Additional material in Traffic Safety (2004)
Chapter 14. TRAFFIC SAFETY IN BROADER CONTEXTS (From 1991 book Traffic Safety and the Driver)
Words only (no formatting, figures, tables, or photographs) from 1991 book
This book, in keeping with its primary goal, has so far focused exclusively on the subject of traffic safety. In the present chapter we place traffic safety in various broader contexts.
TRAFFIC SAFETY IN THE CONTEXT OF SAFETY
There are more than two million deaths per year in the US [National Center for Health Statistics 1989]. About 140 000 of these are due to injury [Injury in America 1985], about 45 000 being traffic fatalities. The numbers of deaths due to the two leading causes of death, three quarters of a million from heart disease and half a million from cancer, far outnumber those due to injuries. Such a comparison ignores the crucial question of the ages at which death occurs. The risk of death from disease increases rapidly with age, whereas the risk of death from injury peaks at young ages; specific functional dependencies are given in, for example, Cerrelli . The comparison also ignores the inevitability of death; every US death certificate cites some trauma or disease, so, independent of safety or medical advances, every mortal must eventually increase the total listed under some cause of death.
When one takes into account the ages at which the deaths occur, a
different picture emerges. Fig.
14-1 shows that about twice as many preretirement years of life are lost as a
result of injury as from each of the
two leading causes of death. The figure also makes the point that research expenditures for the two leading diseases greatly exceed those for research on injury. Later in this chapter I suggest that the resources that are available for traffic safety research tend to be not as wisely used as they might be.
Fig. 14-1 about here
Given that a death occurs, the probability that it was the result of any type of motor vehicle crash is shown in Fig. 14-2. Although many more males than females die in traffic crashes (Figs 2-5 and 2-15), there is little difference dependent on sex in Fig. 14-2 because males are also more likely to die from other injury causes (the main reason for early death) at the ages at which motor vehicle deaths are high. Given that a 19-year-old dies, there is an approximately even chance that the death was due to a traffic crash. The large increase after the minimum at age 15 is not due solely to the commencement of driving; a corresponding increase also occurs in pedestrian deaths (Figs 2-15 and 2-16). The steep decline with increasing age is because traffic fatality risk declines at the same time as the risk from other causes increases. Given that death occurs, it is about 50 times more likely to be due to a motor vehicle crash at age 20 than at age 65.
Fig. 14-2 about here
Effects on longevity
Fig. 14-3 shows calculated increases in longevity from eliminating all
traffic fatalities, without anything else changing [Evans 1988; Evans and
Blumenfeld 1982]. For boys at birth the increase is 242 days, or two thirds of a year; for girls 111 days. At age 65 the increases are 15 days for males and 12 days for females.
Fig. 14-3 about here
Fig. 14-4 shows the total increase in longevity in the US which would result if all fatalities to individuals of one age were prevented. For example, if all fatalities to 20-year-olds were prevented, this would eliminate 1309 male deaths; as life expectancy for a 20-year-old male is 52.7 years, the elimination of these traffic fatalities generates an additional 1309 X 52.7 = 69 000 years of total longevity in the US. The corresponding calculation for 65-year-old males is 173 X 14.6 = 2500, so that the increased longevity from eliminating fatalities to 20-year-old males is 28 times as great as from eliminating fatalities to 65-year-old males.
