Safety · A Century of Progress
The Second Collision
The car has become one of the safest machines humans routinely trust with their lives — the death rate per mile has fallen more than ninety percent in a lifetime. It happened because a handful of people rejected the idea that a crash was an act of God, and treated it as an engineering problem instead. Here is how, decade by decade.
Before the history — a few words on why it took so long
It is astonishing to me that a species this capable — one that had already engineered safety into medicine and hospitals, into the buildings we live in, into the great ships that cross the oceans, a species whose most primal instinct is to stay alive and out of harm's way — looked at the automobile and, for the better part of a century, waved the whole idea of safety away.
Break a car down to its physics. It has mass, a great deal of it. It moves quickly — even the early ones did. And when it strikes something, all of that becomes force. From the time we are children we understand this instinctively: pick up a heavy rock and throw it, and you know, without being told, that it will do more damage than a small one, or than the same rock simply dropped. A car is that idea scaled up to the terrifying — tons of weight, moving fast — with one crucial twist. When the car stops, you do not. You carry your own inertia; you keep traveling until something stops you too. Physics is relentless that way, and, if you let it be, fascinating.
So why the reluctance? Why, for decades, the pushback and the stalling every time someone raised the subject of making a car safer?
There is a story from the Second World War. Bombers were coming home riddled with holes, and the instinct was to bolt extra armor onto the places where the bullet holes clustered. Reasonable enough. But a statistician named Abraham Wald pointed out that they had it exactly backward: reinforce the places with no holes at all. The planes shot there were the ones that never came home; the survivors, by definition, had been hit only where a plane could afford to be hit. It is called survivorship bias, and it is the whole history of car safety in miniature. Some engineers looked at the wreckage and shrugged — nothing to be done, it is what it is. Others looked at the same wreckage and read it backward, and saw exactly where the armor belonged: a bigger brake here, a belt to hold the passenger there, a stronger frame around them. Sometimes it takes legislation to force the point. Sometimes it takes hard experience. It is almost never common sense — which is a strange thing to have to say about staying alive.
That reluctance hardened, for the first half of the automobile's life, into something like doctrine. A crash was simply too violent to survive by design; the forces were beyond anything a body or a vehicle could be engineered to absorb; and so road death was treated the way people once treated death from infection — a tragic constant, not a solvable problem. The idea that finally broke the fatalism has a name, and it is the one this whole story is built around: the second collision.
The first collision is the car striking something — a wall, a tree, another car — and it is largely over in a tenth of a second. The second collision is the one an instant later, and it is the one that kills: the occupant, still carrying all that inertia, thrown against the steering column, the dashboard, the windshield. Name the second collision as the enemy, and suddenly you can engineer against it — the interior, the restraints, the very structure of the car. That reframing turned road death from fate into a design brief, and what follows is the story of how thoroughly it worked — told, because the arithmetic is the argument, largely in dates and numbers.
The great decline: U.S. deaths per 100 million miles driven
The rate has fallen by more than 90% since the 1920s, even as Americans drive vastly more. Longer bars are more dangerous.
Sources: NHTSA / FARS historical series. 2025's 1.10 is the second-lowest rate ever recorded, behind only 2014's 1.08.
The idea that couldn't be built
Road death is as old as the road machine. Mary Ward, thrown from a steam carriage in Ireland in 1869, is the first recorded victim; Henry Bliss, struck stepping off a New York trolley in 1899, was the first in North America. The earliest safety work was quietly mechanical: the 1922 Duesenberg Model A brought four-wheel hydraulic brakes; Ford made shatter-resistant safety glass standard in 1930; General Motors ran the first barrier crash test in 1934. But the cars themselves remained, in hindsight, instruments of injury — rigid steel dashboards, non-collapsing steering columns studded with knobs and levers, plate-glass windshields that broke into blades, and body structures so stiff they transmitted the full force of an impact straight into the people inside. A few voices pushed back early: in the 1930s the Detroit plastic surgeon Claire Straith, who spent his days rebuilding faces smashed on dashboards, and the physician C.J. Strickland, who founded an Automobile Safety League, campaigned for seat belts and padded interiors. Almost nobody listened. The notion that a car could be designed to protect its occupants barely existed.
