New book on the psychology and engineering of traffic

Two summers ago in Paris, I was astounded at the volume of traffic that somehow managed to negotiate the traffic circle at the Arc de Triomphe without incident. Here's the (poor quality) video I made to document traffic flow there:

I learned to drive in my 20s in New York City. Like Paris, New York has a traffic rhythm all its own, where lane markings are mere suggestions. In New York, parking is tolerated nearly anywhere, as long as traffic isn't unduly impeded. I've seen people double-parked, triple-parked, parked on corners, on sidewalks, you name it.

Driving into Manhattan several times a week also helped me developed what I liked to call "highway justice" -- the sense that if someone did something "unjust" as a driver, they would deserve whatever punishment the others meted out to them in traffic.

The biggest injustices in New York always seemed to revolve around highway merges. If traffic was merging onto a busy highway, justice dictated that they should be allowed in every-other-car. If mergers tried to double up, or if highway drivers didn't allow cars in, they were violators who deserved to be punished (although I was smart enough not to attempt to mete out this justice myself).

But it was completely acceptable to wait until the last second to merge (if, for example, a lane was closing ahead due to construction). Highway space was at a premium, so why not use it all? Once I moved to North Carolina, however, I soon learned there was a very different sense of justice. Late merging and horn-honking were considered rude, and the worst injustice was driving too slowly in the left lane. So which justice system is better -- North Carolina's, or New York's (or Paris')? Surely science has something to say about all this.

Of course it does, and Tom Vanderbilt does a fantastic job negotiating through the vast array of studies about traffic and driver behavior in his book Traffic: Why We Drive the Way We Do.

The problem of early- versus late-merging is one of the first examples Vanderbilt uses in the book. As with nearly every problem relating to traffic, it turns out the answer is extremely complicated, and much depends on the context of the merge. Is the merge so disruptive that traffic is slowed considerably? Then late-merging is probably preferable, since this better-utilizes the capacity of the highway. Why back people up for miles when an extra lane is available? But if the highway is relatively clear and people are whizzing along at 70 miles per hour, waiting until the last minute can be downright dangerous. Perhaps the best solution is to install digital signs that can change instructions to mergers based on traffic levels. But even this is problematic -- once regular drivers accustom themselves to one style of merging, it's difficult to get them to change.

This, perhaps, is the central lesson of the book. Traffic and highway management isn't like other engineering problems, because driver behavior adapts quickly to new situations, which can create completely different traffic problems. Consider a narrow road with a speed limit of 30 miles per hour, lined on both sides with imposing oak trees. Each year there are several crashes involving vehicles hitting the trees or cars emerging from driveways between the trees, so engineers decide to cut down the trees and widen the road, creating ample shoulders and improved sightlines. The expected result is a safer road. What actually happens is that drivers increase their speed on the wide (if less scenic) road. There are just as many crashes as before, but because cars are traveling faster, there are more fatalities. The "safer" road ends up being more dangerous.

There are similar issues when engineers decide to add lanes to roads. You might think that adding an extra lane to a busy two-lane road would reduce traffic, but in many cases, the opposite occurs. More people start to use the road because it's available, and capacity doesn't actually increase as much as you'd think. Adding an extra lane in each direction to a two-lane road only increases capacity at intersections from 625 to 1100 vehicles per hour, not the 1250 you'd expect. This is because it takes longer to cross a wider intersection. Adding a third lane adds only an extra 385 vehicles per hour.

Another counter-intuitive rule-of-thumb for vehicles is named after my college Econ 101 professor, Sam Peltzman. The Peltzman Effect says that safety improvements for vehicles such as seatbelts and airbags will never reduce the impact of traffic accidents. As people feel safer in their vehicles, they increase risky driving behaviors: talking on the phone, speeding, or aggressive passing and merging. Early research suggested that the introduction of a third brake light in the center of cars would reduce rear-end collisions by 50 percent. In fact, as people adapted to the new technology, their net benefit was found to be 4.3 percent. As the technology became more widespread, people seemed to reason that if it's easier to see when someone is stopping, why not follow a little closer?

