The unformatted htm text below is from

Copyright © 2004 by Leonard Evans

Information about

How to purchase

Table of Contents

Text

Reviews

The text below is © Leonard Evans.  Portions totaling not more than 400 words my be freely used with appropriate citation.
For more extensive use, permission in writing must be first obtained by emailing the author 

13 Measures to improve traffic safety

This  html version contains only the text (no figures, tables equations, or summary and conclusions) To check printed book appearance see pdf version of Chapter 1 or pdf version of Chapter 16.

Introduction
For over a century measures have been introduced in many countries aimed at reducing harm from traffic crashes. There is extensive world experience, many failures, and many successes. Some of these measures (or interventions, or countermeasures) have been discussed in detail in earlier chapters. In this chapter we address the relative contributions of different countermeasures.
Analogy with health
Traffic fatalities have sometimes been discussed as being comparable to some single disease. I consider such an analogy unhelpful because it tends to suggest that the problem might be solved by the type of elegant knockout blow that conquered smallpox or scurvy. Any such hope tends to divert attention and resources from potentially effective realistic approaches to unrealistic approaches. A more appropriate and fruitful analogy is to health in general, with traffic safety and public health having the same broad goal of reducing death and morbidity. One of the simplest measures of overall health in a nation is average longevity, which has been increasing in nearly all industrialized countries. Longevity is strongly related to national wealth, measured by variables such as gross domestic product per capita. However, people in richer countries live longer for reasons that are more complex than the mere ability to purchase better medical care, important though this is. A country's traffic fatality rate (deaths per vehicle) is also strongly related to its wealth. However, some countries with similar wealth have important differences in longevity, and in traffic fatality rates. Those with the greatest longevity tend also to have the safest traffic.
It is universally accepted that many factors have made major contributions to increasing longevity. Some are technological in nature - surgery, antibiotics, vaccines, organ transplants, and the like. Some involve improved physical and institutional infrastructure such as better housing, sewage, ambulance service, and refrigeration. Some are legislative, such as laws regulating food and air quality, building inspection, and worker safety. Some come from changes in collective human behavior in hygiene, diet, exercise, sexual behavior, and use of alcohol and tobacco. Traffic fatality rate declines reflect contributions from these same broad categories - technology, infrastructure, legislation, and behavior change.
Traffic fatality rates vary widely in time and between countries
Large changes in longevity in time and between countries have close parallels in traffic fatalities. Data in Fig. 3-5 (p. 40) showed that the number of traffic fatalities for the same distance of travel in the US declined by 94% from 1921 to 2002, and the number of traffic fatalities per thousand vehicles declined by 96% from 1900 to 2002. In both cases the comparison is between the earliest and latest data, and the declines were reasonably constant at about 3% per year over each entire period. The number of traffic fatalities per thousand vehicles in some countries is 99% lower than in others during the same year. Understanding the origins of such large variations in time, and between countries, would be a step towards identifying the factors that contribute most to traffic safety. I believe the degree of complexity inherently precludes quantitative analytical models that would effectively explain the factors that produce changes in traffic fatality rates or longevity. Models based on such techniques as multivariate analyses are often inadequate at addressing vastly simpler problems (p. 116). As progress using quantitative analytical methods seems unlikely, we instead combine and synthesize information developed in earlier chapters.
Factors influencing traffic safety
The large number of factors relevant to traffic safety can be conveniently placed into the broad categories shown in Fig. 13-1. Weather is not part of the main structure because there is not much that can be done to change it, and, unlike the other factors, it remains relatively constant over the decades. Improved medicine plays an important role in reducing harm from traffic crashes in all categories, but its more detailed role is outside the scope of this book. If a patient arrives alive at a modern trauma center, survival chances are high. One of the two occupants killed in the 1899 crash described on pages 36-37 would almost certainly have survived similar injuries today. However, FARS 2002 data show that 55% of those killed in traffic crashes died within an hour of their crashes. Recall that a typical fatality involves a single-vehicle crash at 2:00 am in a rural area, so that elapsed time from crash to arrival at a hospital is quite different than for daytime urban crashes.
The dominant role of driver behavior
The variation by a factor of more than a hundred in traffic fatalities per thousand vehicles between different countries (Table 3-1, Fig. 3-8, p. 43-44) cannot
be due primarily to differences in vehicle engineering. While the mix of vehicles is certainly different in different countries, the vehicles in the high-rate (high values of fatalities per thousand vehicles) countries are essentially the same as in the low-rate countries, because the high-rate countries rarely manufacture vehicles, but instead import them from low-rate countries. In the high-rate countries the vehicles are older, but this is of little consequence because many analyses of data from low-rate countries find no large relationships between fatality rates and vehicle age. More recent vehicles may have stricter safety standards, but these could make no more than a modest difference, whereas the lowest rates are more than 99% below the highest rates. Roads, congestion, and other factors are also clearly different in the different countries. But these differences cannot come close to explaining the large variations observed. The only plausible explanation is that drivers are behaving sufficiently differently in the high-rate countries to generate a major portion of the observed difference.
The top photograph in Fig. 13-2 shows traffic in Cairo, Egypt, and the bottom photograph traffic in Adelaide, Australia. Table 3-1 (p. 43) provides the following values:


The Australian rate is 92% lower than the Egyptian rate. The vehicles in the upper photograph are somewhat different from those in the lower photograph, but not in any way that can explain much of the 92% difference. The Australian rate is 25% lower than the US rate. These differences in rates are consistent with observations of traffic in the three countries, which, while difficult to quantify, are nonetheless clear. I observed typical driving in Australia to be at somewhat lower risk than in the US, while driving in Egypt is at considerably higher risk than that in Australia or the US. The risk taking has many easily observable features, including less disciplined and less orderly traffic, as is clear in the photographs, and many more incidents of individual risk taking by drivers and pedestrians. The variations in rates among the countries in Table 3-1 and Fig. 3-8 (p. 44) fit a similar pattern. Low rates correspond to orderly, calm, and disciplined traffic, high rates to more disorganized competitive traffic.
It has been suggested that the downward trend in fatality rates observed in all countries is due to road users learning safer behavior by observing traffic. The evidence that so many countermeasures do change safety is sufficient to refute the extreme claim that all that is occurring is spontaneous learning. However, it is essentially impossible to determine what would have occurred if there had been no interventions, as all societies have taken measures to improve safety. There is very likely a spontaneous learning effect, and if so, it is an example of a safety improvement due to a behavior change.
Some of the fatality rate decreases associated with behavior changes reported in earlier chapters are listed in Table 13-1. The magnitudes of these effects far exceed the types of risk reductions that are associated with other safety measures, justifying the conclusion that how drivers behave is overwhelmingly the most important factor determining overall safety.
Table 13-1. Risk reductions from changes in driver behavior.