The surgeons and the second collision
The breakthrough came, tellingly, from outside the industry. In 1942 an aviation researcher named Hugh DeHaven — who had survived a mid-air collision himself — published a landmark analysis of people who had lived through falls from fifty to a hundred and fifty feet, and argued that a body could withstand enormous force if it were properly "packaged," with the load spread and the deceleration slowed. The Cornell Aeronautical Laboratory turned that intuition into rigorous study, and its conclusion was blunt: the two things that mattered most were seat belts and padded dashboards. The ideas began, tentatively, to reach cars — the 1947 Tucker offered a padded dashboard and a pop-out windshield; the 1949 Saab 92 borrowed aircraft thinking to give its body a rigid safety cage. The second collision, it turned out, could be engineered after all. The industry simply had to be made to care.
Two Swedes and an Austrian
Two inventions, arriving almost together, turned the idea into hardware — and both were given, in effect, to the world. In 1959, the Volvo engineer Nils Bohlin designed the three-point seat belt, the lap-and-shoulder restraint still in your car essentially unchanged, and Volvo made the extraordinary decision to open the patent, letting every rival use it free, judging the lives saved more important than the advantage kept. It is one of the great corporate decisions in industrial history; three-point belts have since saved well over a million lives, and reduce the risk of dying in a front seat by roughly forty-five percent.
At Mercedes-Benz, the Austrian engineer Béla Barényi had already patented, in 1952, an idea that sounds backward until you feel it work: a car should have a rigid passenger cell wrapped in crumple zones deliberately built to collapse, absorbing crash energy by deforming over distance instead of slamming it into the occupants. The first car engineered this way, the Mercedes W111, appeared in 1959 — the same year Volvo standardized the belt and American Motors offered the first head restraints. The physics is a single elegant equation: force multiplied by time equals the change in momentum. Stretch the collision out over a few more milliseconds, and the force reaching the human body plummets. When the Insurance Institute crashed a 2009 Malibu into a 1959 Bel Air fifty years later, the demonstration was almost cruel: the old car's cabin crushed inward onto its dummy while the modern car's cell barely deformed. Same speed, same physics — one had simply learned to buy time.
A crash lasts about a tenth of a second. Nearly the entire history of automotive safety is the art of making it last a little longer.
A century of advances, in order
The progress that followed did not arrive as a single revolution but as a long accumulation — a century of specific inventions, laws, and standards, each buying back a share of the toll. Not every one can be given its own chapter, but every one belongs in the record. Here is the essential timeline.
The road to a safer car
Major milestones in automotive safety, 1922–2029. Passive protection first, then the turn toward preventing the crash entirely.
Unsafe at any speed
For years the ideas outran the industry's willingness to fit them, and the log above hides a fight. It broke open in 1965, when a young lawyer named Ralph Nader published "Unsafe at Any Speed," arguing that carmakers knowingly sold dangerous vehicles and that design — not merely "the nut behind the wheel" — was killing people. When General Motors was caught hiring private investigators to dig up dirt on him, the scandal detonated in the industry's face, made Nader a household name, and helped drive the 1966 law that created the federal agency now known as NHTSA, with the power to set binding standards. The first standards, in effect by 1968, mandated seat belts, collapsing steering columns, padded panels, dual-circuit brakes and side-marker lights; head restraints followed in 1969. Just as important, the agency's first leader, the epidemiologist William Haddon, reframed road death as a public-health epidemic to be attacked systematically. His "Haddon Matrix" split every crash into three phases — before, during, and after — across three factors — the human, the vehicle, and the environment — and insisted that all nine cells offered places to intervene. Safety stopped being a matter of blaming drivers and became a science of prevention.
