Authors: Tom Vanderbilt
Thinking that the key to truly understanding Roman traffic might lie in physics, one afternoon I went to visit Andrea De Martino, a physicist with the Laboratory of Complex Systems at the University of Rome. In his office at La Sapienza, he drew diagrams on the chalkboard and spoke of “network optimalization” and “resource competition.” Then he talked about Rome. “My girlfriend is not from Rome, she’s not Italian,” he said. “She tried to understand the logic behind the fact that a car can just cross the road even if it sees you coming. There is no logic.” He contrasted this to driving in Germany, which he’d found to be “marvelous.” This was not the first time I’d heard a Roman praise driving in some other, more “orderly,” country. I asked him: If everyone likes it so much, why don’t they drive that way here? He said: “I like the German system—
in Germany.
”
One could drive like a Roman in Frankfurt, or drive like a Frankfurter in Rome, only one might not do so well in either situation. But why is that? Where do these norms come from? The simplest answer may be that Romans drive the way they do because
other Romans do.
This idea was expressed in a series of experiments by the psychologist Robert Cialdini. In one study, handbills were placed on the windshields of cars in a parking garage; the garage was sometimes clean and sometimes filled with litter. In various trials, a nearby “confederate” either littered or simply walked through the garage. They did this when the garage was filled with litter and when it was clean. The researchers found that the subjects, upon arriving at their cars, were less likely to litter when the garage was clean. They also found that subjects were more likely to litter when they observed someone else littering, but
only
if the garage was already dirty.
What was going on? Cialdini argues there are two different norms at work: an “injunctive norm,” or the idea of what people should do (the “ought” norm), and a “descriptive norm,” or what people actually do (the “is” norm). While injunctive norms can have an impact, it was the descriptive norm that was clearly guiding behavior here: People littered if it seemed like most other people did. If only one person was seen littering in a clean garage, people were
less
likely to litter—perhaps because the other’s act was so clearly violating the injunctive norm. This is why so many public-service advertising campaigns fall on deaf ears, Cialdini and others have suggested. An advertisement about the many billions of dollars lost to tax cheating draws attention to the problem, but it also whispers: Look how many other people are doing it (and getting away with it).
Who
is violating a norm is also important: Studies of pedestrians have found that walkers are more likely to cross against the light when a “high-status” (i.e., well-dressed) person first does so; they’re less likely to cross when that same person doesn’t. “Low-status” violators prompt less imitative behavior either way.
Traffic is filled with injunctive norms, telling drivers what to do and what not to do. But the descriptive norm is often saying something else—and saying it louder. The most common example is the speed limit. The law on many U.S. highways is 65 miles per hour, but a norm has gradually emerged that says anything up to 10 miles per hour above that is legal fair game. Raise the speed limit, and the norm tends to shift; driving the speed limit starts to seem hazardous.
Some norms seem to hold more strongly than others. Leonard Evans, a trained physicist and traffic-safety researcher who worked for General Motors for more than thirty years, gives an example: “It’s two a.m., some guy’s just been speeding, to save time. He comes to this intersection. There’s no traffic in sight anywhere. He sits stationary for thirty seconds. Objectively speaking, he is causing far more risk by his exceeding the speed limit than he would be if he stopped at the red light, looked this way and that way, and just went through it. But we have a robust social norm in the U.S. You just do not consciously and casually drive through a completely red light. Unfortunately, we don’t have a robust norm against not going fast after it’s turned green.” Both acts are technically against the law, each bear similar penalties, but one act seems more illegal than the other. Perhaps in speeding the driver feels as if he’s in control, while going through a red light, even carefully, puts one at the mercy of others. He may also speed because most other people do (whereas if everyone decided to cross through red lights, anarchy would ensue).
Most traffic laws around the world are remarkably similar. Many places have relatively similar roads and traffic markings. But the norms of each place are subtly different, and norms are powerful, curious things. Laws do not dictate how people should queue up in the United Kingdom or China—nor should they, most would argue—but try queuing up in either place and you will notice a striking difference. In the United Kingdom, queues are famously orderly, but in China, they often exist more in theory than reality—queue jumping, along with jaywalking, was another behavior targeted by the Chinese government for extinction before the 2008 Olympics.