Other factors
Driver performance, while much less important than driver behavior, still has many important impacts on traffic safety. The main reason why various crash rates increase as drivers age is because of deteriorating driver performance. One of the reasons why younger drivers have highly elevated crash risks is performance inadequacies due to inexperience.
The importance of roadway engineering is apparent in Table 5-5 (p. 103). For example, replacing a rural arterial road by an Interstate reduces fatality
risk by 45%. This is much larger than the approximately 20% reduction estimated for the combined effect of all the FMVSS standards examined (p. 115), and larger still than the 10% reduction from frontal airbags (p. 286). The vehicle factor that produces a large change in risk is vehicle mass. However, this is more in the realm of a consumer choice than a vehicle factor, as such. The choice between a motorcycle and a large sedan is not generally considered a vehicle factor, although it is associated with a risk difference approaching a factor of 100. In the context of countermeasures, the term vehicle factors generally implies modifications and safety equipment for what is otherwise an equivalent vehicle.
Studies to identify factor contributions directly
In the 1970's two major studies, one in the US and one in GB, were performed to identify factors associated with crashes. Both studies were based on many thousands of crashes. The US study was performed by Indiana University, and is often referred to as the Tri-level study because crashes were examined in one of three levels of depth, depending mainly on their severity. The British study was performed by the Transport and Road Research Laboratory. , In both studies a team of multi-disciplinary experts conducted a detailed post-crash examination of crashes satisfying specified selection criteria. The crash site was examined for physical evidence, the vehicles involved were examined by an engineer, and the participants in the crash were interviewed in depth. Based on such information, factors contributing to the crash were identified.
The results of both studies are summarized in Fig. 13-3. The interpretation is that, for example, in the US study, the vehicle is identified as the sole factor in 2% of crashes, the interaction between vehicle and road user in 6% of the crashes, the interaction between vehicle, road user, and road (including environmental factors such as weather or darkness) in 3% of crashes, and the interaction between vehicle and road in 1%; the corresponding values for the British study are 2%, 4%, 1%, and 1%. In most cases the vehicle factor was worn tires or brakes, so that it was really due to improper maintenance, a behavior factor.

 

Figure 13-3 Percent contributions to traffic crashes estimated in British and US in-depth studies. Based on Ref. .

