One hundred and seventy deaths per million registered vehicle years. That's the Kia Rio's number in the newest federal fatality tables — the highest driver death rate of any model sold in America. Second place, at one hundred and sixty-four, is the Nissan Versa.
Both of which are sitting in a rental lot near you right now, keys on the counter, no apology attached.
Which is roughly what Daniel was getting at. He wrote in asking about the worst cars you can commonly get stuck with in rental fleets around the world — ordinary cars, rentable on a normal licence, hated, whether for a terrible safety record, being horrible to drive, or some other reason they're complete duds. And the part he underlined: grounded in actual data and real drivers' experiences, not vibes.
The good news is that data exists. The catch is it measures something slightly different from what renters complain about.
So there are two lists here and they only partly overlap.
They diverge more than people expect. One list is danger — cars that kill their drivers at an unusual rate, and for that we have the Insurance Institute for Highway Safety's driver death rate tables, which are about as close to an objective worst-car ranking as anything on earth. The other list is misery — cars that are dreadful to sit in, dreadful to drive, dreadful to look at in a hotel car park. That one's built out of professional reviews and years of renter testimony, which is softer evidence but very consistent evidence.
And the danger list is the one with real numbers behind it.
Deaths per million registered vehicle years. The newest round covers 2023 model year vehicles and fatalities from 2021 through 2024, published at the end of July. It's a rate, not a count, so it accounts for how many of each car are actually on the road. That's the trick that makes it usable.
Walk me through why that matters, because I think people hear "deaths per million" and glaze over.
Fair. If you just counted raw deaths, a car that sold four hundred thousand units would look worse than a car that sold forty thousand, even if the popular car were twice as safe per mile driven. The rate divides that out. So a car with a small sales footprint and a bad record can't hide behind low volume, and a car with huge volume can't be penalised just for being common. It's the difference between "more people died in this car" and "you are more likely to die in this car," and only the second one is a useful sentence.
And the second one is the one that should change what you book.
It should. It rarely does, because the rental counter doesn't show it to you.
The honest limitation, before we go further — this is American data.
Almost entirely. There is no single global ranking of worst rental cars. Nobody has built one. The European and Asian fleet staples — the Fiat Panda, the Dacia Sandero, the Kia Picanto, the Hyundai i10 — they turn up constantly in renter complaints and almost never in structured crash data of this kind. So when we say worst, we mostly mean worst as America measures it, and we should say that out loud rather than pretend otherwise.
Why hasn't anyone built the global version? It seems like the kind of thing someone would have done.
Because it requires three things that don't coexist. You need fatality data that's nationally complete, you need vehicle registration data granular enough to calculate exposure, and you need both of those to be public. America has all three. Most countries have one or two. So the global ranking isn't withheld — it's just not constructible. You'd be stitching together incompatible datasets and quietly guessing at the joins.
Which is exactly the kind of thing that produces a viral chart and a retraction.
Every time.
Right. The list nobody wants to top, then. Because those numbers are stark.
The Kia Rio at one hundred and seventy. The Nissan Versa at one hundred and sixty-four — and the Versa is arguably the single most common economy rental in North America. Then the Chrysler 300 in two-wheel drive at one hundred and thirty-two. Dodge Charger HEMI, one hundred and twenty-seven. Dodge Challenger, one hundred and twenty-five. Kia Forte at one hundred and twenty.
That's the point where the list stops looking like a rental counter and starts looking like a rental counter plus a hen party.
The Charger and the Challenger are a different phenomenon and we'll come back to that, because the reason they're on the list is not the reason the Rio is on it. Keep going: Chevrolet Malibu at one hundred and sixteen. Hyundai Venue, two-wheel drive, one hundred and thirteen. Nissan Altima and Nissan Maxima, both at one hundred and five. Hyundai Kona two-wheel drive at ninety-five. Kia K5 at ninety-four. Nissan Kicks at eighty-six. Hyundai Sonata and Subaru Impreza at eighty-three. Nissan Sentra at eighty-one.
So of the twenty worst, how many can you rent without trying?
