The Car Got Better. The Deal Got Worse.

Right up top, I must admit that I am a huge car guy. I love old cars. I want to feel connected to the machine. I miss rowing through the gears on a manual transmission. In that sense, modern cars are unequivocally worse than their older counterparts.

So I went into this expecting to write an article bashing the industry about the myriad ways that cars themselves had gotten worse. I figured I’d find that components had gotten flimsier. That vehicles now don’t last as long as they used to. But after looking at the data, that just simply isn’t true. By almost every measure, the new car in your driveway is better than any equivalent in history.

A lot did get worse. Just not the car itself.

First of all, price. Adjusted for inflation, new vehicle prices are down roughly 45% since the early 1980s, according to BLS data. Before you type up a reply asking how that can possibly be true, know that the BLS index tracks a like-for-like car. It simply asks what a 1985 car’s worth of metal and capability costs today. That bundle of components costs a little over half of what it cost in 1985, adjusted for inflation. Cars feel more expensive because it is no longer possible to buy a simple car with limited features. I’ll expand on this idea in a moment.

Cars last longer. In 1970 the average American car on the road was 5.6 years old. By 1999 that number had extended to 8.9 years, according to FHWA tables built from Polk registration data. Today S&P puts the average passenger car at 14.5 years. Americans keep cars roughly two and a half times longer than they did in 1970.

They kill fewer people. In 1980 there were 3.35 deaths per hundred million miles driven. By 1999, 1.57. In 2024, 1.19. Of course, better roads, safety laws and emergency medicine all contributed, so the car is one input among several. But it is a large one.

They use less fuel. EPA puts model year 2024 economy at a record 27.2 mpg, up from about 19.2 twenty years earlier. A 41% improvement, achieved while vehicles got heavier, faster and larger at the same time.

And they get recalled less. NHTSA recalled 31.3 million vehicles in 2025, the lowest total since 2013, down every year since 2020.

Cheaper (with a caveat), longer lived, safer, more efficient and recalled less. The machine itself is better for the average consumer.

There is a lot to cover here, so I will be publishing four essays on the automotive sector over the coming months:

This first essay covers the transaction. The vehicle itself is now the least profitable part of the automotive business. Dealers now make nearly as much money arranging your financing as they do selling you the car. The way they make it is by marking up the interest rate the lender already quoted them. The average loan has doubled in length since 1971. On top of that, the cheap, no-frills car has been discontinued outright.

The second covers the software. Basic functions of the car that used to be physical parts you owned as soon as you bought the vehicle are now integrated into the software suite. You access these through the infotainment system, which has become the worst-scoring category in the industry's own reliability study. Phone pairing is the single most reported problem in American cars. The same failure-prone infotainment is also a revenue line. Hardware already fitted to your car, switched off until you subscribe. Not to mention the added perk of being able to sell your driving data.

The third covers the supply chain. Cars have always been assembled from other companies' parts; nothing has changed there. The issue is how concentrated the supplier network has become. When a single supplier fits the same component to nineteen manufacturers, a single defect becomes a market-wide event. Takata's airbag inflators did exactly that. Roughly 67 million have been recalled, in 34 brands from Ferrari to Mazda to Tesla. The NHTSA has confirmed 28 related deaths in the US. That is just one of many similar stories.

The fourth covers our right-to-repair (or lack thereof). Service and parts are now the most profitable department in the dealership. This has created an incentive for manufacturers to make repair harder to get outside of their dealer network. A minor collision on a car with driver assistance sensors costs roughly twice what the same crash cost before them. AAA's guidance is that returning those systems to service "requires special training, tools and information." Meanwhile, Ford's CEO, Jim Farley tells investors that parts and service is a $15 billion business growing 8% a year, and that its great virtue is being anti-cyclical: "when the car business goes down, people tend to repair their vehicles." Sure, but Jim… the difference is we used to be able to repair our cars ourselves.

We start with the financialization of the transaction, which is arguably the most societally damaging trend I have covered to date.

Back to that 45% price-reduction figure I mentioned in the intro.