The studies were performed independently - indeed it appears that neither study group was aware of the activities of the other. The results are remarkably consistent. Each finds that the non-road-user factors of vehicle and road are rarely the sole factors associated with crashes - the British study finding 6% of crashes not linked to the road user, and the US study 7%. In both studies, when only one factor was identified, it was overwhelmingly the road user (65% in
the British study, 57% in the US study). Road user factors were found to be the sole or contributory factors in 94% of crashes in the British study, and 93% in the US study.
A smaller scale study using exclusively urban injury crashes found that vehicle factors played an even smaller role of about 1%. A study examining crashes in China attributes the following factors - human factors, 92.9%; motor vehicle, 4.5%; road 0.1%; and other 2.5%. In China about 70% of crashes involve bicycles, but only about 1% involve alcohol.
Studies such as the above should be interpreted with caution because identifying the mix of factors is not the same as identifying the most effective mix of countermeasures. An effective analogy points out that finding that mailed items are damaged by the human factor of careless handling does not mean improved handling is the most effective countermeasure. Better packaging is far more effective. Similarly, the finding that human factors are overwhelmingly involved in traffic crashes does not logically imply that countermeasures should focus primarily on human factors. However, decades of research have shown that, unlike the packaging analogy, only modest additional improvements in acceptable packaging of humans in vehicles are now feasible. The most expensive and intrusive packaging intervention, the airbag, reduces driver fatality risk by 10%, a reduction not competitive with those listed in Table 13-1. It is changes in driver behavior that have the potential to make, by far, the largest improvements in traffic safety.
The dominant role of driver behavior
Figure 13-4 shows the relative importance of factors schematically. The figure is a non-quantitative judgmental estimate in which the areas indicate the importance I attach to the different factors. Given the enormity of the losses from traffic crashes, all factors are of immense importance, but some are more important than others. Road user (or human) factors are more important than engineering factors. Among engineering factors, the roadway produces larger changes in risk than are associated with engineering changes to given vehicles. The human factor that plays the largest role is driver behavior, and changes in driver behavior are the key to changes in traffic safety. Behavior-change theories may offer guidance that could contribute to harm reduction.
In 1949 the comment a man drives as he lives entered the traffic safety literature (p. 220). The things that affect driving behavior are almost as diverse as those that affect life in general; they include family, personality, the era in which one lives, socio-economic status, religion, beliefs, traditions, etc. Below we address four areas of influence on driver behavior:
· Fear of adverse consequences.
· Social norms.
· Mass media.
· Legislative interventions.
These all interact with each other. What people fear is highly influenced by what their peers fear, and by what concerns the mass media stress, and so on.
Fear of adverse consequences
The most severe punishment for risky driving is not one issued by a court of law. It is a swiftly administered death penalty. While drivers are well aware that many people are killed in traffic, concern about joining their ranks is not normally present while driving. A subjective sense of control leads to a feeling that it cannot happen to me. Even if the occupant of a vehicle has no control over how it is driven, familiarity with road travel still generates a level of comfort higher than for far safer activities. Fear of terrorism and crime keeps many people from visiting certain cities and countries, yet the same people are happy to travel in a bus, taxi, or car driven by someone else, all far riskier activities. While driving is riskier than almost anything else a typical member of a motorized society is likely to do, there are reasons why it does not appear so to individual drivers.
The frequency of occurrence of a number of adverse events to drivers is
listed in Table 13-2. More than 26,000 drivers were killed in the US in 2002. In the same year, US vehicles traveled over 4 billion km. Thus, on average, a driver is killed for every 172 million km of driving. A driver traveling 25,000 km per year would take 7,000 years to accumulate this total. From the perspective of the individual driver, a fatality can appear to be an event
too improbable and remote to cause concern. This impression is further reinforced by driving day after day without any adverse consequences. The impression of safety does not appear to vary between countries with risks differing by a factor of more than one hundred. However, a passenger from a low-risk country does feel at risk in a taxi in a high-risk country, even though the driver does not.
Even when an adverse event does occur, it is rarely interpreted to be a natural consequence of normal driving behavior. Interpretations involving unpreventable bad luck are more prevalent. Notwithstanding the individual driver's perception of no risk, a by-product of current US average driving is over 40,000 traffic fatalities per year. In a paradoxical sense, individuals may not perceive there is a problem, although there clearly is.
While the rarity of the events in Table 13-2 may convey a personal sense of low risk, vastly rarer events, such as falling aircraft, shark attacks, tornadoes, and lightning cause greater concern. This is in part due to a media paradox. It is only rare events that make lots of news, and the more we hear about something, the more it enters our consciousness. Most people in the US know about an incident at the nuclear power plant at Three Mile Island in 1979. No death has ever resulted from a mishap in a US nuclear power plant, yet since 1979 over a million people have died on US roads. It seems possible that if nuclear power had killed a few thousand people (like the production and delivery of conventional power has done) it would be feared less.
In traffic, non-fatal adverse events of course do influence behavior. After a major scare in traffic, drivers tend to slow down, but the effect is of short duration. More serious harm is likely to have a larger influence. As Shakespeare notes "The injuries that they themselves procure must be their schoolmasters." (King Lear: Act 2, Scene 4). Direct evidence of behavior change as a result of being injured is provided by interviews of former patients who had been treated in a trauma center for motor vehicle injuries. Self-reported belt wearing rates were 85% after their crashes, compared to 54% before. While some learning occurs in response to adverse outcomes, experiencing traffic crashes is an unsatisfactory way to learn how to avoid them.
Social norms
Individual behavior is enormously influenced by the behavior of the members of the society in which the individual is embedded. People are particularly inclined to behave in ways that they believe will win approval of those whose approval they value. In the case of teenagers such peer pressure is particularly strong.
Autonomous behavior, or habits. As a society becomes more motorized, institutional changes are initiated in response to increasing casualties, which in turn influence social norms relating to driving. Adherence to many traffic safety measures eventually evolves into habit, or autonomous behavior, without any conscious safety consideration. Stopping at red lights is so ingrained that drivers rarely proceed after stopping at a red light until it turns green, even in the absence of crash risk, traffic, or police. There is a strong social norm against proceeding through a red light after stopping at it, even in conditions in which it would be no more dangerous to do so than proceeding after stopping at a stop sign. In Sweden and Switzerland pedestrians rarely cross roads against traffic signals, whereas in the US, Canada, and the UK they tend to cross when they judge it safe to do so without regard to traffic law or control devices. In motorized countries even the most impatient risky drivers rarely consider driving on the sidewalk to save a few seconds, although the practice is quite common in many less motorized countries, though illegal. In motorized countries driving even for short distances on the wrong side of the road attracts negative attention even if it poses little risk. When a social norm violation is unambiguous, condemnation is assured, and the practice tends to become rare.
However, the main behaviors that pose major risks involve continuous variables. Most motorists have on some occasions gone through traffic signals later than is prudent or legal, and have moderately exceeded speed limits. This clouds the intensity and clarity of their disapproval of other drivers who commit traffic light and speeding offenses that pose vastly greater risks than their own less frequent, less egregious violations.
Original motivations for improved personal hygiene were partially rooted in disease prevention. Such motivation rarely plays a role today. Rather, we find it difficult to conceive that prior highly civilized societies, such as Elizabethan England, did not share our fastidiousness. The cultivation of safe driving as habit seems a more productive approach than expecting drivers to be motivated by fear of consequences. A society that can evolve a social norm in which it is as unacceptable to speed or run red lights as it now is to proceed when facing a red light or drive on the sidewalk will enjoy an enormous safety improvement. There are precedents for large changes in social norms relating to safety behavior that would have seemed improbable before they actually occurred.
