few million Americans make that same decision every year. The guessing behind it is one of the least watched businesses in credit.

Right now the guessers are winning. Ford Credit booked residual value gains of 14.2% this year. GM Financial booked 7.1%. Amy Sze at JPMorgan put those figures in a research note on August 28.

Cars are coming back worth more than the contract said they would be.

So the lenders write next year's contracts off that. Black Book expects three-year residuals on this year's models to be set in the mid-50s as a share of sticker price, about six to seven points above where residuals sat before 2020.

The wholesale market backs them up. The Manheim index read 215.3 in March, up 6.2% from a year earlier and the highest since the summer of 2023.

So a contract signed this month assumes the car holds more of its value than cars used to hold, because cars have lately been holding more of their value.

That reasoning has one weakness. It faces the wrong way.

A residual is not a measurement. It is a forecast with a signature on it.

Every one of those gains comes from a lease written in 2022 or 2023. There were no cars to buy then. The guess was set low against a market that kept rising for two more years.

The guess being made this month gets made into something else entirely. More than a million extra vehicles a year are coming back off lease by 2027, and the mix is shifting toward the part of the market with the worst resale record.

I have no idea when this turns.

But a gain on a three-year-old guess tells you about 2022. It tells you nothing about 2029. Those two things are being written into the same contract.

This connects directly to how car lease paper gets priced, and to why the reported numbers look calm.

Ford Credit and GM Financial are the two regular issuers of auto lease securitisations. Both reported residual value gains this year. Ford at 14.2%, GM at 7.1%, per an August 28 note from JPMorgan's ABS research desk.

A gain means the car sold at auction for more than the contract assumed. It is the scoring of a three-year-old decision.

The leases maturing now were signed in 2022 and 2023, when dealer lots were empty and nobody knew whether used prices would hold. The residuals went in low. Then prices climbed for two more years.

Fitch made the same point about its rated pools. Most auto lease deals it covers have been running residual gains, with a handful showing small cumulative losses, concentrated in the all-electric transactions.

So the scoreboard says the underwriters were right. About cars they priced when there were no cars.

And here's where it spreads.

The Inflation Reduction Act treated a leased electric car as a commercial vehicle. That meant the full $7,500 credit applied, with none of the battery and assembly sourcing rules that a buyer had to satisfy.

The finance arms passed it straight through as a reduction in the capitalised cost. A $50,000 electric car leased for about what a $35,000 petrol car cost per month. People signed in record numbers across 2022 and 2023.

Those contracts assumed the car would be worth roughly half its price after three years. Black Book's measure of two- to four-year-old battery cars had fallen to 39% of original retail by June 2024, from a peak near 87% in mid-2022.

Then the credit ended. Black Book now projects new electric vehicles at 7.25% of US light-duty sales, down on the year.

So the policy that created the returning supply also removed the demand that was meant to absorb it.

Black Book reported in July that off-lease electric cars are going to wholesale rather than being bought out, because the market value sits below the contract buyout price. When that happens the driver hands back the keys and the lender owns the car.

Meanwhile, the same collateral is sitting inside European auto lease securitisations.

Fitch measured battery cars at close to 11% of its rated European loan and lease pools by the third quarter of 2024, up from around 6% in 2023. Pool composition tracks new registrations with a lag, so the share keeps climbing after the sales do.

Moody's has made the structural point for years. As leasing replaces buying, residual risk stops sitting with the dealer and starts sitting inside the bond.

There is one offset that sits outside all of these models. Fuel.

Black Book's mid-year read noted that high fuel costs are pushing buyers toward hybrids and electric cars, which should absorb part of the returning inventory. With crude where it is, the used battery bid may hold up better than the 2025 forecasts assumed.

That cuts both ways. The residual gap closes while petrol stays expensive. It reopens the week petrol does not.

Three more things worth keeping track of…

China is buying soybeans again. Daily sales of 272,000 tonnes to China and another 206,500 tonnes to an undisclosed buyer pushed bean sales past 1.2 million tonnes in a week, with November futures up to 1332.25 cents. The trade is reading it as positioning ahead of a Trump–Xi meeting and a possible tariff reduction. Grain markets tend to price political outcomes before political reporters do, because somebody has to load a vessel either way.

Two canola markets stopped agreeing. Winnipeg November canola rose C$7.80 a tonne to C$839.30 on crude spillover and frost risk across the northern Prairies. Paris canola sat almost unchanged at 557.75. Same oilseed, same week, two prices moving apart. When a physical market splits from its European contract, the reason is usually sitting in a field somewhere rather than on a screen.

American output keeps outrunning American hiring. Real GDP grew at a 1.8% annual rate in the first half, while real private domestic final purchases, a cleaner read on underlying private demand, grew at 3.0%. Fed governor Christopher Waller pointed at data centre and technology investment as the driver. Firms are producing more without adding people at the rate the old relationship would predict. It is too early to call that a productivity boom, and too consistent to ignore. We'll see.

In 1999 American factories shipped 348,671 manufactured homes. By 2004 they shipped 130,802.

The industry did not lose to a competitor or to a technology. It lost to its own used inventory.

Through the 1990s the lenders cut down payments from ten or twenty percent to five, and stretched repayment from twelve years out to twenty or thirty. The monthly number got small. Shipments climbed.

Green Tree Financial wrote much of that paper, and booked the profit on each loan the day it was made rather than as it was repaid. Conseco bought Green Tree in 1998 for about $6 billion. By 2000 the combined firm was originating 54% of every manufactured housing loan in the country.

Then the borrowers stopped paying. Through the boom years about 20,000 homes came back annually. Repossessed inventory on lender books went from $300 million in January 1999 to $1.3 billion by the end of 2002.

Every home that came back competed with the ones still on the factory floor. Cheaper, and already built.

That is what killed it. A repossessed house at auction is not a bad loan sitting in a file. It is a rival product with a price tag.

Resale values fell. Lower resale meant thinner recovery on the next repossession, so lenders wrote less. Fewer buyers could get financed, so demand fell, so prices fell again, so recovery thinned again.

At least eight sizeable lenders left the business between late 1998 and early 2002. Conseco filed for Chapter 11 in December 2002 with $52.3 billion of assets against $51.2 billion of debts, the third largest bankruptcy in American history at the time. Oakwood Homes, once the largest builder in the industry, filed the same year.

Employment in manufactured housing went from 43,234 people in 1998 to 27,027 in 2004.

The houses were built well enough. The financing decided what they were worth. We'll see.