The difference between “theoretically possible” and “important”

Academics, myself included, love coming up with counterintuitive arguments that change the way people see the world. The media probably loves to publish such articles even more than the academics like to find them.

Sometimes, though, these same academics/media do a big disservice by raising issues that are theoretically possible, but not at all important in reality.

A great example appears in this article in Time magazine about how maybe seat belts don’t save lives because drivers become more reckless when belted because they feel safer. This effect could reduce, or even reverse the safety benefits of seat belts.

The theory is sensible. When I am unbelted, I am at greater risk for injury, so I may drive more cautiously. In economics, this idea is attributed to my friend and colleague Sam “Seatbelt Sammy” Peltzman in the 1970s. Economists call this tendency the “Peltzman Effect.”

In practice, though, the evidence could not be clearer that seat belts are an incredibly cost effective way of saving lives. (See for instance, this study of mine and the citations therein.) Whatever small offsetting impact that more reckless driving due to seat belts may have, it is trivial compared to the benefits of wearing a seat belt. Articles like the one in Time Magazine encourage people to come to completely the wrong conclusion on this question.

If, however, I’m wrong and compensating behavior on the part of drivers really does undo or reverse the benefits of seat belts, there is an easy public policy solution: the government should mandate the installation of a razor sharp knife on every steering wheel aimed directly at the heart of the driver. Just think how carefully we would all drive then.


The same argument was made about ABS (brakes) years ago. The problem is that even if it is true, the benefits outweigh the costs. Further, one has to remember that many accidents involve a causing driver and one or more victims. So, even if true, the victims are more likely to walk away if they have seatbelts on, ABS, structural cages, and other safety devices.
As to a knife on the steering wheel, the same victim problem occurs (I know you were not serious, but the concept is the same). If you want that kind of compensating behavior, make it far more costly in other ways to be the "cause" of an accident. Time in jail or community service for example. Of course, then you get far more lies and disputes and paid off witnesses, etc.

John Hawks

I'm really surprised that you think that your article answers the points raised by Adams. He *doesn't* claim that seat belts make their wearers less safe in a crash, or that the majority of people who wear seat belts will be more likely to drive aggressively.

His claims are about the global costs of seat belt laws: (1) Drivers have a "risk thermostat" that makes some people likely to drive more aggressively when they adopt risk prevention devices such as seat belts. Those people make the road globally more dangerous, even for safe drivers. (2) Saved lives on the part of drivers and passengers are offset by costs to the system that aren't counted in the same way, such as pedestrian and cyclist fatalities.

You are quite right that razor blades on the steering wheel would improve road safety; Adams makes the same point. You seem to agree.

You disagree (apparently) about the safety statistics following the adoption of seat belt laws, but your data doesn't address the international comparisons (countries with and without laws, and effects on individual countries after adopting seat belt laws) that Adams presents in his 1995 book.

Frankly, I think this is a ripe topic for the Freakonomics treatment -- his international seat belt comparisons are similar in form to the abortion data, with the advantage that there is no 18 year delay.


Bill Conerly

Hey, there's a tradeoff between safety and speed (with "speed" being a proxy for speed, recklessness, drunkenness, etc.). When there's an improvement in the opportunities open to me, and I choose to load all of the gains into speed, leaving safety unchanged, what's wrong with that? I get to drive faster. I'm better off. So what if I didn't use the gains in the way you would like me to; it's my choice to make.

Whether I use seat belts to reduce my risk or to increase my speed, there is still a gain from wearing seatbelts.

Steven D. Levitt

John Hawks,

I probably could have made my logic clearer. My point was that seat belts have huge direct effects on safety. I can imagine they also have small indirect effects through the Adams channel, but it is totally implausible
to me that wearing seat belts would lead to nearly a doubling in serious crashes, which it is what it would take to offset the direct effect.