Fig. 14-4 about here
Car safety compared to safety using other modes of travel
Airline passengers awaiting take-off are sometimes told that they have
already completed the most dangerous part of their trip -- the drive to the
airport. The impression that air
travel is so much safer than car travel arises from the most widely quoted
death rates per billion miles for each -- 0.6 for air compared to 24 for road. There are three reasons why such a
comparison is inappropriate. First, the airline rate is passenger fatalities
per passenger mile, whereas the road rate is all fatalities (any occupants, pedestrians, etc.) per vehicle mile. Second, road travel that competes with air travel is on the rural Interstate system, not on average roads. Third, driver and vehicle characteristics, and driver behavior, lead to car-driver risks that vary over a wide range. In contrast, airline fatality risk is similar for all travellers. Evans, Frick, and Schwing  investigate influences due to driver age, alcohol use, safety belt use, car mass, and roadway type. The age distribution of airline passengers fatally injured in the eight worst 1975-1985 US airline crashes is used to infer car-driver fatality risk for drivers with the age distribution of airline passengers. Because risk of death on a flight does not depend on flight distance, airline fatality risk per mile decreases with flight distance. Expressions derived to compare risks for drivers with given characteristics to those on airline trips of given distance showed that 18-year-old, unbelted, intoxicated, male drivers of cars 700 pounds lighter than average are substantially more likely to be killed on the trip to the airport than on the flight. However, drivers with the age distribution of airline passengers are less likely to be killed on the trip to the airport than on the flight. It is further concluded that 40-year-old, belted, alcohol-free drivers of cars 700 pounds heavier than average are slightly less likely to be killed in 600 miles of rural Interstate driving than in regularly scheduled airline trips of the same length. For 300-mile trips, the air travel fatality risk is about twice that for driving. Hence, for this set of particularly safe drivers, car travel provides a lower fatality risk than air travel for trips in the distance range for which car and air travel are likely to be competing modes.
Many sources compare safety for different travel modes; data on p. 87
of National Safety Council  are typical.
The data for 1987 show 9.7 car
and taxi occupant deaths per billion miles of travel; the comparison rates for passengers using other modes are, for busses 0.3, for passenger railroads 1.3, and for scheduled airlines 0.7. In examining such averages, the massive contribution to the car rate from intoxicated youths should be kept in mind; if you are not one of them, then the relative risks are quite different. If you are as safe a driver as a bus driver, you are still at lower risk in the bus because of vehicle mass (including rollover) and seating position effects (Chapter 4). However, if you are a safer driver than the bus driver, you may be safer travelling in your car than by bus, especially as the bus trip necessarily involves some additional travel. As in the study by Evans, Frick, and Schwing , the wide variation in driving risk should be kept in mind in comparing travel risks in different modes. In the words of old adage noted by Feinstein , "Statistics are like a bikini bathing suit: what is revealed is interesting; what is concealed is crucial."
TRAFFIC SAFETY IN THE CONTEXT OF MOBILITY
Traffic safety is all too often discussed as if the only goal in
creating traffic systems was safety [Haight 1985]. On the contrary, the goal is mobility;
crashes are an unwanted by-product to be minimized in the context of this
primary goal, not by abandoning it. While
the goals of safety and mobility are often in conflict, this is not always the
case. It is helpful to discuss
safety changes with regard to whether they reduce mobility, do not appreciably
affect mobility, or actually increase mobility.
Such categorization does not imply anything about the relative
desirability of different interventions.
An intervention which reduces mobility, such as drunk driving laws,
might be more desirable than an intervention which
increases both mobility and safety, such as building freeways. In the discussion we ignore the effect of crashes, as such, on mobility. Thus, we assume that decreasing speed decreases mobility without taking into account the additional mobility enjoyed by those not killed or injured as a result of the lower speeds.
Safety measures which reduce mobility
Safety measures which reduce mobility include:
Drunk driving laws
The safety intervention which is most in conflict with mobility is speed regulation. Eqns 6-1 to 6-3 show that crash risk, and the probability of injury and death in a crash, increase steeply with speed. There is no reason to think that similar relationships do not apply at all speeds. Therefore, the safest speed is zero. This safest speed would indeed eliminate all harm from traffic crashes, and also eliminate all traffic. As expected harm increases steeply with increasing speed, whereas mobility increases more slowly, optimum speeds exist which can be estimated in terms of various assumptions. Miller  performs such an examination for the US rural Interstate system and finds that the 55 mph speed limit may be close to the optimum choice. He further estimates that the higher 65 mph may not even save time when all the delays associated with injuries and crashes are added to the years of life lost through fatalities.
Driver licensing reduces the mobility of some individuals by denying them the right to drive. The main reasons for such denials are age (per se, before a specified age, but by less straightforward means for older drivers), performance impairment (for example, blindness), or criminal sanction. There are many studies which show that denying some group of individuals (usually the young) the right to drive reduces crashes. Successfully prohibiting any fraction of the population from driving is naturally expected to reduce crashes. The percent reduction in crashes will exceed the percent reduction in drivers if those prevented from driving have above average crash rates. However, this in itself does not justify denying driving privileges to a group of people. If it did, the logical consequence would be to prohibit males from driving (Fig. 2-7). If one had sufficiently detailed knowledge about all drivers, successive application of such a philosophy would eventually eliminate all except the one safest driver.