Changes in social norms relating to smoking. One of the most dramatic changes in social norms is in relation to smoking, an activity that caused more than 20 million Americans to die prematurely in the twentieth century. From 1950 to 2000, the US adult smoking rate dropped from 44% to 23%. Changes relating to smoking are obvious and dramatic. In the 1960s, when smoking in hospitals and doctors' offices (often by the doctor!) was common, few could have imagined a world in which smoking would be prohibited on airliners, and in airports, public buildings, restaurants, and even bars. No single event caused these changes - they were gradual and due to the cumulative effects from many inputs. The various Surgeon Generals' reports stressing the link between smoking and a host of illnesses stimulated increased discussion about the nature and origins of the habit, and its path to addiction. Health warnings on packages and the growth of smoking-cessation businesses and products resulted in reduced smoking. Various US commercial companies prohibited smoking in offices and cafeterias at a time when smoking was still allowed in the offices of the Environmental Protection Agency, an agency of the US Federal Government. Northwest Airlines prohibited smoking on all its North American flights long before any government-owned airline had adopted such a policy.
Speeding compared to drunk driving. While the social norm relating to drunk driving has changed, contributing to major reductions in drunk driving, the same cannot be said about harmful driving in general. Table 10-5 (p. 252) shows that 78% of traffic crash harm involves no alcohol. 87% of drivers in FARS 2002 do not have measured BAC > 0.08%. Even if all driving after consuming alcohol were eliminated, 62% of those killed in traffic would still be killed (Table 10-4, p. 251).
One of the largest reductions in traffic harm in the US can be attributed to Mothers Against Drunk Driving (MADD), which became influential under the leadership of a mother whose child, while walking, was killed by a drunk driver (p. 257). In the US, 367 child pedestrians (14 years or younger) were killed in 2002 in single-vehicle crashes. 28% of the involved drivers were administered alcohol tests, and of these, 81% had no detectable alcohol (measured BAC = 0). Recall that being intoxicated increases the risk of being tested, so the untested majority likely had lower BAC levels than the tested minority. It therefore is likely that well over 90% of the child pedestrians killed were killed in crashes with vehicles driven by sober drivers.
The behaviors associated with most of the harm in traffic do not attract the moral opprobrium focused on drunk driving. Driving 65 mph when the speed limit is 55 mph increases the risk of involvement in a fatal crash by a factor of two, similar to driving with BAC = 0.08% compared to driving at BAC = 0.
A prominent US politician (former Congressman and former Governor of South Dakota) boasted about his lead-foot driving, and received 12 speeding tickets from 1990 to 1994. In 2003, while speeding through a stop sign, he killed a motorcyclist, for which he received a 100-day sentence. In the post-MADD era nobody in US public life could have boasted about driving while drunk, nor retained a driving license after a long series of drunk driving convictions. If the social norm and legislative response to speeding had been more like that to drunk driving, the motorcyclist would not have died and the politician would not have had his own life devastated by the crash.
An activist movement to focus grief and anger on risk-taking sober drivers who harm others, especially children, has the potential to produce safety benefits like those produced by MADD.
Road rage. The claim that wildly aggressive driving has become more prevalent, as suggested by the term road rage, is not supported by any evidence. Certainly, many drivers behave badly today, but so did many in the past. There are no reliable data on non-crash aggression in traffic. However, the downward trends in fatality rates are more consistent with the notion that, while risks are presently high in traffic, they were much higher in the past.
Education and persuasion. The normal role of education is to impart know-
ledge. As lack of knowledge is not at the core of traffic crashes, safety benefits from education are necessarily limited. This conclusion has tended to foster the incorrect notion that persuasion is similarly ineffective. One problem is that the word persuasion has negative connotations, so it is often instead called education (which is known to be relatively ineffective). At the other extreme, calling it propaganda would assure its instant dismissal. Efforts to persuade still generally include specific items of information. For example, attempts to persuade drivers to fasten safety belts should appropriately mention that belts reduce risk, and that wearing is required by law, even though these facts are already well known by nearly all of the intended audience of non-wearers.
The advertising industry is successful at persuading people to do all sorts of things that they would not otherwise do, yet it is often difficult to show specific behavior changes resulting from advertising campaigns. The assertion that driving cannot be changed by the same methods known to change other behaviors is unsupportable. The potentially large, but uncertain, gains from persuading drivers to behave differently seem a better investment than expensive minor vehicle engineering changes that do not produce benefits even large enough to be measured in field data.
Mass media
Of all the many factors that contributed to declines in smoking, the largest contribution was likely from prohibiting cigarette advertising on television (in 1965 in the UK, 1971 in the US). Once television advertising was banned, smoking largely disappeared from made-for-television programming. The situation was different in movies, where tobacco companies pay large placement fees to have their brands incorporated directly into the movie plot, resulting in smoking appearing in 2002 movies about as often as in 1950 movies, even though smoking was twice as prevalent in 1950.17
Changes by the media in response to legislation and business considerations, together with voluntary restraints, contributed to reductions in deaths from two major causes - smoking and drunk driving. No corresponding changes have occurred relative to the driving behaviors that are responsible for the majority of traffic deaths. As noted above, even if drunk driving were completely eliminated, most traffic deaths would still occur. Yet the media continue to portray positively the use of vehicles in ways that are known to cause harm. Vehicle manufacturers often advertise their products in ways that glorify irresponsible driving. Movies and television programs, especially those
aimed at young people, often contain scenes depicting unrealistic occupant kinematics under crash conditions. For example, an unbelted driver, often the hero, may crash into a solid object at 60 km/h, jump out, and, uninjured and undaunted, pursue the chase by other means. In such fiction, car crashes are even presented as humorous events. The possibility that they can destroy lives is generally ignored.
There is unmistakable evidence that the massive portrayal of casual violence on television and movies leads to increased violent behavior. It is hard to imagine how the massive portrayal of the irresponsible use of vehicles in programs and advertisements would not lead to increased driver risk taking. The claim by media executives that what viewers observe does not affect their behavior is absurd. Commercial television receives all of its revenues from clients who pay billions of dollars in the confident belief that what viewers see in their advertisements will indeed influence their behavior towards purchasing their offerings. Most television viewers seeing someone eat a cookie in an advertisement or program do not copy the portrayed behavior any more than most people who see reckless driving by unbelted drivers copy the behavior. But some do - otherwise there would be no advertising. While automobile manufacturers would likely oppose government advertising standards, those manufacturers that are currently already behaving as better public citizens would likely benefit from such restraints. It seems to me undeniable that the media causes a large number of traffic deaths by portraying risky, irresponsible, and illegal driving as an exciting, glamorous, heroic, or humorous activity without harmful consequences. The period known as McCarthyism ended in 1954 when Senator McCarthy was asked in front of a large live television audience "Have you no shame?" I believe media executives should be asked this same question, even if it is unlikely to be given much coverage in the media they control.
Legislative interventions
Traffic law has existed since the beginning of vehicular traffic. It has two goals - to insure efficient traffic, and to promote safety. One of the earliest laws adopted in all countries to further both goals was to require all traffic to travel on one side of the road. There are no known safety or efficiency advantages in choosing one side over the other. 166 countries, including all in North America and continental Europe, require traffic to drive on the right. 74 countries, including the UK, India, Japan, Australia, Ireland, and Malaysia, require traffic to drive on the left.
Since the earliest days of motorization, it became clear that some drivers were behaving in ways that threatened public safety. Law contributes to safety by requiring drivers to behave in ways that professionals, based on analyses of data and collective expert judgment, have concluded are safer. Despite the subjective confidence of so many drivers, individual personal experience and common sense are inadequate guides to safe driving, just as they are inadequate for identifying safe food or safe drinking water.
Because traffic law relating to speed, intoxication, signal devices, and the rules of the road has been evolving for so long it is not possible to estimate its total influence on safety today. The law has many effects, including education and influencing social norms. If all traffic law were to be suddenly abolished, traffic safety would change, but it would not revert to what it would have been if the laws had never existed. The case of belt wearing in the US will suffice to illustrate. Prior to belt laws, US wearing rates remained fairly constant for many years at about 14%. After the passage of belt laws, rates increased over many years to 75% by 2002. There is of course no way to know what would happen to wearing rates if all belt wearing laws were suddenly abolished. However, a sudden drop to 14% is implausible, whereas settling to a new level below 75% but far above 14% seems far more likely. Habits molded in part by prior law seem likely to persist for most drivers.