Similarly, economists have long been puzzled by the fact that, in most places, restaurant patrons tip their server
after
they have already been served—which may boost the incentive for the server to give good service but hardly increases the incentive for the patron to tip well. Mysterious, too, is that patrons tip even in the face of further erosion of these incentives—if their service was less than desirable or if they don’t plan to return to the same restaurant. Studies have shown the link between tip and service quality to be slight. People seem to tip because it’s seen as the right thing to do, or because they don’t want it known that they’ve not done the right thing. There’s no law that says that patrons have to tip; they simply follow the norm.
In traffic, norms represent some kind of subtle dance with the law. Either the norms and laws move in time or one partner is out of step. In Florence, observes the writer Beppe Severigni, the locals have a phrase,
rosso pieno,
or “full red,” for a traffic signal. This implies that there are other reds that are less “full.” These distinctions are not noted by law, but they help explain actual behavior. Yet where do these norms come from? How do they adhere to or depart from the law? It seems that the most significant norm of all, as the legal scholar Amir Licht has noted, is the “deeper, more general norm of obeying the law.” When you step off a curb because you have the “Walk” light or drive through a green light expecting not to be hit by another driver, it is not the law per se that protects you but other drivers’ willingness to follow the law. Laws explain what we ought to do; norms explain what we actually do. In that gap dwells a key to understanding why traffic behaves the way it does in different places.
In 1951, some 852 people were killed on the roads in China. In the United States in that year, 35,309 people were killed in traffic. In 1999, traffic fatalities in China had risen to nearly 84,000. The U.S. figure, meanwhile, was 41,508. The population of both countries had almost doubled in that time. Why did fatalities rise so much higher in China than in the United States?
The answer lies in the number of vehicles in each country. In 1951, there were about 60,000 motor vehicles in China, while in the United States, there were roughly 49 million. By 1999, when China had 50 million vehicles, the United States had over 200 million—four times as many. And yet
twice
as many people were killed on Chinese roads than American ones. How could the country with so many fewer vehicles have so many more deaths?
This strange equation has become known as Smeed’s law, after a 1949 paper, humbly titled “Some Statistical Aspects of Road Safety Research,” by the British statistician and road-safety expert R. J. Smeed. What Smeed’s law showed was that, across a number of countries, ranging from the United States to New Zealand, the number of people killed on the roads tended to rise as the number of cars on the road began to rise—
up to a point
—and then, gradually if not totally uniformly, the fatality rates began to drop, as, generally, did the absolute numbers of fatalities.
Smeed suspected that two things were going on: One, as the number of deaths grows higher, so too do people begin to clamor for something to be done about it (as began to happen in the United States in the 1960s, when fatalities were topping 50,000 people a year). Second, Smeed proposed that a sort of national learning curve was at work. The more cars on the road, the more people are “growing up” and learning how to sort out the problems of traffic—with better highway engineering, stronger laws, safer vehicles, and a more developed traffic culture itself (and perhaps more congestion, which tends to lower traffic fatalities).
In China, one sees things that make the hair stand on end—like bicycles traveling on restricted highways, scooter drivers carrying several children without helmets, and drivers stopping on the highway to urinate—but presumably, a number of years down the road, these things will largely be only memories. The dynamics of Smeed’s law may help explain a curious phenomenon noted by Rong Jiang, a Beijing Institute of Technology transportation engineer. Studies had suggested that the crash rate was actually higher on the high-speed, divided “luxury roads” of the new China, he said, than on the two-lane rural highways. This is exactly the opposite of what happens almost everywhere else. He suspected that drivers were not adequately trained for the new high-speed roads. “The drivers were used to low speed on the open road,” he explained. “But if they travel along the freeway, they keep the same habits. If their vehicle has a malfunction they will just park on the shoulder, without any alerting equipment. There are many such collisions.”