Most of them. That's the part that should stop you. This isn't a list of exotics. It's the aisle you walk down when you've booked "economy or similar" and you don't get to pick which similar.
And there's a class number underneath the model numbers that matters more than any single badge.
Minicars average one hundred and fifty-eight driver deaths per million registered vehicle years. The fleet-wide average is thirty-seven. That's roughly four times the risk, and it's a property of the segment, not of any one manufacturer. Small cars as a class average seventy-nine — better than minicars, still more than double the fleet.
So swapping the Rio for a Corolla helps. Swapping the Rio for a different minicar doesn't.
Barely at all. You've moved within the dangerous class, not out of it.
Why? Because physics doesn't care about your build quality.
Exactly that. The IIHS president, David Harkey, put it plainly: even the safest small cars don't fare well in crashes with large pickups and heavy SUVs. That's mass mismatch. When a two-thousand-eight-hundred-pound hatchback meets a vehicle that weighs double that, the small car absorbs the difference, and no amount of engineering in the small car changes the arithmetic of the collision. So the minicar death rate is largely a statement about what else is on the road.
Think of it as a conservation problem. In a collision, momentum has to go somewhere. If a heavy vehicle hits a light one, the light one gets a much larger change in velocity over a much shorter time, and the occupants inside it experience that as force. A heavier car in the same crash experiences a smaller velocity change, so its occupants feel less. That's it. That's the whole mechanism. Airbags and crumple zones manage the inside of the box. They don't change how much the box gets shoved.
So the crumple zone is doing its job. It's just that the job is smaller than the problem.
Beautifully put, and slightly depressing.
Which is a strange thing to sit with. The car isn't badly made. It's badly outmatched.
And the downstream effect of that is uncomfortable. If you make the small car safer — better structure, more airbags, better crumple zones — you add weight and cost, and it stops being the cheap car. There is a version of this where the safest small car is simply a midsize car, which is the entire problem in one sentence.
So the segment can't be fixed without being abandoned.
That's the trap. And you can watch the industry walk into it in real time.
Now the historical comparison, because it's worth a beat rather than a victory lap.
The earlier round of these tables, reported in the middle of 2023, had the Mitsubishi Mirage G4 at two hundred and five driver deaths per million and the Mirage hatchback at one hundred and eighty-three — the two worst in America by a wide margin. The Hyundai Accent at one hundred and fifty-two and the Chevrolet Spark at one hundred and fifty-one were close behind.
And the newest tables put the Rio at one hundred and seventy, below the Mirage's old two hundred and five.
Which reads like progress until you notice that the Mirage mostly left the market rather than improved. The Rio at one hundred and seventy is still the worst car in the country. The number came down because the worst car went away.
The list got better by subtraction.
There's a much more interesting wrinkle in this data, and it's the one that stops you reading the whole thing as an engineering verdict. The Volkswagen Atlas.
The three-row family bus.
In two-wheel drive form, the Atlas is among the safest vehicles in the fleet for its own driver. In four-wheel drive form, it's among the worst for the people it hits — one of the highest other-driver death rates going. Same platform, same badge, same crash structure, wildly different position depending on which column you read.
How does the same car end up on both ends of the table?
Because the two-wheel drive and four-wheel drive versions aren't bought by the same people or driven the same way. The front-wheel-drive Atlas is the suburban family hauler. The all-wheel-drive one skews to a different buyer in different conditions. Same metal, different population behind the wheel, and the table is measuring the population as much as the vehicle.
Which means the table is partly measuring who drives the car.
And how they drive it, and where, and how many miles. Exposure. A vehicle used as fleet transport or high-mileage commercial work shows up differently from one doing the school run. The death rate is a real number but it's a measurement of a population, not a property of the metal.
So the correct sentence is "people who drive this car die at this rate," not "this car is dangerous."
Which is the sentence nobody puts on the rental listing.
Now the muscle cars, because they're on the same list for a completely different reason.