The long build-out
From there the protections compounded. The airbag had a decades-long gestation — patented in the 1950s, offered by GM as an experimental option in 1974, and made genuinely practical by Mercedes in 1981 as a supplement to the belt. The Porsche 944 Turbo made dual airbags standard in 1987; Chrysler put driver's airbags in every U.S.-built car by 1989; a federal mandate for dual front airbags took effect for 1998. They are credited with saving 28,244 lives in the United States through 2009 — though the same period exposed their danger to children and small adults sitting too close, which is why modern "smart" airbags meter their force by occupant size and seatbelt use. Antilock brakes, with roots in aviation, reached the road in 1978 and marked the first turn toward preventing crashes rather than merely surviving them. And accountability arrived through the courtroom: the Ford Pinto's rupture-prone fuel tank and the landmark Grimshaw verdict of 1978 established that cutting safety corners carried ruinous legal cost, not merely moral cost — a lesson that did as much to change corporate behavior as any regulation.
What actually saves you: reduction in risk, by system
Approximate reductions credited to each measure. They stack: a belted occupant in a car with airbags and stability control is protected several times over.
Sources: NHTSA and IIHS effectiveness estimates. Figures vary by crash type and vehicle.
The other half of the problem
Engineering can only do so much when the driver is drunk, unbelted, or speeding, so a parallel campaign attacked behavior — and it may have saved as many lives as any device. Mothers Against Drunk Driving, founded in 1980, helped push the drinking age to 21 and, eventually, a national 0.08 blood-alcohol limit. Seat-belt laws, beginning in New York in 1984 and now in force in every state but one, did something quieter and larger: they drove belt use from roughly one American in seven to more than nine in ten. NHTSA credits that single behavioral shift with saving on the order of 10,000 lives a year — a device that already existed, finally being used.
The belt that was always there: U.S. seat-belt use
The three-point belt existed since 1959. Getting people to wear it took laws, campaigns, and forty years.
Source: NHTSA observed-use surveys. The gains track almost exactly with enforcement laws and "Click It or Ticket" campaigns.
The star that finally sold safety
For decades the industry's excuse was that safety didn't sell. What changed that was information. Government crash-test ratings began in 1979; the insurance-funded IIHS introduced its punishing offset test in 1995 and a brutal small-overlap test — in which only the outer edge of the front end takes the whole hit — in 2012, humbling even cars rated "good" elsewhere. Europe's program, launched in 1996, added pedestrian-protection scoring that reshaped the fronts of cars. Once a poor star rating could visibly sink a model, safety became a marketing weapon: the one-star Rover 100 was pulled from sale within months; the Isuzu Trooper, rated "Not Acceptable" for its rollover tendency, never recovered and was discontinued; a two-star result nearly buried the Daewoo Kalos. Cars are now crashed in a whole battery of standardized ways — full-frontal, moderate and small overlap, side barrier, side pole, and rollover — each exposing a different weakness, and each result printed where a buyer can read it.
The ninety-four percent
By the 2010s the car had grown so good at surviving crashes that attention turned to preventing them, because the remaining problem is overwhelmingly human. The quiet giant of that turn was electronic stability control, developed by Bosch and Mercedes in 1995 and spread partly by embarrassment — a new Mercedes famously tipped up during a Swedish "moose test" — which senses an impending skid and brakes individual wheels to keep the car pointed where the driver steers. Often called the most important advance since the seat belt, it cuts fatal single-vehicle crashes by roughly a third and was mandatory on every new U.S. light vehicle by 2012. Behind it came a whole suite of electronic guardians: traction control, brake assist, tire-pressure monitoring, adaptive cruise, lane-keeping, blind-spot monitoring, driver-drowsiness detection, and automatic emergency braking. The reason for all of it fits in a single statistic.
What causes crashes
The share involving human error — the single fact driving the entire push toward automation.
Source: NHTSA/Stanford analyses of crash causation. The figure is the founding premise of the driverless-car project.
That single number is why the frontier moved to active safety — and, ultimately, to the autonomous car, whose entire pitch is the removal of the one component every crash statistic keeps implicating. U.S. regulators have now ordered automatic emergency braking, with pedestrian detection, onto every new passenger vehicle by 2029; the European Union already requires a longer list. The car is, slowly, being handed the job of not crashing at all.