The BLS price index measures a constant car. It asks what a fixed bundle of metal and capability costs this year against last year, and adjusts for all of the new features that now come standard with vehicles. Backup cameras have been federally required on every American vehicle since 2018, so the index treats them as a quality improvement and subtracts their cost before comparing. On that basis, yes, cars are cheaper. But they feel more expensive because you can no longer buy the no-frills version.

If you measure instead what people actually pay, against what people actually earn, you can clearly see the decline in affordability. Cox Automotive counts how many weeks of median household income it takes to buy the average new vehicle, financing included. In December 2020 it was 32.2 weeks. This June it was 35.3.

Comerica Bank ran a similar index for decades before Cox launched theirs. It used median family income rather than household income, so the two series should not really be compared directly. But between 2007 and 2012, Comerica's index never once went above about 25 weeks, and it bottomed out at 21.8 weeks in 2009.

However you handle the methodology gap, the direction is not in question. A new car costs materially more of your working life than it did fifteen years ago.

A like-for-like car got cheaper. The car available for sale got much more expensive.

A couple of factors explain what is going on.

The first is what the manufacturers decide to build. In model year 2010, sedans and wagons accounted for 54.5% of American vehicle production. By 2024, they had dropped to 23.7%. Truck-based SUVs went from a fifth of the market to almost half of it over the same period. J.D. Power has trucks and SUVs at 83.1% of retail sales as of January 2026. The industry moved the median buyer out of the cheaper category and into the more expensive one.

The second is what stopped being built. Nissan made the last cheap car in America. The Versa started at $18,585, and production ended in December 2025. Nissan's explanation: "In line with Nissan's product strategy, the Nissan Versa ended production in December 2025 for the U.S. market."

For the 2026 model year, nothing sold in America starts under $20,000. The cheapest new car you can buy is a Hyundai Venue at $22,200, after destination charges.

The industry discontinued the no-frills compact car, and sold you the larger one with more equipment in it. It then went looking for a way to make the payment feel survivable.

The Federal Reserve has tracked the average length of a new car loan since 1971.

In June 1971 the average new car loan ran 35 months. By 1980 it was 44. By 1985, 50.7. By January 2011, when that particular series ended, the average loan was 62.3 months long. The successor series put it at 66.1 months in December 2025.

Today it is about 70.

Two firms measure car loans independently today, using different methods and different data. Experian reads credit files and puts the first quarter of 2026 at 69.48 months and $770 a month. Edmunds reads dealer transactions and puts it at 70.3 months and $773. They agree to within a percent.

Experian also found 35.55% of new loans that quarter had terms longer than six years, up from 30.83% a year earlier.

Edmunds' second quarter 2026 came in higher again. The average buyer financed $44,156 at 7.0% and paid $777 a month, a record for the third consecutive quarter. Over the life of that loan they will pay $9,811 in interest.

This did not happen because borrowing got more expensive. It got much cheaper. A new car loan at a commercial bank cost 13.37% in 1985. The same loan costs 7.47% today.

Hold the loan amount steady and change only the length of the term. Financing $44,156 over the 36 months customary in 1985, at 1985's rate, costs about $1,500 a month and $9,688 in interest. Today's buyer at today’s rate over a 70-month term pays $777 a month, and $9,811 in interest.

The rate fell by nearly half. The interest did not fall at all. The length of the loan ate the entire benefit, and what the buyer got instead was a payment that looks half the size.

The cheapest car on the lot got more expensive, and the loan got long enough to keep the monthly payment roughly where buyers expect it. That is what the extra years are for. Not to save you money, but to move a price you could not afford into a payment you can.

The strain shows in who is signing. Subprime has gone from 14.40% of vehicle financing to 15.75% in a year. Nearly a quarter of buyers now commit to seven years or longer. And a buyer who trades in a car they still owe money on will pay an average of $16,270 in interest.

This is why I say the financialization of the automotive industry may be the most societally damaging trend I have covered to date. A bad frying pan costs you eighty dollars once. Americans now owe $1.685 trillion on their cars. They added $182 billion in new debt last quarter. In most of this country, owning a car is not optional. Demand is inelastic. A car is the thing that gets you to the job that pays for the car. Auto-makers understand this. They know that opting-out is rarely an option.