I've read before about how ABS brakes will cause someone to simply drive faster. Another example that actually makes the point better goes like this: When a Scandinavian country switched to driving on the opposite side of the street (sometime in the 1970s if I remember the example correctly), a national campaign for weeks beforehand urged pedestrians to look both ways. The fear was these people would be blindsided when crossing the road as they habitually look for oncoming cars on the wrong side.

What happened was the big switch came and went, and the period that followed was the safest for pedestrians than any time prior. This grace period lasted three weeks or so, if I recall. Then people stopped being vigilant and started being traffic victims again.

The point of the anecdote, which was in a story on why mishaps happen in The New Yorker, was to show that people tend to "spend" risk when they feel they have more safety -- or protection -- in some theoretical bank account. When that balance runs low, as with not buckling a seat belt or walking across the street in a more perilous traffic situation, they are more frugal with risk.

With this banking metaphor, I suspect it's almost impossible to outspend the additional protection a seat belt provides.



It has been shown that automobile drivers passing a bicyclist leave a narrower gap when the cyclist is wearing a helmet. So, as hard as you try to be safer by wearing the helmet, it may induce drivers to take risks with you, and you have no control over it!,,2-2353650,00.html


Are there areas where the opposite is true - where the Peltzman effect outweighs the direct effect on safety? I've heard people make the claim about hockey. Sometimes the claim is that helmets didn't increase overall safety, and their introduction didn't reduce injuries. Sometimes the claim is that anti-fighting rules resulted in more injuries (this isn't quite the Peltzman effect, but it's something like it). The idea, I think, is that fighting was a good mechanism for preventing cheap shots or dirty hits that aren't always caught by referees. Fighting itself, meanwhile, is relatively safe.

Jun Okumura

"I can imagine they also have small indirect effects through the Adams channel, but it is totally implausible..."

I agree, Mr. Levitt, it sounds so counterintuitive. But isn't that the gist of what a lot of people used to say said about many of the things you write in your book? I agree wth Mr. Hawk; if there's anybody better equipped to get to the bottom of this "international data", it's you.

I'm not so sure that the article has done a great disservice to safety though; it is also conterintuitive to think that there are people who will read this article, get in the car, then say to themselves, from now on, I'm not buckling up because it'll make me safer.

I do believe, however, that it does a disservice to the journalists' trade when you basically write up a summary of one man's research and publish it, without first seeking out the other side's view. It conjures for me an image of a high school student on an assignment cribbing from a single Wikipedia article.



Well...guys, this is kind of the similar to the Health Warning on a pack of fags...and now the gross pictures showing cancerous tumors etc...has it actually got people to stop ? I do not have access to emperical evidence, but I cant see any tobbaco companies giving any profit warnings !! On another note, back home in India, there are a few states which still have a a prohibition law ie liquor cannot be sold . Period. However, has that stopped anyone ..on the contrary, as a bystander I think it has only succeeded in increasing bootlegging, sale of illicit home brew, as well as fostered a higher level of corruption in the law enforcement my mind they keep up with the corruption because it is a money making machine for the regulators ! So I kind of buy the argument that there is a counter intuitive argument at play , most times ..the question is to have the ability to measure the tipping point .



I would just wear a metal chest plate and go on driving like a maniac.


Didn't Sam base his logic also on some data? I believe he did and what he found was seatbelt regulation had no effect on total number of deaths.

Also you have to remeber the pedestrians who do not wear seatbelts and suffer more from the increased riskiness of drivers.


This post reminds me of a story a contractor friend of me told me. He said that he knows electricians who will ONLY work with the power ON. This way, he says, they always know the state of the wires they're interacting with.

Sounds weird to me, but does speak to the kind of psychology discussed here.