Drunk driving laws do not have the catastrophic effect on mobility which speed or driver licensing do when taken to hypothetical limits. They do, however, still inhibit the mobility of those who would otherwise drive and drink. Although increasing the minimum drinking age does not affect mobility as such, it does amount to a "selective prohibition", in which the majority musters enough political muscle to force on a relatively weak political minority a policy which they would not accept for themselves, notwithstanding its potential contribution to traffic safety. The majority are presently unwilling to pass other laws relating to alcohol which are expected to reduce traffic crashes. In the case of alcohol use, as for driver licensing and speed laws, it is the legitimate role of the political process to balance the goals of mobility and safety when they are in conflict.
Safety measures which have little or no effect on mobility
Measures which have little or no effect on mobility include:
Active occupant protection devices
Passive occupant protection devices
Vehicle safety improvements
Safer highway furniture (break-away signs, etc.)
Improved emergency medicine
Passengers sitting facing rearwards
Passengers selecting rear rather than front seats
While passive occupant protection devices such as airbags do not influence mobility, it could be argued that devices which require action on the part of the occupant do. Fastening and unfastening safety belts does take some time, as does the deployment of motorized belts. These delays have such trivial effects compared to, say, a change in speed limit, or modest changes in chosen speed, that it seems more appropriate to ignore them than to think of them as realistic impediments to mobility. Motorcycle helmets might have a sufficiently larger effect on mobility to justify inclusion in the mobility reducing category. Not only do they take longer to put on, but they might be stored separately from the vehicle. If required by law, trips could be cancelled if helmets were lost, stolen, or not available for a passenger.
In Chapter 11 it is concluded that devices which reduce injury given
that a crash occurs, but which are relatively invisible to the road user, have
little or no influence on behavior. As
all the other items listed above fit this definition, they are considered
unlikely to influence mobility. Facing
rearward, or sitting in a rear compared to in a front seat, are choices
related to convenience, comfort, pleasure, etc., under which heading they are discussed later.
Safety measures which can increase mobility
Safety measures which can increase mobility include:
Upgrading roads (for example, replacing two-lane by freeway)
Improved brakes, tires, headlights, and other technical equipment
Improved vehicle performance and handling
New technological approaches to night vision, drowsiness
The largest changes in safety associated with technical differences are those associated with different types of roads (Table 4-4). Thus upgrading a rural two-lane roadway to a divided freeway is expected to reduce crash, injury, and fatality risk by large amounts. Upgrading roads is accompanied by reduced delays due to congestion and traffic control devices, and by higher speed limits. Thus, upgrading roads sharply increases both safety and mobility. Mobility is sufficiently increased that a natural consequence is the generation of more traffic [Mackie and Bonsall 1989] which will discount some of the expected safety increases.
Insofar as improved brakes, tires, headlights, and other changes in
technical equipment are used to increase mobility, their safety-increasing
influence is discounted (Chapter 11). Increased
acceleration reduces the time taken to reach cruising speed after stops, and
reduces the time to merge into freeway traffic.
It also increases overtaking opportunities. Unquestionably,
for a fixed level of safety, the greater the performance characteristics, in terms of acceleration, handling and braking, the shorter is the trip time.
Devices being developed to provide the driver with a view of what is ahead under extreme visual conditions, such as fog, have the potential to increase mobility; similarly, devices to alert the driver of the onset of sleep. While it is not possible to apportion benefits between safety and mobility without actual experience with the devices, it does seem probable that they would permit more exhausted drivers to drive in worse weather, leading to benefits more in the realm of increased mobility than increased safety.
TRAFFIC SAFETY IN THE CONTEXT OF OTHER VALUES
While most discussions of factors which are in conflict with safety focus on trade-offs with mobility, this is by no means the only goal which may clash with safety.