The Charger, the Challenger, the Chrysler 300 — Harkey's explanation is not mass mismatch. His line is that powerful engines and marketing that attracts aggressive drivers is the issue, and that they're dangerous mainly because of how they're driven. That's the cleanest split in the data. The small cars are dangerous because of what hits them. The muscle cars are dangerous because of who's holding the wheel.
Same table, two entirely different failure modes. One is a physics problem, one is a behaviour problem.
And it's why the rental angle on the Charger is quietly grim. A rear-wheel-drive V8 with three hundred horsepower, handed to somebody who has just flown six hours and is driving an unfamiliar road at night — that's the exposure profile the table is describing.
And it's marketed as the upgrade. You pay extra for the car with the death rate.
It's the only line item in travel where the upsell is statistically worse for you than the base rate.
Which raises the obvious question — why does the industry keep buying them?
Because rental fleets buy what rents. The Charger rents. It's the car people photograph in the lot. Nobody upgrades to a Versa. So the fleet keeps stocking the car with the behavioural death rate, and the customer keeps paying the premium to be in it, and both sides feel like they won.
That's the danger story. Danger and misery are two different lists, though, and the second one is where renters actually live.
And here the leading character is the Nissan Versa, which is the most consistently named rental dud anywhere you look. The Car Connection's line on it is worth quoting exactly: few cars feel built down to a price these days, but the Nissan Versa is certainly one of them. Their complaint about the engine is that it's gutless but not especially fuel efficient — so it's slow and it doesn't even buy you anything at the pump.
That's the piece that would annoy me. Slow I can live with. Slow and thirsty is an insult.
The renter testimony is where it gets specific. A FlyerTalk user describing a filthy 2016 Versa — their words — said the entire car shook if you got over sixty miles an hour. Not rattled. Shook.
Sixty is the speed limit on half the roads in America.
And my favourite line in the entire research pile, from a commenter calling himself Cheap Bastard on The Autopian, describing a Versa as powered by an underfed hamster with a moped's transmission.
That's not a review. That's a diagnosis.
The Mirage gets similar treatment and earns it more thoroughly. Seventy-eight horsepower from a three-cylinder. The Car Connection rated the 2019 model at three point three out of ten, citing dismal acceleration, scary highway handling, very noisy at speed, and lousy safety ratings. Edmunds gave the 2020 model five point seven out of ten for poor build quality and an underpowered, noisy engine.
Three point three out of ten. On a scale where most cars cluster between seven and nine.
And it wasn't free. At the time The Drive wrote about it, a Mirage was going out the door at seventeen thousand six hundred and fifty dollars. Cheap, but not free.
So you're paying real money for a car that scores like a rented bicycle.
Then there's the supporting cast. The Chevrolet Spark — ninety-eight horsepower, a cramped two-door layout, and weaker fuel economy than the Kia Rio and Hyundai Accent it competes against. The Chrysler 200 and its Sebring ancestor, which Jeremy Clarkson named the worst car ever built: a gutless two point seven V6, a transmission he described as horrid, handling that was floppy to the extreme.
Floppy is such a specific word for a car.
It means the body moves after the wheels do. That's the sensation.
So you turn, and then a moment later the car agrees with you.
And by the time it agrees you've already made the correction, so it's constantly one beat behind you. That's what people mean when they say a car feels unsettled. It's not that it can't grip. It's that it responds late.
The Jeep Compass and Patriot are on the list too.
Loud, slow, and inefficient, and they went about a decade without a redesign, which is how you end up with electrical faults like wipers that come on by themselves and cruise control that engages when it feels like it. And the Chrysler Pacifica has a complaint I find more serious than any of the comfort ones — drivers reporting that the digital instrument cluster and the infotainment screen go blank occasionally.
Blank cluster means no fuel gauge and no gear indicator.
Which is not a comfort problem. That's a you-don't-know-what-gear-you're-in problem.
And it underlines the split we've been drawing. Most of the misery list is annoyance. A blank instrument cluster is not annoyance. That's a car failing at the one job it has.
Which is why the misery list needs its own internal ranking, and nobody's built that either. A slow engine and a blank dashboard are not the same category of complaint, and lumping them together is how you end up with a list that reads as snobbery rather than warning.