GM's 2025 results put GM Financial at $2.8 billion in earnings before tax on $17 billion of revenue. A margin of 16.4%. The same release puts GM as a whole at $12.7 billion on $185 billion. That is only a 6.9% margin for the whole business.

Ford's own release tells the same story, but worse. Ford Credit earned $2.6 billion before tax on $13.8 billion of financing revenue, a margin of 18.8%. Ford as a whole earned $6.8 billion of adjusted EBIT on $187.3 billion of revenue. That is only a 3.6% margin.

Lending is more than twice as profitable as manufacturing at GM, per dollar of revenue. At Ford it is more than five times as profitable. Ford Credit accounted for 38% of Ford's adjusted profit last year.

Some of that gap is Ford having a bad year. It lost $8.2 billion on a GAAP basis in 2025 and its adjusted profit was barely half of GM's. But that is exactly what a finance arm is for. It earns when the car business does not.

Presidio-NCM benchmarks more than 4,000 American franchised dealerships and reports their profit by department. In the first quarter of 2026, selling a new car produced $1,781 in gross profit per unit, down 11.2% from a year earlier.

Arranging the financing and insurance on that same car produced $1,727 per unit, up 7.1%.

Service and parts, meanwhile, generated 53.8% of all dealership gross profit, the highest share in the benchmark's history.

At the biggest dealer group in the country, the story is the same.

AutoNation made $2,769 per unit on finance and insurance in 2025 against $2,564 on the vehicle itself, and by the first half of 2026 the gap had widened to $2,826 against $2,444. Its CEO Michael Manley told investors that "approximately 80% of our profits come from CFS and after sales."

So financing is a bigger share of this business every year, which should strike you as strange. The dealership is not lending you anything. A bank or a credit union is. The dealership introduces you to them.

Which raises the question: what is the dealership finance office actually being paid for?

In 2013 the CFPB and the Justice Department ordered Ally Financial to pay $98 million. Explaining how the arrangement worked, they wrote this:

"Ally sets a risk-based interest rate, or 'buy rate,' and then allows auto dealers to charge a higher interest rate when they finalize the deal with the consumer. This is typically called 'dealer markup.' Ally then shares some or all of the revenue from that increased interest rate with the dealer."

The lender decides your rate based on your credit history. The dealer then decides how many additional points they can get you to agree to. The resulting difference is split between the dealer and the lender.

Dealer profits are falling, by the way. Presidio has average net pretax profit per dealership down 11.2% in the first quarter of 2026 and 11.8% in the second. Selling cars is a hard business and it is getting harder. I have sympathy for that.

What I have less sympathy for is where the shortfall went. A business under pressure is entitled to raise its prices. That is what prices are for, and a dealer who told you plainly that the car costs more this year would be doing nothing wrong at all.

The finance office recovers the money a different way. It seeks only to obfuscate. It arranges the transaction so that you cannot see what you are being charged. The markup on your rate is never quoted, and the products arrive presented as conditions of the sale.

At AutoNation, three quarters of cars leave with a finance contract. Michael Manley says buyers there purchase "on average more than 2 products per vehicle with extended service contracts."

Nobody publishes the margin on those. It sits with the administrators, JM&A and Zurich and Assurant, and is not in any filing I could find.

What is public is what happens when regulators take an interest. In November 2023 the CFPB fined Toyota Motor Credit $60 million. Bundled products had added between $700 and $2,500 to individual loans. More than 118,000 calls from customers trying to cancel were routed to a dedicated retention line, where representatives were instructed to make them ask three times before accepting a written request. Refunds, when they came, were applied as principal payments rather than issued as checks.

In December 2025 the FTC mailed $9.6 million to 168,179 people who had bought vehicle service contracts from CarShield. They had paid up to $120 a month for coverage that did not cover their repairs.

In December 2023 the FTC finalized the CARS Rule on a 3-0 vote. It banned bait and switch pricing on vehicles and financing terms. It banned charging for add-ons that provide no benefit. It required dealers to disclose the actual offering price, to say plainly that add-ons are optional, and to obtain express informed consent before charging for one.