So how is it that people manage to navigate the unbelievably crazy traffic circle at the Arc de Triomphe? Vanderbilt says it's precisely because people know it's so crazy. They're more alert, and they drive slower and more carefully than they would in other "safer" situations. Sometimes it's precisely these "dangerous" places that are actually some of the safest places to drive, because people slow down and pay attention.

There's lots of other great stuff in Vanderbilt's book -- it'd make an excellent late Christmas gift. Just make sure the person you buy it for isn't the sensitive type -- they might take it as a subtle hint that you don't think they're a very good driver.

Tom Vanderbilt also has an excellent blog with frequent updates and discussions of the psychology and engineering behind traffic and the way we drive.


More like this

People act reasonably. Purpose of cars is to get from one point to another as fast as possible. Safety is not most important. This is why improving security causes faster driving.
Article suggests, that acceptable level of risk is constant. Are there any studies to demonstrate this?

I bought this book a few months ago and it's been sitting on my shelf waiting to be read. You've inspired me to move it to the next spot on my list.

Traffic has always been something of interest to me

Having driven in Caracas, I know no fear. My wife would refuse to ride with me for at least a couple of weeks after I would come back from Venezuela.

By Jim Thomerson (not verified) on 23 Dec 2008 #permalink

"Vanderbilt says it's precisely because people know it's so crazy. They're more alert, and they drive slower and more carefully than they would in other "safer" situations. "

Speaking from experience I totally agree. I've spent some time in Iran lately. Traffic in Tehran makes what we see on that Paris video look like a walk in the park. At first I was in shock. I didn't know how to cross the street. Forget the lights.
But after some time I realized something: they actually drive according to the rules, or, better said, one rule, which is "the one in the front is the one that is right". That means no driver would ever stop to let me cross the street, but as soon as I stepped in front of them I was fine.
Of course, my sensation-seeking is probably above average, and I also have a specific understanding of "safe" :)

And here's another video to support the Vanderbilt hypothesis. Not only is this intersection busy, but the number of vehicle types (car, pedestrian, motorbike, trucks, etc.) passing through is large. And yet, everyone manages to make it.

Of course, I just checked on the fatality rate, and it appears India kills over 100,000/year in traffic accidents. I don't have a figure for fatalities/mile to compare with other nations.

By briang467 (not verified) on 23 Dec 2008 #permalink

I get what you're saying about the most dangerous places being the safest. I live in Vietnam and traffic is absolutely insane. There is only one rule: don't hit anyone or anything. Right away is assigned on an ad hoc basis; it's mostly determined by speed, honk volume/duration, size of vehicle, and how crazy you look.

I am not sure what you would say to this, and of course Vietnam doesn't have any reliable traffic statistics, but annotatly it seems that traffic is actually slower here, and more dangerous.

Has anyone done a study comparing traffic in areas with a heavy reliance on fixed rules (US) to those without a reliance on fixed rules(Vietnam)?

On the Peltzmann effect, I've never seen why it should completely remove the safety effect - travellers want speed AND safety. Give them more safety, and they'll trade some of it away for extra speed, but there's no reason they'll trade away all of it.

As a fellow Saigon resident, I've noticed that the chaos on Vietnam's roads is organised chaos. Traffic moves slower and feels much safer as a result, because you have more time to react to the many events happening around you. And of course the collisions are less traumatic too.

While right of way usually goes to the bigger/louder vehicle, everyone takes a collective responsibility to avoid collisions. It helps that the traffic is mainly motorbikes, which are small and highly manoeverable, and that you can make eye contact with other riders on the road.

And although traffic is slow, the number of vehicles that move is incredibly high. I would suspect that the capacity utilisation of Saigon's roads is amazingly high, and that on a passenger mile basis the fatality rate wouldn't be that high.