> "seatbelt regulation had no effect on total number of deaths"

This is not actually clear, as there are no cases of identical cars/environment, just seat belts. Seat belts get phased in as new cars come in. But, newer cars also have other safety factors. So, it is very hard to get an instantaneous measure (as far as I can tell from his data, he found no country that required retrofitting all cars immediately).
Further, seat belts are supposed to decrease injury more than decrease deaths. That is, by stopping you from slamming into the steering wheel or dashboard, or being thrown within or even ejected from the vehicle, you will likely suffer far fewer injuries for most kinds of accidents. Regardless of whether you drive more safely or less, the reduction of injuries leads to a huge reduction in costs to society (insurance costs, loss of work time, those who suffer lifetime injuries, etc).
A funny example of the “Peltzman Effect" is the Nintendo Wii remotes. Because they have a safety strap (which is poorly designed), people feel comfortable using much larger force on the remotes (swinging and throwing motions). But, the straps tend to break, and so people are causing serious damage (to their TVs mostly).


scott cunningham

An interesting bit in the NYT Magazine today talked about this (sort of). A psychologist and avid cyclist installed some kind of sensor to his bike and rode several thousand miles - half the time wearing a helmet, half the time not. He found that when he wore his helmet, he was on average 3.4 inches closer to cars than when he didn't. It's not clear whether automobiles get closer to him when he's wearing the helmet, or if he unconsciously gets closer to them.


I don't really believe that some manifestation of the Peltzman effect being absorbed in 1970 is inherently inconsistent with seatbelts having a positive effect on safety today.

I would argue that the novelty of seatbelts early in their life added to the fact that anyone who believed in them enough to wear one also believed that they would make them invincible somehow. I'm relatively young however, (college freshman) and therefore have no experience with cars that didn't have seatbelts and my parents barely had any experience without seatbelts. There was no novelty to it, for as long as I've been alive, when I sit in a car there is a seatbelt for me, and it is a reflex to put it on. I don't even think about it making me safer honestly, I just feel wierd without one. So maybe once people are used to them, they stop compensating for them since they have no benchmark for comparison.


This link ([url][/url])
is to an article about about traffic engineers in Europe and the United States. From the article, "Build roads that seem dangerous, and they'll be safer." Accident rates have dropped in towns where they have removed all traffic signals and signs. Highways have become safer where they have removed the center line markings. Pedestrian traffic has increased and property values have increased where wide thoroughfares have been replaced with narrow two lane roads.

To me this all seems at odds with my understanding of the early years of the automobile. I thought traffic signs and modern road design were a response to the high number of accidents. Could it be that much of the effectiveness of a new safety device or idea is derived from its "novelty?"


That makes me remember an article on city planning I read some time ago (in Wired, I think), where they definitely went the counterintuitive route, removing the barriers between drivers and pedestrians- streets and sidewalks were delineated with color, but no curbs, etc.- everything was one similar, even surface. the result was fewer accidents, because it forced drivers to be far more cautious.


I feel like part of changing the road is that it makes things immediately more interesting, regardless of if you crash or not. There is just more that a driver needs to watch. A similar reason motivates curvier interstates. [according to the History Channel, maybe Modern Marvels] Initially, the engineers laying out interstates, being engineers just drew them as straight lines from point a to point b. However, this sort of driving will put drivers to sleep. They found that adding an occassional curve kept things exciting and thus saved lives. If this is the factor, then I'm not sure that making the car less safe would add any extra stimuli like this. Stuff like airbags and to a lesser extent seatbelts, are passive safety, there is no ongoing action by the driver to make them work. Turning the car when the road bends (or watching the pedestrians walking right next to you) requires immediate action.



I think that's a good part of it, for sure- it's so easy to zone out on a straightaway, nobody has a perfect attention span, so novelty is a lot more important than is usually accounted for.


In the 1960s the baby boomers started driving.

So the US experienced a very large number of young drivers in the 1960s--both in absolute numbers and as a share of mile driven.

But young drivers -- 16-24 years old -- have much higher accident rates.

So what Pelzman actually found was the impact of the baby boomers on accidents. But he called it risk adjustment to demonstrate his anti-regulation priors.