The sobering turn
The story does not end in triumph, and honesty requires the harder chapter. Even as people inside cars grew dramatically safer, the people outside them did not. U.S. pedestrian deaths have climbed since 2009, pushed higher — a New York Times analysis found — by the relentless growth of taller, heavier pickups and SUVs, whose high, blunt front ends strike a person in the torso and head rather than the legs, and whose blind spots hide small children entirely. Europe moved against this a generation ago, banning rigid "bull bars" and writing hood-height and pedestrian-impact rules into law; the United States largely has not. Roughly two-thirds of the world's road deaths are people who were never in a car at all.
Safer for whom? U.S. pedestrian deaths, indexed
Occupant deaths fell for decades; pedestrian deaths reversed course after 2009 as the vehicle fleet grew taller and heavier.
Source: NHTSA/GHSA pedestrian-fatality series. The rise coincides with the "SUV-ification" of the U.S. fleet.
There is a national dimension, too. The United States, once the world's safety leader, stopped improving in the early 2010s and now trails its peers. The European Union cut road deaths by nearly half between 2001 and 2018; Sweden and Japan pulled far ahead, their roads engineered around the premise — Sweden calls it Vision Zero — that no traffic death is acceptable. The American toll surged during the pandemic to nearly 43,000 in 2021 before receding to an estimated 36,640 in 2025. Real progress, but from a plateau the rest of the rich world had already cleared, and for reasons that are as much about the vehicles Americans choose and the miles they drive as about the engineering.
The room they weren't in
There is one more pattern in the numbers, and it points at the future as sharply as any crash-avoidance system. Look at who dies on the road, and who gets hurt. Men account for roughly 72 percent of America's traffic deaths — 29,584 of them in 2023, against 11,229 women — and among drivers in fatal crashes the ratio runs closer to three to one. Men speed more, drive impaired more, and buckle up less; a large body of research, including a much-cited 2006 study in Judgment and Decision Making, finds that women consistently perceive more risk and take less of it. And yet — as the companion to this piece lays out — the crash test dummy that defined "safe" for half a century was modeled on the average man, and women, tested as an afterthought if at all, turned out to be markedly more likely to be injured in an equivalent crash.
Sit with that a moment. The people who take the most risks behind the wheel built the car, and tested it on a proxy shaped like themselves. The people who take the fewest risks — and who paid for that oversight with higher injury rates — were largely absent from the rooms where safety was designed. It is not a claim about biology; it is a claim about who was in the room. And it opens onto a larger, stranger history: the women whose fingerprints are all over the modern car — the heater, the windshield wiper, the turn signal — and whose names the record somehow misplaced. That is a story for another day, and we will tell it soon.
Where do we go from here?
The next gains will be harder-won than the last, because the easy structural wins — the belt, the cage, the airbag, the stable chassis — have largely been banked. What remains is behavior and geometry. The behavioral frontier is being handed to the machine: automatic braking, driver-monitoring, alcohol-detection interlocks, and eventually full autonomy, all aimed squarely at that ninety-four percent. The geometric frontier is political as much as technical — whether regulators will do to hood heights and blind spots what Europe already began, and design cars as survivable for the person on the sidewalk as for the person behind the wheel. And the ethical frontier runs underneath all of it: the promise that a driverless fleet could erase most crashes, weighed against the new and unproven ways it might fail.
A century ago, the industry insisted the crash could not be beaten. The death rate has since fallen by more than ninety percent per mile — one of the great quiet victories of modern engineering, and proof that it could. The unfinished part of the sentence is who, exactly, that victory has reached, and how much further the same stubborn refusal to accept the toll can carry it.
Part One
Stand-Ins for the Living: the crash test dummy, and the dead who came before it →Part Three — coming soon
Forgotten Pioneers: the women who built the car's safety, and the record that misplaced them →And for the argument that the machine has now outrun the driver: Cars Are Safer Now, But Is Driving? — a standalone companion to all three.