> 0 The Most Dangerous Machine (Ep. 147) - Freakonomics Freakonomics

The Most Dangerous Machine (Ep. 147)

Listen now:
(Photo: Simon Davison)

(Photo: Simon Davison)

The latest Freakonomics Radio podcast is all about our long, risky, and mostly unrequited love affair … with the automobile. (You can subscribe to the podcast at iTunes, get the RSS feed, or listen via the media player above. You can also read the transcript; it includes credits for the music you’ll hear in the episode.)

Every year, more than a million people die worldwide from traffic accidents. Last year, there were roughly 33,000 such deaths in the U.S. And while it’s widely believed that modern distractions — our mobile devices, our lattes, our podcasts — make driving ever more dangerous, here’s the surprising thing: in most places, the death rate has fallen, by a lot. Ian Savage, a Northwestern economist who studies transportation safety, tells us that this number has declined by two-thirds since 1975. And self-identified terrible driver Steve Levitt says that the data show “an amazing success story.”

A sign in Ohio City marking the site of the John Lambert crash. (Photo: Karen Bennet Miller)

A sign in Ohio City, Ohio, marking the site of the John W. Lambert crash. (Photo: Karen Bennet Miller)

You’ll hear about the many factors that have contributed to this success: seat belts, tougher drunk driving laws, demographic changes. You’ll also hear about some of the earliest safety inventions, some of which were much more successful than others. Museum curator Alison Federik tells us about a pioneering doctor who created lanes in her hometown of Indio, Calif.; and automotive writer Brett Berk fills us in on a short-lived invention from auto entrepreneur Preston Tucker.

So the good news is that we’ve gotten a lot safer behind the wheel. The bad news is we’re our own worst enemy. More than 90 percent of crashes involve some kind of human error. Greg Fitch, of the Virginia Tech Transportation Institute, ran naturalistic driving studies (including the impact of hand-held and hands-free) in which video cameras were put in cars to see what people were actually up to while driving:

FITCH: We have evidence of people talking on two phones at once, while smoking a cigarette while driving through a work zone and running a red, yellow light. We’ve seen videos of drivers relieve themselves behind the wheel. For over the road drivers we’ve seen examples of them actually meeting with a prostitute behind the wheel.

Freakonomics Radio producer Gretta Cohn.

Freakonomics Radio producer Gretta Cohn with the Jaguar F-TYPE at Monticello Motor Club.

And no show about driving would be complete without going out for a spin. We’ll take you along on two rides – one high-tech, the other high-speed.

Autonomous vehicles may well become a reality in the coming years; you’ll hear host Stephen Dubner take a ride in a driverless car developed by the engineering wizards at Carnegie Mellon University. (You’ll also hear insurance veteran Fred Cripe explain how liability might be sorted out in a driverless-car future.) And the entrepreneur and car enthusiast Ari Straus takes us on a 140-mph  Jaguar joyride at the Monticello Motor Club, a country club built for cars. And yes, it’s a sweet ride.

Tal F

What about automobiles causing injury and death to those outside the motor vehicle? Seatbelts, air bags, and crumple zones are all great for the driver and passengers inside, and it is not surprising that the fatality rate has gone down so far so fast.

But are drivers overcompensating for the perceived safety by driving faster and more irresponsibly? Similarly to what Ed Glaeser explained in your podcast on bad environmentalism, just as making driving cheaper has resulted in far more driving, so, too, making driving safer has resulted in at least a partial blowback effect so that overall traffic fatalities are not down nearly as much as the fatality rate. Both the number and the rate are not down as much as if driving behavior itself had not changed.

Has the negative externality of cars on the surrounding environment gotten much worse over time (in terms of death and injury, not pollution)? The figures I've seen suggest that, even as traffic fatalities overall are decreasing, the share of pedestrians and cyclists of traffic fatalities is increasing.

I believe this also relates to your previous podcast on rinkonomics. I explain it a bit more fully on my blog, but in essence the mutuality has been removed from traffic collisions. Motorists have little to no incentive to watch out for vulnerable road users, as the likelihood of damage to themselves or even to their car is extremely low.


W. P. Kelpfroth

@Tal F:
I realize that the tech innovations in cars makes it less likely I'll be hurt through inattention, so to keep myself sharp for driving, I alternate between my car and my motorcycle.

Tal F

My logic would certainly suggest that motorcyclists would be much more careful, and the average motorist probably would be more careful in a motorcycle. Potentially outweighing this is the fact that many (certainly not all) motorcyclists, particularly in the US, are thrill-seekers and/or gang members, not simply commuters looking to get from point A to point B. Nevertheless, your point is well taken, but I think it would be far better to reintroduce mutuality via the justice system, with stiffer penalties for reckless driving, than to promote potentially unsafe yet more vulnerable modes of transport such as motorcycling.

W. P. Kelpfroth

You're a constant channel switcher? I listen to your podcasts, and all the way through, too.

Julien Couvreur

Consider two broad categories of improvement: the vehicle and the road.

The road is not simple, it has many different qualities: it's structure (how many lanes, how wide, how many intersections, what types of intersections), it's rules (minimum/maximum speed per lane), it's enforcement (how many cops, how many cameras, methods of social enforcement), and it's costs (payment structure, wait time in congestion, etc).
The only reason why we think the road is simple and uniform is because governments tend to provide vanilla grey goods. In the realm of food, it would be porridge and potatoes.

We have quite a bit of competition in the former (vehicle), and we get innovation and diversity as a result. Imagine if roads were not subjected to government monopoly. How would road owners respond to competitive pressure? How much experimentation and innovation could we get there?