By the way, there is an interesting case study of the Peltzmann Effect in Saigon. Very rough data suggests that the law mandating helmets for motorcycle riders (except kids!!) reduced fatalities by 16% in its first few months. Presumably, motorists went faster, but not by enough to remove all the safety gains that wearing helmets give.

I second Jans comments re Tehran.
At first glance it was mayhem, but there are unwritten rules. I asked a driver about the traffic light colours.
"Green means go into the intersection, but check if someone is coming from the red side.
Red means check if someone is coming from the green side, then go into the intersection" (!)
It also takes a fair bit of nerve to step out there into that traffic, but it does seem to work.
And the road toll is about equivalent to what it was in Australia in the 70s (as far as I could work out). We have improved since then, but we found that level of death and injury acceptable 30 years ago.

The Arc de Triomphe roundabout is a challenge even for french people. The only rule which applies is right-hand priority.
That said, it is one of the least structured intersection in Paris (there is another similar one Place Dauphine) and is not typical.

At the opposite end of the spectrum - Saskatchewan Canada. It has a low population and good highways - straight flat paved roads without obstruction of vision and very low traffic. You would think that there would be no accidents at all. But they have a high number of 'single car accidents'. People also run straight into railway trains that can be seen coming for miles. If traffic is safe enough - people will pay so little attention that they literally go to sleep.

Another example of the "Arc de Triomphe" effect or situation was what happened during the Feria of Seville. Until a few years ago, there was *no* driving under the influence checking during the Feria. This despite the huge amount of alcohol consumed during the Feria by almost everyone. As a matter of fact it was said that the police did not do it because there would not be enough place in police stations and jails to hold people convicted for DUI. And there was no relative increase of death and accidents during the Feria. Of course everybody knew that everybody was drunk and driving so everybody was very, very cautious. Even youngsters with tuned cars were riding carefully.

Now the police has implemented lots of DUI checks during the Feria and the only effect is that public transport are much crowdier and taxis are much harder to catch. But no effect on accidents.

By Pedro Godfroid (not verified) on 29 Dec 2008 #permalink

The Peltzmann effect could go the other way around. Very good examples are custom bikes.

At first sight they seem to be completely absurd. Customizing a bike is basically transforming it to make it much more difficult to drive, handle, brake, speed, etc. Good progressive suspensions are substituted with "low profile" hard as hell shocks that bounce the bike around. Or simply removed for a true hard tail (nor rear suspension). Front brakes, the most effective, are often removed. Seating position is made extremely unnatural and uncomfortable. Moreover the weight is shifted towards the rear axle rendering the bike extremely hard to handle and with very bad stability in turns. Often the easy to use stock shift and clutch is substituted by a automotive stick shift extremely hard to use on a bike. Good telescopic forks are changed for either "springer" nostalgic thingies with no rigidity or handling whatsoever or/and with absurd lean angles that further impede normal handling. And so on. Everything that makes a bike "custom" seems to have been designed to make the bike harder to handle, drive and simply stay on the road.

Why? Because speed limits make driving a bike uninteresting. It's telling that the custom thing began in the States who had speed limits very early and only moved to Europe when speed limits were implemented here. And got stronger as speed control got serious.

By Pedro Godfroid (not verified) on 29 Dec 2008 #permalink

NC drivers, yes, but the Raleigh area is sort of a bubble in itself. Maybe it's the relocated Northerners?

Take a trip to Glenwood Ave. in north Raleigh and people routinely go 5+ below the posted limit. It drives me crazy.

I've seen a study showing that bicyclists in the city are safer without helmets than with helmets, because without the helmets cars give them a wider berth, since they look so vulnerable.


I've seen that study as well -- it's unclear whether you can draw the conclusion that cyclists without helmets are safer. Yes, cars give them a wider berth, but in the case of an accident, helmets are safer. The study doesn't show how many accidents are prevented, and whether that completely compensates for the helmet effect.