Traffic laws regulating behavior by such measures as speed limits and
requiring the use of restraints (belts, helmets) have been criticized as
diminishing personal freedom. Such issues tend to arise more in prospect than in
retrospect. Little concern is
expressed because drivers are legally prohibited from driving on whatever side
of the road they wish. The rule,
rigorously enforced by custom and law, that all motorists drive on the same
side of the roadway expands the freedom of mobility for all. That we all drive on one side of the road is a particularly
interesting case, because the choice is arbitrary, as there are no convincing
reasons for choosing one side over the other.
Some nations choose the left, some the right, but there are
none which have opted to allow their citizens to choose freely for themselves each time they drive. Having to stop at a red traffic light is a similar apparent impediment to freedom, yet traffic signals increase mobility.
The argument that mandatory restraint use laws are of a different nature has more merit, in that they affect the safety primarily of only the involved individual. There are some secondary effects. Unbelted rear occupants pose increased threats to front-seat occupants. Unbelted front-seat occupants pose increased threats to other front-seat occupants in side-impact crashes. It has been suggested that belted drivers may be able to control their vehicles better when crashes are imminent, but there is no supporting evidence. Apart from these considerations, the use of a belt or helmet affects mainly the user's risk. However, the non use of restraints imposes additional economic burdens on society. It could even be argued that unrestrained occupants increase the fatality and serious injury risks faced by restrained occupants by diverting away from them limited emergency medical resources. There does not appear to be any feasible mechanism by which the increased economic burden the unrestrained place on the restrained could be avoided. It is not conceivable, apart from any questions of desirability, that any modern society would recognize a contract by which a motorcyclist would sign away his right to medical care in return for permission to ride without a helmet. It seems to me that the society that pays the medical bill does have some legitimate right to take reasonable measures to reduce that bill. Each such issue is intrinsically political in nature; the only way to decide whether to require helmets or safety belts, impose speed limits, ban heroin or skiing is through the political process. The argument that it is nobody's business seems to me to have lost credibility as society becomes increasingly intertwined.
The claim that motorists have a right to own and operate devices to
warn them of the proximity of police radar speed monitoring seems devoid of
the valid points which can be raised in favor of allowing occupants to please themselves about restraints. The only purpose of such devices is to facilitate breaking laws which are (formally, at least) supported by the majority. The manufacture, sale, and use of such devices seems to me about as legitimate as the manufacture, sale and use of burglary tools.
Unlike refusing to wear restraints, there is no question that speeding kills road-users other than the speeder. Radar detectors are advertised in airline magazines, presumably because airline passengers desire to save time, and can afford such devices. I wonder how these purchasers, whose time seems so precious, would react if the pilot of their plane walked on board surreptitiously carrying an electronic black box which facilitated circumventing government regulations promulgated to increase safety, but which increase flight time as an unavoidable by-product. The pilot could explain that the flight would of course be safer, because monitoring the additional sensor would prevent boredom, and the exhilaration of the sport would ensure high adrenaline levels. I suspect that the speeding driver, who seems unconcerned at putting other road users at risk because of his or her own law breaking, may not necessarily find this make-believe situation so appealing.
Adams [1981; 1985] discusses many ways in which increased mobility for
vehicle occupants may be in conflict with other human values. One of these is the relative emphasis
placed on the safety and freedom of different categories of road user. In the sense in which I have been
using safety in this book, a reduction in pedestrian fatalities is interpreted
as an increase in safety. However,
such a reduction may be obtained by preventing children from even
crossing roads in which they may have played with relative safety in earlier decades. Indeed, the main mechanism which reduces the number of children playing in the road is the very increase in danger brought about by increased traffic and increased speeds. It is likely another example of a perverse effect in which factors adverse to safety generate safety increases because of large human behavior responses on the part of the children and their parents. While increasing motorization expands the mobility of those who can drive, it stimulates other processes which tend to diminish the mobility of those who cannot, mainly the old, those too young to have licenses but old enough to seek independence from adult control, and those with physical handicaps.
There are other approaches, such as separating pedestrians and vehicles, and building limited-access freeways, which enhance safety and mobility but may diminish a sense of community and interaction between people. All these considerations place such decisions properly in the political process; they are not to be decided by technocrats with narrow goals, even those as laudable as saving lives and reducing injury. The goal of traffic safety research is to inject into the discussions objective estimates of the most likely safety effects. Safety research should not be a lobby for adopting safety measures, but for providing knowledge.