Although it is not possible to estimate the aggregate effect of law on safety, we can get a sense of how enormous it is by measuring the effects of adding a law, or strengthening or more strongly enforcing an existing law. Random breath testing in New South Wales, Australia, decreased total fatalities by 19% (p. 254). This large effect was relative to a prior period in which laws against drunk driving were already in place. Reductions in speeds on the US rural Interstate system in response to a change in the speed limit were largely responsible for a 34% reduction in the fatality rate (p. 213). Prior to the change, a speed limit was already in place, albeit a higher one.
Many studies have estimated the casualties that result from violating present law. A review estimated that about 50% of traffic crashes in Europe could have been prevented if road users were completely dissuaded from committing traffic violations. Considerably more than 50% of fatalities would be eliminated, because fatal crashes involve more egregious law violations than typical crashes. Eliminating the single violation of driving at illegal BAC limits would prevent about 34% of US traffic deaths (p. 251). The UK's belt wearing law decreased fatalities to affected occupants by 20% (p. 296). Repeal of
US laws requiring motorcyclists to wear helmets increased motorcyclist deaths by 25% (p. 299).
These outcome changes produced by behavior changes resulting from changes in law overwhelm contributions from any of the other factors listed in Fig. 13-4. A road engineering safety improvement will apply to only a small fraction of all roads. A vehicle safety improvement takes years to design and manufacture, and much longer if it is part of a regulation. So a mandated device that reduces driver fatality risk by 10% will prevent no deaths for some years after it is proposed, and only about 1% of the driver deaths in the first
year it is installed, because about 90% of the vehicles on the roads date from before the change. As about 60% of traffic fatalities are drivers, it will reduce
traffic deaths by 0.6%. For a vehicle safety regulation, typically about 15 years elapses between proposal and essentially all vehicles on the road containing
the regulated change. On the other hand, laws applying to all drivers can be passed quickly, and can immediately start generating large safety benefits for all road users.
Effect of enforcement of laws on casualties
The fact that a law is on the books will, by itself, change the behavior of some drivers who believe that obeying the law is a canon of good citizenship. Other drivers change their behavior to avoid the penalties specified in the law. The extent to which they do so depends on their perception of how well the law is enforced. This is related to how well it really is enforced. When new laws come into effect, compliance is often initially high, only to decline later as drivers discover by personal observation, or reports from acquaintances, that detection is unlikely. The goal of enforcing traffic law is, of course, not to achieve compliance with some set of rules for its own sake, but to reduce casualties. The effect of enforcement on casualties is therefore a central issue in traffic safety.
A direct relationship between a driver getting a traffic ticket and that same driver being subsequently involved in a fatal crash was established using data from Ontario, Canada. The case-crossover method was applied to 8,975 licensed drivers involved in fatal crashes during an extended study period. The fatally injured drivers had received convictions at a rate of about one every 5 years before their fatal crash involvement. The risk of a fatal crash in the month after a conviction was found to be (35 ± 8)% lower than in a comparable conviction-free month. The effect quickly dissipated, and had effectively disappeared after 3 months. However, it shows large reductions in the risk of fatal crash involvement as a direct result of law enforcement. The data suggest that about one death was prevented for every 80,000 tickets issued.
Other studies have associated enforcement with casualty reductions. Based on synthesizing the results from a large number of studies in many countries, a 14% reduction in fatal crashes was associated with manual enforcement.23, Manual enforcement is the traditional arrangement of a police officer in a stationary or moving vehicle with speed-measuring equipment who stops and issues traffic citations to drivers violating speeding or other traffic law. While such enforcement is shown to make large differences, it is costly and resources permit its application to only a few locations at a given time.
Enforcement using newer technology. Vastly more effective enforcement is possible using the new technologies of photo radar and red-light cameras. These can enormously increase the probability of detecting law violations. I suggest in Chapter 16 that such technologies can form the core of a new approach that would produce huge reductions in traffic harm.
Interactive effects
Interactive effects can make it difficult to evaluate interventions, especially those that produce relatively small effects. The problem arises because the traffic in which crashes take place is a highly interactive system. Every component is connected to every other component to some extent. It is rarely possible to address the influence of one factor, all other factors remaining the same, because all other factors rarely remain the same.
There can be little doubt that driving on snow presents greater risks than driving on dry pavement, all other factors being equal. Yet Chapter 5 showed that fatality risks are systematically lower when it snows. Speed reductions in response to the perceived greater risk more than neutralize the increased risk from reduced roadway friction. Similarly, antilock brakes do not lead to harm reductions even though they improve braking. Less safe older drivers pose less threat to the safety of others because they drive less. Drivers react to changes in perceived safety, whether environmental, personal, or technological in origin.
Such interactive effects have generated much controversy and passion over the years, generally because the originators of devices with effective engineering performance or of seemingly effective policies are reluctant to accept evaluation results that show they did not work as planned. On the other hand, interactive effects have been blown out of proportion, and spawned much theoretical nonsense and needlessly complex explanations. Yet surely it has been obvious since antiquity that humans change their behavior in response to the perceived probability and severity of harm. We walk more carefully when the ground is icy than when it is not, and we walk more carefully on rough surfaces when barefoot than when wearing shoes. A warrior clad in armor may accept a greater risk of being struck by a weapon than one not so clad, and so on. In the 1988 movie Dangerous Liaisons, Valmont states, "But it is always the best swimmers who drown." (Data confirm this; if you want to reduce your children's chances of drowning, do not teach them to swim). Shakespeare writes, "Best safety lies in fear." (Hamlet, Act I, Scene 3).
Interactive effects in traffic have been recognized for a long time
In 1938 a paper titled, A Theoretical Field-Analysis of Automobile Driving, was published in the American Journal of Psychology, and contained the following:
More efficient brakes on an automobile will not in themselves make driving the automobile any safer. Better brakes will reduce the absolute size of the minimum stopping zone, it is true, but the driver soon learns this new zone and, since it is his field-zone ratio which remains constant, he allows only the same relative margin between field and zone as before. (p 458)
In 1949 the following appeared in the Journal of the Royal Statistical Society:
It is frequently argued that it is a waste of energy to take many of these steps to reduce accidents. There is a body of opinion that holds that the provision of better roads, for example, or the increase in sight lines merely enables the motorist to drive faster, and the result is the same number of accidents as previously. I think there will nearly always be a tendency of this sort, but I see no reason why this regressive tendency should always result in exactly the same number of accidents as would have occurred in the absence of active measures for accident reduction. Some measures are likely to cause more accidents and others less, and we should always choose the measures that cause less. (p 13)
Given how long interactive effects in traffic safety have been discussed in such reasonable terms in the technical literature, and recognized much earlier in other contexts, it is surprising how many later claims of discovery there have been, and more surprising still, how often the very existence of such effects has been hotly denied.
Human-behavior feedback - the technology/human interface
Various terms have been used to describe interactive effects in traffic, including human-behavior feedback , (p 283) and the technology/human interface. Fig. 13-5 compares two models, labeled the naive model and the realistic model, of how technological changes affect safety. The names listed reflect historical usage, and do not imply that engineers deny interactive effects. Terms such as risk compensation and danger compensation have been used for the realistic model. I consider these inappropriate because mechanisms other than risk may be involved, and changes may not be only compensation.
Realistic model. Let us suppose that some change is introduced into a traffic system that is expected to change safety by some fraction, say DSeng, assuming users continue to behave exactly as they did before the change. The subscript denotes that the change is of an engineering nature. For example, if design changes to a guardrail were estimated, by engineering methods, to reduce the probability of driver death on impact by 10%, then DSeng would be 10% for drivers killed crashing into the guardrail. We use DSeng more generally to indicate fractional reductions in some harm measure expected from changes to a system if users do not alter their behavior in response to these changes. The change might be modifying vehicles, reducing speed limits, etc. While safety interventions always aim at producing positive values of DSeng, there are other changes motivated by different considerations, such as saving fuel in the case of smaller cars, for which the values of DSeng are negative. 