On the question of why these cars keep showing up on the lot —
It's not the Versa.
... Sorry?
It's not the Versa. You've given it four minutes and it isn't the worst thing on a rental lot. It's the Dodge Journey.
Hilbert.
I've seen the paperwork. Regional depot, back office, and the reason the Journey kept appearing long after they stopped selling it new was a bulk-purchase clause tied to a discontinued trim level. They bought the run, they had to take the run, and the run was fourteen hundred of them. All the same colour. Not one with a working USB port.
Fourteen hundred Dodge Journeys, one colour, no USB.
Silver. And the USB ports worked when they left the factory. They stopped working because nobody was maintaining them, which is a different failure and a more honest one.
Hang on, how do you end up reading a rental company's regional depot paperwork?
I was there about something else.
About what?
A door. A door that didn't close properly on a vehicle I was inspecting for an entirely separate reason, and while I was waiting for someone to come back from lunch I sat in the office, and the file was on the desk. I don't think anyone minded. I don't think anyone knew I was there, now that I say it out loud.
The fourteen hundred figure is from a file you read while nobody was in the building.
The figure is correct. The circumstances were incidental.
What was the separate reason?
Doesn't matter. The point is the Journey. Everybody piles on the Versa because it's slow. The Journey is a vehicle they kept building for eleven years with the same engine and the same four-speed automatic, and there's a reason it turns up in every worst-rental list ever written and nobody at the counter will ever say its name.
Fourteen hundred, though.
Fourteen hundred and six. I rounded down.
You rounded down to a more suspicious number.
It's the same number to anybody who isn't being difficult about it.
Right. Well. Setting that aside — the thing that stays with me is what happens when the segment gets killed instead of fixed.
That's the real knock-on effect here, and it's the one that should worry you. The Mirage, the Spark, the Rio — all of them are ending production. The segment is being discontinued, not improved.
The Auto Wire put the structural version of that better than anyone: the cheap small car is the most dangerous seat in America, it's what you can afford if you're new, young, or working two jobs, and the industry solved the problem by discontinuing the cheap small car rather than fixing it.
The people who needed it don't stop needing a car. They buy it used, five years later, in worse condition, with the safety record intact.
The danger doesn't leave the road. It just changes hands.
It changes hands downward, in the direction of the people with the least room to say no. That's the bit that turns a data story into a policy story, and it's why the table matters more than the misery list even though the misery list is more fun to read.
There's a complication worth sitting with, though, because it cuts against the "worst ever" framing.
The nostalgia inversion. The Autopian ran that question — what's the worst car a rental company has forced on you — and got two hundred and sixteen comments, and a striking number of them made the opposite argument. That the Sebring and Impala era was worse. That there are no horrid cars any more.
Which is probably true on the misery axis and definitely not true on the danger axis.
Both things are real. The average rental car today is better built, better equipped, and more reliable than the average rental car of fifteen years ago. It's also, if it's in the minicar class, roughly four times as likely to kill its driver as the fleet average. Comfort improved. Physics didn't.
There's a reason those two facts can coexist without contradiction. The things that improved — materials, infotainment, build quality, reliability — are all things you can fix with money and engineering. The thing that didn't improve is mass, because mass is the thing you're paying not to have.
Which is the whole argument in one line. You can make a cheap car nice. You can't make it heavy without making it expensive, and if you make it expensive it isn't the cheap car any more.
The one thing that stays with me out of all of it — the rental counter is where you meet the used-car market of five years from now. The cars being killed off today are the ones being handed out tomorrow, to the people with the fewest options.
Nobody has ever built the global ranking Daniel was effectively asking for. It doesn't exist. What exists is a very good American dataset, a lot of very consistent renter complaining, and a purchasing-economics thread we can only infer rather than source properly.
A thread worth pulling on another day.
Thanks to our producer, Hilbert Flumingtop.
This has been My Weird Prompts. If you've got a prompt of your own, send it to us on Telegram at t dot me slash MWP listener bot.
We'll be back soon.