Smeed’s law, if history serves as a guide, is why one cannot simply look at the current horrific numbers of road deaths in countries like China and India and the relatively low levels of car ownership and assume that fatalities will continue to rise proportionally as more people get more cars. It may seem hard to imagine, but there is already progress of sorts even in China’s massive death toll: While more people are dying on China’s roads than ever before, the Chinese fatality
rate,
as measured in number of deaths per thousand registered vehicles, has actually been dropping.
Smeed’s law is complicated, however, by a few factors that make China and India different from the countries Smeed considered. The first is that most people dying in traffic in the developing world are dying not in cars but outside cars. More than half of the people killed on the road in the United States are drivers or passengers, whereas in a country such as Kenya the figure can be as low as 10 percent. In Delhi, the occupants of cars represent only 5 percent of fatalities, while pedestrians, cyclists, and motorcyclists make up a staggering 80 percent. In places like the United States and England, motorization was an evolutionary process. However novel they may have been, the first automobiles, the “horseless carriages,” could still be understood in terms of what had come before. The speeds were slow, the number of cars few.
China and India, by contrast, are seeing a vast flood of modern cars surging onto what are, in some cases, premodern roads. The Lexus and the rickshaw are thrust onto the same thoroughfare. Another consequence of this dizzyingly fast motorization is that people of all ages who have never before driven in their lives are being put on the road at once. In 2004 it was estimated that nearly one out of every seven drivers on the road in Beijing was a novice. The rapidly evolving Chinese insurance industry was dealing with customers who were reporting as many as
thirty claims
in a multiyear period. Some insurers reported accident risk for certain classes of individuals at nearly 100 percent—virtually moving them from the category of “accident risk” to the paradoxical “accident certainty.”
In the harsh language of economics, the massive traffic fatalities and unsafe road systems in developing countries might be seen as temporarily necessary “negative externalities.” In other words, like pollution or poor working conditions, they are just another price those countries have to pay in order to “catch up.” Indeed, one might read the frenetic traffic behavior as somehow expressing the soul of noisy, dirty, clamoring entrepreneurial and industrial cities. Calm and safe traffic, the argument might go, is fine for those who can afford it (e.g., Switzerland). Let us get the cars and motorcycles on the road first, let us get people commuting to jobs, and then we can worry about safety. This is why, even as the rates for things like diseases begin to drop as countries get wealthier, traffic fatalities—a “disease of development”—rise until that point, as formulated by Smeed’s law, where they begin to drop. When East Germany was reunited with West Germany in 1990, the traffic fatality rate in the former Communist country quadrupled: More people bought cars, drove them more often, and at higher speeds (the East German speed limit of 100 kilometers per hour on autobahns was raised to West Germany’s 130). While the fatality rate is still higher in the eastern half of the country, it began to drop again after 1991.
It is eerily striking how closely fatalities can be tracked in economic terms. A country’s motorization rate is linked in a somewhat linear fashion to its gross domestic product: the more money, the more cars. Researchers use the rough benchmark of a $5,000 per capita GDP as the point at which car ownership rates begin to accelerate. As work by the World Bank economists Elizabeth Kopits and Maureen Cropper shows, countries with very low GDPs have low numbers of fatalities per population (there are simply not that many cars, even if the rates per
vehicle
might be high). As the GDP grows, there is a sharp upward curve in fatalities. The rate per vehicle begins to drop with minor increments in the GDP—for example, when per capita GDP climbs from $1,200 to $4,400, the fatality risk per vehicle drops by a factor of three. After studying the data from eighty-eight countries from 1963 to 1999, Kopits and Cropper concluded that the fatalities per person begin to drop only when a country’s GDP hits $8,600 (in 1985 dollars); they eventually hit levels lower than those of countries with much smaller per capita GDPs. Projecting these numbers outward, Kopits and Cropper concluded that India, for example, where the GDP (using that same 1985 standard) was $2,900 in 2000, will not see its road death rate decline until 2042.