Figure 13-5. Contrast between a naive model that ignores interactive effects and a realistic model that includes them.

Because road users may alter their behavior, the actual realized percent safety change that is observed, represented by DSact, may differ from DSeng. The quantities can be considered to be related in the following simple way:
13-1
where f is a feedback parameter which characterizes the degree to which users respond to the safety change. In this context, feedback is synonymous with user reaction, behavior change, or interactive effects in the system. If users do not change behavior in response to the safety change, then f = 0, and the safety change is just as expected on engineering grounds. If the safety change is in the expected direction, but of lesser magnitude than expected, then -1 < f < 0, and the safety change is discounted compared to the expected amount. If the safety change has no effect, then f = -1.
The naive model assumes that f = 0, while the realistic model accepts that, in principle, f is not restricted to any range of values. It is a parameter that is determined by comparing observed outcomes with those expected if there were no interactive effects.
Rich variety of responses observed. An analysis of 24 studies provided examples in which interventions aimed at increasing safety produced the following observed outcomes:29,30(p 284-290)
· Safety increased even more than expected (f > 0).
· Safety increased about as much as expected (f = 0).
· Safety increased, but less than expected (-1 < f < 0).
· Close to zero effect (f = -1).
· Perverse effect - safety decreased (f < -1).
Measures expected to reduce safety, but introduced for other reasons, had similarly varied outcomes.29,30(p 290-294) For example, when Sweden changed from driving on the left to the right side of the road in 1967, and Iceland did the same in 1968, the naive expectation was that such a change would increase crashes. In fact, crashes declined in both countries. (p 139), (p 215)
Behavioral responses are likely for any intervention. Driver behavior changes
are usually reliably observed in response to technologies that provide clear feedback to the driver. Antilock brakes provide such an example (Chapter 5). For technologies that affect only injury risk when crashes occur, behavior effects are expected to be smaller, and therefore more difficult to measure. Indeed, there is no definitive body of empirical evidence demonstrating such effects. This has given rise to some claims that devices that affect only the risk of being killed in a crash do not influence driving behavior. I believe that the following thought experiment establishes that they do.
Consider two hypothetical vehicles, identical in all respects except that one has the magical property that neither it nor its occupants can be hurt in any crash, while the other is wired with dynamite to explode on the slightest impact. Even in the unlikely event that a few drivers would ignore the difference, there can be no doubt that, on average, the vehicles would be driven differently. The same conclusion applies even if the vehicles were in fact identical, but falsely believed to possess the hypothesized properties.
Changes in perceived protection can be viewed as lying along a continuum bounded by these hypothetical extremes. It seems likely that any change in perceived outcome will have some effect on driving. In Chapters 11 and 12 we discussed the possibility of small increases in driver risk taking in response to the perceived increased protection from belts and airbags. It was suggested that a modest reduction in the benefits of belts is possible from such an effect. Because of the perception that airbags are much more effective than they are, it is possible that small behavior changes could be of sufficient size to even reverse the small effectiveness values (DSeng in the present terminology) listed in Table 11-4, p. 286. Although it is not generally explicitly stated, such effectiveness estimates are all based on assuming the naive model.
To place the difficulties of measuring the behavioral parameter, f, in context, recall how difficult it was to obtain even the DSeng values for occupant protection devices derived in Chapter 11. The effect that a device has on outcome when a crash occurs provides estimates only of DSeng, which may not have even the same sign as DSact, and it is DSact that is crucial for safety.
Comment on "risk homeostasis theory"
This subject is included only because it keeps getting the occasional mention in the literature, and additional people encounter it for the first time and find it seductive. Risk homeostasis theory claims that drivers have a target level of risk per unit time, so that physical changes to the traffic system stimulate user reactions that reset safety to its prior level.33, The finding that changes in safety systems generate outcomes all the way from effects greater than expected to opposite to expected is sufficient to dismiss this claim, which is categorically refuted by voluminous data. The distance fatality rate on local rural roads is over 400% higher than on urban Interstates (Table 5-5, p. 103). For the US, the rate was over 1,400% higher in 1921 than in 2002. Risks vary by more than a factor of 100 between countries (Table 3-1, p. 43). Such enormous variation in risk on different roads, in different countries, and at different times hardly invites an interpretation in terms of a target level of driver risk.
Risk changes by large amounts during an individual trip. Crash, injury, and fatality risks are enormously higher when driving through intersections than when driving between intersections. Although most drivers must surely be aware of this, they are in little better position to equalize these risks than is a pilot to equalize the risk per unit time at landing to that when cruising at 35,000 feet. Replacing roads containing intersections by limited-access freeways reduces crash risk as certainly as eliminating the take-off and landing risk would reduce air-travel risk.
When the homeostasis notion first appeared around 1970 it played a positive role in stimulating thinking about interactive effects, and highlighted the importance of motivational factors. The original specific claims were quickly demolished by data. The notion kept getting reformulated in one of two ways. In the first, revised specific claims were made which were quickly demolished by data. In the second, a more metaphysical formulation led to claims that could not, in principle, be falsified by data. So, one can regard this in either of two ways. First, either as theory in the usual way science uses the term, but one that has been convincingly refuted by data, and ought therefore to have quickly disappeared. Second, one can regard it as a collection of non-scientific metaphysical claims not addressable by data, in which case the term theory must not be used, and one must wonder why it is discussed at all. In my own view it has for far too long been much ado about nothing. Frank Haight comments:
There is some question as to whether the theory is meaningless (since incapable of testing) or simply false. Evans' conclusion that "there is no convincing evidence supporting it and much evidence refuting it" is if anything generous. In my view, a sufficient argument against the validity of risk homeostasis is provided by the incoherence of its "theoretical" formulation. (p 364)
Traffic safety and mobility
Traffic safety is all too often discussed as if the only goal in creating traffic systems were safety. On the contrary, the goal is mobility, and crashes are an unwanted by-product to be minimized while achieving, not abandoning, this primary goal. An important measure of mobility is trip time. While the goals of safety and mobility are often in conflict, this is not always the case.
Safety measures that increase mobility. Roadway engineering changes generally increase both safety and mobility. Replacing a rural arterial road by an Interstate reduces fatality risk by 45% while at the same time increasing travel speeds. Replacing signalized intersections with modern roundabouts increases safety and under most conditions also mobility. Replacing intersections with overpasses and providing elevated or underground pedestrian crossings increase safety and mobility. Vehicle technology improvements such as improved braking, tires, lighting, and navigation have the potential to improve both safety and mobility. Devices such as ABS designed with safety in mind can end up improving mobility more than safety. Night vision systems will certainly increase mobility even if their effect on safety is uncertain. Likewise, an effective drowsy driving detection device will likely increase mobility and safety.
Safety measures that have zero, or minimal, effect on mobility. Vehicle safety improvements and genuinely passive safety protection devices should increase safety at no mobility cost. Safer highway furniture (break-away signs, etc.) that reduce injury risk on impact improve safety without affecting mobility. The two seconds it takes a driver to fasten a safety belt increases the duration of a typical 15 minute trip by 0.2%, a mobility reduction, but one of inconsequential magnitude.
Safety measures that reduce mobility. Driver licensing reduces the mobility of those denied licenses. The main reasons for such denials are age (per se, before a specified age, but by less straightforward means for older drivers), performance impairment (medical, failure to pass test), or criminal sanction. Successfully prohibiting any fraction of the population from driving reduces crashes. The percent reduction in crashes exceeds the percent reduction in drivers if those prevented from driving have above average crash rates. However, an above average crash risk is not grounds to deny a license. If it were, the logical consequence would be to prohibit males from driving. If one had sufficiently detailed knowledge about all drivers, successive application of such a philosophy would eventually eliminate every driver except the one safest driver.
Speed control is the intervention with the greatest potential to reduce casualties, but it is also the one most in conflict with mobility. Reducing speed limits leads to real increases in trip time, which has economic consequences. In societies like the US in which most goods are transported by trucks, lower
speed limits increase the cost of just about everything. As discussed in Chapter 9, such matters can be addressed in terms of socially optimum speeds that take into account safety and mobility costs. Stricter enforcement of existing speed limits likewise increases average trip times, and major reductions in casualties are obtainable through compliance with present speed limits. If tailgating were to be reduced, the maximum flow on freeways would drop producing increased delays at peak periods. Drivers who obey traffic lights have longer trip times than those who run red lights.
Interactive effects attenuate these mobility costs. Crashes produce long, but infrequent, delays to all motorists, made all the longer by gawkers who slow down even more than the altered traffic conditions require. Manual enforcement leads to substantial delays to those ticketed, but the ticketing process also slows other traffic (mostly by encouraging greater adherence to the speed limit). For many trips the expected variance is crucial. My average trip time to the airport is 45 minutes, but the standard deviation is 15 minutes. So I must allow 90 minutes to have a 99% chance of arriving in 45 minutes or less. If enforced safety measures increased my average trip time to 50 minutes but reduced the standard deviation to 10 minutes, then allowing 80 minutes would produce the same 99% probability of completing the trip in 50 minutes or less. The yellow phase in traffic signals is usually set with some allowance for vehicles running the red. If there were confidence that this would not occur, shorter yellow phases would lead to higher throughput and reduced delay. Increased enforcement generates more orderly traffic, which is often accompanied by travel efficiencies even beyond the avoidance of delay from crash incidents. Which traffic is more efficient - that in the upper or lower picture in Fig. 13-2? Also, it should not be forgotten that many of those injured in traffic lose much of their mobility, and those who die in traffic lose all of their mobility.
Contrast with airline safety
While traffic fatality rates have declined steeply in time, those for airline travel have declined far more steeply. On average, the risk of death for the same distance of travel on a scheduled US airline is about 99% less than in road travel. However, it should be kept in mind that the risk is the same to all on board an aircraft, whereas some road drivers have crash risks much lower than the average. The main contributors to the road average are high-risk drivers. Intuitively, it would appear to be far more dangerous to fly 35,000 feet above the ground at a speed of 1,000 km/h in a vehicle designed with little safety margin due to overriding weight considerations than in a more robustly constructed road vehicle. The skies being empty while the roads are dense with threatening vehicles is not an explanation. Single-vehicle road crashes led to the deaths of 18,476 vehicle occupants in 2002 (see also Table 3-3, p. 48), whereas not even one occupant of a US scheduled airline was killed in 2002 in any type of crash. It is therefore not surprising that one often hears the question, Why does road traffic safety not make the dramatic progress of airline safety? Table 13-3 presents a list of differences between the two modes.