The Commission estimated the rule would save Americans more than $3.4 billion a year and about 72 million hours of their time.

Thirteen months later the Fifth Circuit vacated it. They held: "The FTC violated its own regulations when it failed to issue an ANPRM for the CARS Rule."

An ANPRM is an advance notice of proposed rulemaking. It is a procedural step that comes before the proposal that comes before the rule. The court did not vacate the CARS Rule because it was not an effective policy. It vacated because it ruled that the FTC did not file the right document.

Nothing replaced it. The FTC now pursues the same conduct one dealer at a time.

The standard bumper to bumper warranty in America is three years or 36,000 miles. Chevrolet, Ford, Toyota, Honda, Nissan, Subaru, Mazda, Jeep and Chrysler all sit there.

More than a third of new car loans now run past six years, which means the warranty expires with three years of payments still left.

Covering warranty liabilities costs big money, and manufacturers have to set it aside in advance, before they know what will fail. Ford put $6.71 billion into its warranty reserve last year. GM put in $5.07 billion.

Both figures are climbing. Measured against sales, the American industry ran warranty stockpiles at about 1.26% between 2015 and 2019. Ford is now at 4.0% and GM at 3.0%, two-to-three times the old rate.

Some of that can be explained by the price of a repair going up rather than more repairs happening. Some is manufacturers revising estimates on cars they sold years ago. The direction of the trend is the important thing, and this is one area where the claim that the cars are becoming less reliable has some evidence behind it. It also varies enormously by company: Toyota set aside $561 per vehicle in 2024, Stellantis $1,230.

It is money the manufacturer wants back. Kumar Galhotra, Ford's chief operating officer, calls warranty "the largest component of our competitive cost gap" and "a major cost opportunity for us."

There are two ways to reduce a cost like that. Either build cars that break less often, or have them break later, once the warranty window has closed and the bill has become the owner’s.

Whether anyone engineered this transition is a question I won’t answer here. It requires evidence about how cars are actually built. That is the second essay.

Arrange your financing before you go to the dealership. A credit union or bank approval in hand turns your time in the finance office from a negotiation into a comparison. This single step removes the markup described above, because there is no buy rate to mark up.

If you do finance through the dealer, ask what the buy rate is. Ask what rate the lender approved you at, as distinct from the rate on the contract in front of you. You are entitled to know both, and the question itself tells the person across the desk that you know the difference exists.

Judge the loan by total interest, not the monthly payment. Every lengthening of the term I described above was sold as a lower payment, and each one cost more money. $9,811 is the current average (I recommend you aim well below that). Ask for that number, in dollars, before you sign.

Decline the add-ons in writing. Extended service contracts, gap insurance, tire and wheel protection, paint sealant. If you want a service contract, buy it later from someone who is not also selling you the car, and know that you can almost always cancel one for a prorated refund. Toyota Motor Credit built a phone system to make that hard. Cancelling is still your right.

Check the term against the warranty. If the loan outlives the coverage by three or four years, you have agreed to keep paying through a period where every repair is yours. That may still be the right deal, but it should be a decision rather than a surprise.

And never roll negative equity into the next loan.

Above all, if you can, avoid the loan altogether and pay cash. Nowadays, a 10-year-old used car that was well looked after will still be in fantastic shape, with all of the features you could want.

If that is the difference between you having a loan and avoiding one, I’d go used over new.

The car, in many ways, is the best it has ever been. The transaction wrapped around it is the worst it has ever been.

The cars and the production processes around them got so good that the industry stopped making its money on the vehicles themselves. It discontinued the cheap version, moved you into a bigger one, stretched the loan until the payment looked survivable, and took its margin on the lending, the paperwork and the service bay instead.

I still miss the manual gearbox… but that is a complaint for another day.

Next in this series: where the car actually got worse. Cars fail more often now, and less catastrophically. The drivetrain lasts; the screen, the sensors and the electronics do not.

Join the discussion, share product recommendations and suggest topics for me to investigate next in the Worse on Purpose community. Become a member to join.