Tal F

You don't have to imagine. In the early days roads were private, and they were either a horrible mess or extremely expensive and underutilized. Unlike cars, roads are somewhat of a public good. Until fairly recently (e.g., zipcar), my occasional use of a car prevented you from ever using that car, whereas my occasional use of a road only barely impinges on your ability to use the road, too. Economists widely acknowledge that a private market will provide too little of a public good, hence the need for a government to step in and provide it.

Having said that, I do believe we have gone too far and now the government is providing too many roads, not too few. Nevertheless, going to the opposite extreme of zero government provision cannot be the answer.

Michael Bischoff

While the podcast was fascinating and well-researched, as others have pointed out you focused almost entirely upon what happens inside the car and the driver of the car. As an avid cyclist who rides for both transportation and sport, I share the road with these "dangerous machines" that outweigh me by at least a factor of 10 and are capable of inflicting serious injury or death upon me.

I do my best to stick to bike lanes, ride in a predictable manner, and signal my intentions, but all of it is for naught if a driver who is texting or otherwise distracted (or just simply doesn't care to check their blind spots) drifts or turns into my direction of travel. I really don't care how safe automobiles are for the drivers, because while all of those safety features result in fewer motorist fatalities, it means jack to the cyclists or pedestrians who are struck and almost certainly seriously injured or killed as a result.

So, I'd ask that a follow-up podcast be done that focuses on the injury/fatality rates of vulnerable road users.


Cassandra hartblay

I can't be sure, becaue I haven't looked in to this, but as a Medical anthropologist, I suspect you're missing something in your analysis of why automobile fatalities have fallen since the 1970s. You hinted at it, but didn't quite hit on it. And that's that the contemporary medical technology that we now generally call "life support" - respirators and so forth that keep people alive after major traumas - first came into widespread use around the 1970s. So that what your data may be reflecting is not an actual decrease in the severity of car accidents (or that alone) , so much as an increased capacity to survive - although potentially with severe injuries and subsequent disabilities, accidents that previously would have been fatal. Those people who survived such accidents wouldn't necessarily show up in the fatality data -- at least not if it's formatted in the way that I understood it from the podcast. So, where are these people? And what do they think about car safety?



Freakonomics seems to be operating on the premise that only the people IN the cars matter. You laud the grand advancements in safety for car drivers/passengers and even entertain the notion that those advancements may outweigh the risks of distracted driving. But you never even so much as mention the fact that none of the advancements to date do anything for the safety of pedestrians and cyclists. There is a brief mention of driverless cars being able to avoid them, but it's only in passing. The incredible threat cars pose to pedestrians and cyclists, children in particular, should have been a major part of this piece but was completely ignored.

You even make light of the great danger Steve Levitt's grossly negligent behavior while driving poses to such people, especially in a pedestrian-heavy city like Chicago. He might as well have admitted to driving drunk.

Further, you acknowledge that the most dangerous machine we interact with is the car and that it does in fact kill more people than the machines intended for killing: guns. But then you go on to paradoxically praise how safe the car has become? Not your best work, Freakonomics.



While they're not the biggest threat to health and safety, it's ironic how little most people worry much about them, in contrast to what they do worry about.

Ben Von Handorf

Amusingly, while catching up on my Freakonomics Radio episodes and listening to this exact episode, I was struck from behind at a stoplight by a catering van where the driver was trying to catch falling food.

Luckily, no real damage and no apparent injury yet... all thanks to technology. If I'd had an old metal bumper on shock absorbers, I strongly suspect I'd be in the hospital this AM and have a very damaged car.

Dan Tani

Steven Dubner - I have a bone to pick with you! At the beginning of this podcast, you ask Levitt to rank his driving skills "on a scale from 1 to 10". As someone who knows the value of accurate data and the inherent flaws in human's subjective ratings, you should know that a 1-to-10 scale is pretty much useless. Too many choices! A MUCH better rating scale would be 1-4 (1 - Terrible; 2 - Bad; 3 - Good; 4 - Great). Too many options leaves too much "play" in the evaluation, and allow the data to stack up in the middle and edges where is it hard to interpret.

I think the subject of rating scales would make a great Freakonomics podcast topic. A good example (good in the bad sense) is the ratings that are given to products on websites. It seems to me that people that leave reviews either love or hate the product, and so the list of reviews are either 1 or 5 stars (never use an odd number of choices - it leave the middle choice open - worthless data in the signal detection sense). So inevitably when the ratings are averaged out, the resultant score is 3-4 stars. On the good side, the ratings are also displayed on a histogram -which allows the user to see the distribution - which is almost always more useful than the average.

Love your podcast and I learn something every time I listen. Keep up the great work.



I'm all for driverless cars especially with the potential benefits to traffic and commute.

As a software engineer, I worry about who will be responsible if there's an accident. During the podcast, you speculate that fault would be based on a user's GPS that malfunctioned. It's one thing if I as the driver turn the wheel or accelerate or brake and cause an accident, but it's another when it's something that's software based. With a car doing millions of calculations, what's to say that a sensor or a calculation that actually caused the accident. Then who's the responsible? The car manufacturer? The company that coded the software? The sensor maker? I don't see any of these companies willing to be held responsible.

Take the case that Toyota is going through right now about some cases of sudden acceleration, http://online.wsj.com/news/articles/SB10001424052702304403804579264261779925166 . The software has been evaluated and yet no one has found any bugs, but yet Toyota is still being found guilty because they can't prove that it couldn't happen.

I am very intrigued to see how car manufacturers handle this as we move closer to this reality.



Did the number of deaths calculated take into account tailpipe pollution from cars? I know that this is an issue worthy of consideration in London, where diesel cars produce most of the nasty pollution and the city is often beyond the EU recommended limits of pollutants. I daresay the number of deaths attributable to emissions is far lower than caused by collisions, but it's worth mentioning nonetheless.