Table 13-3. Comparison of safety characteristics of US commercial air carriers and US road transportation.

commercial airline road traffic
deaths per billion km of occupant travel 0.07 4.9
countermeasures with most success and potential crash prevention crash prevention
main US policy emphasis crash prevention crashworthiness
impact of vehicle designor manufacturing flaws vitally important minimally important
driver selection strict essentially everyone
importance of driver skill and knowledge high may increase or decrease crash risk
main influence on driver behavior following increasingly effective procedures experience and personal judgment
violation of safety laws rare typically, many times per trip
use of alcohol/drugs rare alcohol in about 40% of fatal crashes
value of high technology driver training simulators enormously high value zero or minimal value
time to react to crash-threatening situations often more than many seconds or minutes usually less than a second
value of crash-avoidance advanced technology enormously high value minimal value
key to making largest improvements in safety safer aircraft flown by better trained pilots adhering to better procedures behavior changes resulting from changes in social norms, legislation, and enforcement
Deaths per billion km is not a good measure for airline safety because risk, being so concentrated at take-off and landing, is largely independent of trip distance.37, The illustrative value, 0.07 deaths per billion km of occupant travel, is based on an airline trip of average distance, and fatalities averaged over a ten-year period.39 The comparison, while approximate, leaves no doubt that average airline travel is dramatically safer than average road travel. The crucial difference is that pilots adhere to rules specified by combining the knowledge and experience of many experts. Pilots may not, for example, take off after another airline has taken off based on what their personal experience tells them is a safe following headway. In the airline case, learning to avoid crashes by experiencing them is unacceptable, yet it is a large component of the approach to road safety.
Much attention has been devoted to increasing cabin safety in the event of an airliner crash. There are standards for fire-retardant fabrics, seat strength, safety belts, emergency exits, emergency cabin illumination, and evacuation equipment and procedures. Important though such measures are, they are not the key to why airline travel is so safe. Improvements in crashworthiness never led to any ambiguity about what the main focus of aviation safety should be. The number one priority is to avoid the crash rather than marginally increase the probability of surviving it. When road safety transfers its main focus to the prevention, rather than the survival, of crashes, it can start moving in the direction of the extraordinary safety achieved in airline travel.
Relative importance of factors
The schematic in Fig. 13-1 summarizes what I consider to be the main components relevant to traffic safety, with my judgment of their relative importance indicated in Fig. 13-4 (p. 339). Safety has been conceptualized in other formalisms, including particularly the Haddon Matrix.10 This is a 3 3 classification in which all factors are either human, vehicle, or environment, and either pre-crash, crash, or post-crash. While this has been helpful in the past, I believe it now places too much emphasis on the event - the crash. Pre-crash normally implies the few seconds prior to the crash, whereas I consider the really important pre-crash period to be the decade of socialization prior to the day of the crash. The focus needs to shift away from the details of the event, and more in the direction of eliminating it. The classification into nine cells may tend to place more emphasis on relatively less important factors at the expense of more important factors. Once occupants wear belts (a behavior and legislative matter), only modest further improvements in the crash phase features of vehicle design and occupant protection beyond those already in place are likely to occur. Likewise, improvements in rescue and medicine will save lives in the post-crash phase, but the potential from further improvements is limited by the already high level reached. In sharp contrast, no nation has more than touched the tip of the iceberg of harm reductions that can be achieved by behavior changes. Herein lies the opportunity for the breakthrough advance in traffic safety that is proposed in Chapter 16.