Convenience, comfort, pleasure, etc.
Because frontal impact is the most common crash impact direction,
injury risk to passengers would be less if they were seated in appropriately
designed rearward-facing seats. Such
a change has little direct effect on mobility, except insofar as it could
affect navigation and also remove the benefit of other pairs of eyes looking
for danger; such considerations have little applicability to rear-seat
passengers. The desire for
human interaction, and
the greater pleasure of looking forward overwhelm an otherwise substantial safety increase available, in principle, at essentially no monetary cost. Rearward-facing seats in trains and aircraft, despite their increased safety, have never proved popular.
Passengers can presently choose between seats with substantial differences in fatality risk. A passenger who does not wear safety belts and makes the normal choice of sitting in the right-front seat, rather than an outboard-rear seat, thereby increases fatality risk by a substantial 35%. Even if the front seat is protected by an airbag, it still has an 18% higher fatality risk than the rear seat (Fig. 3-5 and Chapter 9). If the passenger uses belts, and there is a lap/shoulder system in the front and a lap-belt only in the rear, then each seat has a similar fatality risk. If both front and rear seats have lap/shoulder belts, the rear is expected to have lower fatality risk. Notwithstanding the greater safety of rear seats, passengers are likely to continue to chose the front because of such considerations as closer interaction with the driver and a better view.
Devices like radios and telephones may compromise safety, though each of these may also contribute to safety under particular circumstances. As in the discussion on freedom, concerns seem much more intense in prospect than in retrospect. While nearly all cars have radios, which undoubtedly distract drivers to some extent, there is little desire to prohibit them, or restrict their use to, say, when the vehicle is not in motion.
Any cursory examination of the success of those making predictions invites the following advice:
Never make predictions -- especially about the future.
(Such advice assumes a goal of accuracy. If, however, the goal is to become rich and famous, then different advice applies: make lots of predictions, the more unconstrained, outrageous, and alarmist the better, and keep in mind that the market for bad news far exceeds that for good news.) Notwithstanding the pitfalls surrounding any attempt to estimate what will happen in the future, I offer below a few thoughts on where I think traffic and traffic safety are heading.
The future of traffic
The field of transportation has had many interesting past predictions. In 1979 a group of experts
predicted the US price of unleaded gasoline in 1990 using a
computer-interactive approach called the Delphi method [UMTRI Research Review
1987]. Their estimate, over $4.00
per gallon in 1990 dollars, is 300% higher than the 92.9 cents to $1.079 per
gallon I have paid from January to June in 1990! Macrae [1988, p. 18] reports that, in 1903, Mercedes thought
there would never be a world market for more than one million automobiles,
because there were not one million artisans in the world trainable as
chauffeurs (also quoted in slightly different form by Mackay ). Between then and now there have been a
succession of predictions of saturation levels of vehicle ownership that have
been overwhelmed by a reality reflected in a world now containing about 700
million vehicles [Mackay 1990]. Many
mechanisms in addition to the difficulty of training artisans have been
invoked to infer asymptotic saturation levels; these have included limited
amounts of roadway, economic constraints, energy constraints, and natural
limits, such as one vehicle per family. Recently
Haight , Mackay
, and Lave  have included in their thinking saturation levels of vehicle ownership similar to present US levels, with the implication that motorization in the US has saturated, and anywhere else approaching the same level will also saturate.
The upper graph in Fig. 14.5 shows the growth of all vehicles, and of privately owned cars, in the US since the beginning of the century. Because growth in the human population contributes to the growth of vehicles, it is more illuminating to examine the growth in vehicles per capita, which is shown as the lower graph in Fig. 14-5. Apart from declines during the depression and the Second World War, the curves show increases almost every year. There is no hint of a kink corresponding to one vehicle per family, or one privately-owned car per family. The curves evince no indication of becoming horizontal, nor does the value one appear to constitute any particular impenetrable membrane.
Fig. 14-5 about here
One interesting attempt to address saturation phenomena in general is
that of Marchetti  who finds that a three parameter equation identical
to one describing the time dependence of many ecological systems can be
applied to the evolution of such manufactured items as telegraph wire and
roads. One of the parameters
derived from historical time series is an estimate of a future saturation
level. In applying the approach
to cars, Marchetti  estimates a saturation level of 200 million for the
US (the 1988 total is 139.5 million). However, he points out that the goodness of fit to the
equation is relatively insensitive to the choice of final saturation level. Because a simple linear equation, with
only two parameters, fits the US data
substantially better than the more complex three parameter equation, the third parameter, the estimated saturation level, is of questionable validity.
It seems to me there is no more a "natural" limit on vehicle ownership than there is a natural limit on the ownership of radios, televisions, bathrooms, shoes, or houses. In the early days of radio, one radio per family might have seemed a natural limit. A stronger case could have been made for one per person, given that we each have only one pair of ears. The number of products of any type sold depends on such factors as how much they are desired, how much they cost, how conveniently they can be stored, and how much money people have; it cannot be inferred from an abstract general principle. I see no reason why this should not also be the case for vehicle ownership. US evidence shows clearly that there is nothing special about one vehicle per family. There is likewise nothing magical about an average of one vehicle per person -- many families now possess more than one vehicle per family member. My guess is that US vehicle ownership per capita will continue to grow without any natural upper bound, although in future decades the rate of growth might slow down. If history is any guide, motorization in other countries will tend to follow the US pattern. Vehicle registrations in the US are presently increasing at about two and a half million per year.
The total distance of travel, which is the denominator in the fatality rates plotted in, for example, Fig. 13-2, is the number of vehicles times the average distance of travel per vehicle. This (Fig. 14-6) has been remarkably stable at close to 16 000 km/year since World War II, although there is an indication of an increasing trend in the late 1980's. Figs 14-5 and 14-6 suggest ongoing increases in total US travel.
Fig. 14-6 about here
In the mid 1970's there were dire predictions that the world's energy
resources could not support increasing motorization, and that by now
catastrophe would long since have overwhelmed us. While we know that petroleum resources
are limited, as is the life of the earth and of the solar system, the earlier
noted present price of gasoline (in real terms less than in 1950) speaks
eloquently to the immediacy of this problem. The 1970's "energy crisis" has been replaced by the
1990's "congestion crisis", the concern that all these vehicles are
going to be locked in "gridlock", meaning immobilized in some
gigantic traffic jam. Such
concerns yet again ignore the enormous feedback that characterizes large
complex social structures. Congestion
today is not demonstrably worse in New York City than it was a hundred years
ago, nor worse in London than three hundred years ago, nor worse in Rome than
two thousand years ago. Congestion
is caused by people's desire to have access to amenities that require large
concentrated populations; it is regulated by how much aggravation they are
willing to pay for such benefits, not by the number of vehicles. When congestion reaches a certain level of unpleasantness, a
natural balance arises between those willing and those not willing to suffer
it. The tolerable level does not
seem to have changed much over the centuries, or for that matter the,
millennia. It is based on the
number of hours in the day, and the desire of humans to use their time in
productive and pleasant ways. I
see no reason why it should change in future decades. Places which are already as congested
as the human spirit can bear, and there are a great number of them, will not
become more congested. Because all market equilibrium processes are imperfect and
have time lags, there may be some fluctuations above and below what people
will tolerate; however, any net increase in congestion in, say, the world's
100 most congested cities seems implausible, although the fraction of the city
subject to congestion could increase. What
is much more likely is that more cities will join the
already long list of those that have reached just bearable levels of congestion. Although congestion impedes mobility, it increases safety, as measured by serious injuries and fatalities.
When land is available, road-building is the most effective way to reduce congestion, although the increased mobility thereby provided likely attracts additional traffic. Even when land is unavailable, building double-decker roads could be economically attractive in some circumstances. Although road-building has become politically unpopular in recent years, I suspect that this will change when the effectiveness of other approaches, such as electronic measures to better control traffic, are found to be modest and far less cost-effective than building roads.