Summary and conclusions (see printed text)

References for Chapter 13 - Numbers in [ ] refer to superscript references in book that do not correctly show in this html version.  To see how they appear in book see the pdf version of Chapter 1.

[1] Hertzman C. Health and human society. Am Scientist. 2001; 89: 538-545.

[2] Evans L. Driving and human health (letter to the editor). Am Scientist. 2002; 90: 4-5.

[3] Treat JR. A study of precrash factors involved in traffic accidents. The HSRI Research Review. Ann Arbor, MI; May-August 1980.

[4] Sabey BE, Taylor H. The known risks we run: The highway. In: Schwing RC, Albers WA, editors. Societal Risk Assessment – How Safe is Safe Enough? New York, NY: Plenum Press; 1980, p. 43-63.

[5] Sabey BE, Staughton GC. Interacting roles of road environment, vehicle and road user in accidents. Presented to the Fifth International Conference of the International Association for Accident and Traffic Medicine, London, UK; 1975.

[6] Rumar K. The role of perceptual and cognitive filters in observed behavior. In: Evans L, Schwing RC, editors. Human Behavior and Traffic Safety. New York, NY: Plenum Press; 1985, p. 151-165.

[7] Carsten OMJ, Tight MR, Southwell MT. Urban accidents: Why do they happen? Basingstoke, UK: AA Foundation for Road Safety Research; 1989.

[8] Wang, Z. Road traffic accidents (RTA) in China – 1998 in retrospect. J. Traf Med. 2000; 28: 15.

[9] Wang Z, Jiang J. An overview of research advances in road traffic trauma in China. Traf Inj Prev. 2003; 4: 9-16.

[10] Haddon W Jr. A logical framework for categorizing highway safety phenomena and activity. J Trauma. 1972; 12: 193-207.

[11] Gielen AC, Sleet D. Application of behavior-change theories and methods to injury prevention. Epidemiologic Reviews. 2003; 25: 65-76.

[12] Slovic P, Fischhoff B, Lichtenstein S. Facts and fears: Understanding perceived risk. In: Schwing RC, Albers WA, editors. Societal Risk Assessment – How Safe is Safe Enough? New York, NY: Plenum Press; 1980, p. 181-214.

[13] Ruscio J. Risky business. Skeptical Inquirer. 2000; 24: 22-26.

[14] US Nuclear Regulatory Commission. Fact sheet on the accident at Three Mile Island (revised March 2004). http://www.nrc.gov/reading-rm/doc-collections/fact-sheets/3mile-isle.html

[15] Passman C, McGwin GJ, Taylor AJ, Rue LW. Seat belt use before and after motor vehicle trauma. J Trauma. 2001; 51: 105-109.

[16] Koop CE, Richmond J, Steinfeld J. America’s choice: Reducing tobacco addiction and disease. Am J Public Health. 2004; 94: 174-176.

[17] Glantz SA, Kacirk KW, McCulloch C. Back to the future: Smoking in movies in 2002 compared with 1950 levels. Am J Public Health. 2004; 94: 261-263.

[18] CNN.com Law Center. Janklow sentenced to 100 days in jail. 22 January 2004.

http://www.cnn.com/2004/LAW/01/22/janklow.sentencing

[19] Fumento M. Road rage versus reality. Atlantic Monthly Online. August 1998.

http://www.theatlantic.com/issues/98aug/roadrage.htm

[20] Anderson CA, Bushman BJ. The effects of media violence on society. Science. 2002; 295: 2377-2378.

[21] Lucas B. Which side of the road do they drive on? http://www.brianlucas.ca/roadside

[22] Sivak M. How common sense fails us on the road: Contribution of bounded rationality to the annual worldwide toll of one million traffic fatalities. Trans Res Part F: Psych Behav. 2002; 5: 259–269.

[23] Zaidel DM. The impact of enforcement on accidents. The “Escape” Project Contract No. RO-98-RS.3047; February 2002. http://www.vtt.fi/rte/projects/escape/escape_d3.pdf

[24] European Transport Safety Council. Police enforcement strategies to reduce traffic casualties in Europe. ETSC report. Brussels; May 1999. http://www.etsc.be/strategies.pdf

[25] Redelmeier DA, Tibshirani RJ, Evans L. Traffic-law enforcement and risk of death from motor-vehicle crashes: Case-crossover study. The Lancet. 2003; 36: 2177-2182.

[26] Elvik R, Mysen AB, Vaa T. Trafikksikkerhetshåndbok (Traffic Safety Handbook). Oslo, Norway: Institute of Transport Economics; 1997.

[27] Gibson JJ, Crooks LE. A theoretical field-analysis of automobile driving. Am J Psych. 1938; 5l: 453-47l.

[28] Smeed R. Some statistical aspects of road safety research. J Roy Stat Soc, Series A. 1949; 112: 1‑34.

[29] Evans L. Human behavior feedback and traffic safety. Hum Factors. 1985; 27: 555-576.

[30] Evans L. Traffic Safety and the Driver. New York, NY: Van Nostrand Reinhold; 1991.

[31] Evans L. Transportation Safety. In: Hall RW, editor. Handbook of Transportation Science, Second Edition. Norwell, MA: Kluwer Academic Publishers; 2002, p. 79-86.

[32] Näätänen R, Summala H. Road-User Behavior and Traffic Accidents. Amsterdam, Netherlands: North Holland; 1976.

[33] Wilde GJS. The theory of risk-homeostasis: Implications for safety and health. Risk Analysis. 1982; 2: 209-255.

[34] Wilde GJS. Notes on the interpretation of traffic accident data and of risk homeostasis theory: A reply to L. Evans. Risk Analysis. 1986; 6: 95-101.

[35] Evans L. Risk homeostasis theory and traffic accident data. Risk Analysis. 1986; 6: 81-94.

[36] Haight FA. Risk, especially risk of a traffic accident. Accid Anal Prev. 1986; 18: 359-366.

[37] Evans L, Frick MC, Schwing RC. Is it safer to fly or drive? – A problem in risk communication. Risk Analysis. 1990; 10: 239-246.

[38] US Department of Transportation, Bureau of Statistics, Table 2-1 – Transportation fatalities by mode.

http://www.bts.gov/publications/national_transportation_statistics/2003/html/table_02_01.html

[39] Sivak M, Flannagan MJ. Flying and driving after the September 11 attacks. Am Scientist. 2003; 91: 6-8.