Most of the increase in traffic will occur in areas which are presently less congested than the maximum tolerable level. In the US context, when life in California becomes a little less attractive because of increased congestion and other ills associated with high population density, the population of that state may stabilize in favor of growth in such less populated states as Michigan and Ohio. Even many countries with much higher population densities than the US still have large uncongested areas; in many cases the reason why people will not move to them is that they contain too few people!
The future of traffic safety
Increasing motorization has led to expectations of ever increasing
numbers of traffic fatalities. In
1975, when US traffic fatalities were 44 525, the National Highway Traffic
Safety Administration  estimated that traffic fatalities in 1985, one
decade later, would be 72 300. The
actual number turned out to be 43 825. In
other words, a 62% increase over a 10 year period was forecast, compared to
the 2% decline that actually occurred. While
 comments on many of the details of the calculation, it seems to me that details are rarely crucial in making predictions. In the face of high uncertainty, adding complexity generally conceals the truly crucial one or two assumptions, which often, out of necessity, are little more than educated guesses. It seems particularly inappropriate to invoke computers in attempts to predict the future, as if the ability to do lots of multiplications quickly (about the only thing a computer does) is of any more relevance than a crystal ball. The most one can hope to do is have some broad understanding of the processes, and some general view of some factors which exhibit some degree of historical stability.
The decline in fatalities per unit distance of travel (Figs 13-2 and 13-3) shows long term stability. Assuming that the logarithm of the rate declines linearly with time would rarely have generated large errors. Broughton  finds a similarly scatter-free relation for British data. Trinca et al.  write, "In the most motorized countries of the world, the traffic safety rate is already nearly stabilized, so that even if percentage improvement in rate is forthcoming, the possibility for decline in absolute numbers is limited." [Trinca et al. 1988, p. 30] The fatality rates in Figs 13-1 and 13-2, and those reported by Broughton  show no stabilizing; the simplest interpretation is of a constant percent decline per year. If the fatality rate declines faster than the distance of travel increases, then a reduction in fatalities necessarily follows; if the human population increases, then fatalities per capita necessarily declines more. Thus, although Trinca et al.  in the above quotation and elsewhere appear to suggest otherwise, reductions in fatalities can occur in motorized countries even as mobility increases. In the US distance of travel doubled in the two decades from 1968 to 1988, yet fatalities in 1988 were 5000 less than in 1968.
Hutchinson [1987, p. 6] reproduces the following 1979 statement from the World Health Organization: "Little change may be expected in road accident rates within the next 5-10 years because alterations in the behaviour of road users and improvements in the environmental infrastructure of roads can come about only gradually." Hutchinson writes, "Happily, this prediction has turned out to be overly pessimistic," and points out that 22 of 28 jurisdictions for which he displays data show a decline in traffic deaths per million population from 1980 to 1985, with the average change being a 14% decline in the five year period.
While total distance of travel in the US increased at an average rate of 2.85% per annum from 1979 to 1988, the fatality rate (fatalities per unit distance of travel) decreased at 3.69% per annum in the same ten-year period. The slightly greater decline in the fatality rate compared to the increase in the distance of travel suggests that total fatalities are more likely to drift downwards than upwards. However, fluctuations depending on economic conditions are to be expected, as discussed in Chapter 11. In the absence of any dramatic discontinuity, such as occurred in October 1973, and based on the 1989 preliminary total of about 45 500, I would expect US fatalities to remain in the range 40 000 to 51 000 throughout the remainder of the twentieth century.
The above discussion, in focusing on broad trends, has ignored details. However, details are of the utmost importance, and it is the combined effects of many details that generate the broad trends. Interventions that change fatalities by 1% are not going to be visible in the broad trend. However, such an intervention prevents over 400 deaths annually in the US. The extent to which reductions beyond the general trend are achieved depends crucially on better understanding and knowledge; in other words, on research in traffic safety.
THE STATE OF TRAFFIC SAFETY RESEARCH
In the more than 50 years since driving behavior and traffic safety were first analyzed in a technical way, much has been learned. However, when compared to advances in the traditional sciences, increases in knowledge about traffic safety are less impressive. Below I offer some thoughts on why this is so, and offer suggestions for changes, building upon comments in three previous articles [Evans 1985; 1988b; 1988c]. I hope that my occasional oversimplification in the interests of brevity and clarity will not be mistaken for naivety.
The comments focus mainly on hopes that traffic safety research might in the future acquire more of the method, style, values, attitudes, and institutional structures which have proved so successful in the traditional sciences, and that in the decades ahead understanding might increase at a greater rate than in the past. In advocating the application to traffic safety research of what might be called the normative model of science, I am not claiming that this model is in fact all that closely followed, even in the physical sciences. Even ignoring outright fraud, which Broad and Wade  indicate is more common than generally recognized, the normative model of science is one which is subject to many criticisms [Kuhn 1970; Feyerabend 1975], being rarely the model used in the discovery process, but more likely used to organize events after they have occurred. My claim is not so much that traditional science really follows this normative model, but that it is an understandable ideal standard. I believe that activities that have such an idealized goal will make more progress than those that do not, because possession of such a standard enables more effective evaluation of contributions.
The goal is knowledge
The primary motivation in science is curiosity, a desire to know. All too often in traffic safety there appears to be insufficient recognition of the difference between knowledge and what might be done with it. When Karl Marx draws the distinction in stating, "Philosophers have merely interpreted the world; the point, however, is to change it," he is stressing action over understanding. The goal of science is understanding; hopefully, increased understanding will illuminate the process of making changes.
The scientific community should be dispassionately knowledgeable about, for example, the effect of alcohol on the probability that a driver will crash and on the probability that an intoxicated driver will be arrested. On the other hand, those pursuing the policy goal of trying to reduce harm from drunk driving may legitimately publicize the first of these items more than the second. It would be counterproductive to their goal to publicize that the probability of arrest for this offence is typically about one in a thousand, and that doubling police enforcement is consequently expected to increase the chances of arrest to one in five hundred. While the scientist as a human being may indulge in trying to change the world, it is imperative that this entirely different activity be recognized as something separate from science.
Advocacy, by its very nature, implies making as good a case as you can. This involves selecting supporting
evidence, and leaving the task of presenting contrary evidence to opponents. Although it is argued that advocacy
performs some functions well, it is nonetheless a process which is
different from, and indeed inimical to, the normative processes in science. The goal aspired to in science is that all relevant evidence be evaluated in a detached and objective manner, and that the inquirer be "disinterested" in the result.
The question of objectivity is often confounded by questions of organizational affiliation. For good or ill, the days of the gentleman/scholar of independent means are gone; essentially all researchers are paid for their efforts by somebody. To admit this is not to conclude that therefore all research is but a mere affirmation of the interests and beliefs of the body supporting it, and even less to assume that such a model applies to some bodies but not to others. All institutions, including universities (and the government bodies supporting them) and government-supported laboratories develop their own interests. As effectively pointed out Hauer  in his paper "The reign of ignorance in road safety: a case for separating evaluation from implementation", the notion that, say, government bodies charged with evaluating safety countermeasures are focused solely on the public good, without any agenda of their own, does not withstand much scrutiny.
There is little basis for judging the quality or objectivity of
research based on the type of organization in which it was performed. Research is primarily an individual
effort; the individual researchers within any large institution differ in such
characteristics as ability, understanding, compassion and integrity by
enormously more than any possible differences in these same characteristics
between types of institutions. In
this book, quality research from many types of institutions has been cited; in
many cases the same institutions have also supported blatantly self-seeking
and incorrect studies. Research
must be judged by its content, and the reputation -- especially the long term
track record -- of the
researcher. Work with the
outward appearance of being technical is sometimes used for purely advocacy purposes and to buttress ideological beliefs. Some social "science" is little more than political advocacy packaged in unintelligible jargon.
The question "What change in fatalities is associated with the passage of a mandatory belt law in some jurisdiction?" is, in its structure, a simple factual question. Most would agree that there is an objective answer. Although the answer may be difficult, or even impossible, to determine, it does not depend on the belief system, discipline, or training of the inquirer. This is not to suggest that science operates in a social vacuum; complex social factors determine what individuals, if any, are addressing this particular question, what methods they use, and what resources are available. In contrast, the question "Ought we to pass (or more strictly enforce) mandatory belt-wearing laws?" is of a quite different nature, and cannot be answered by science. Questions of this type are properly addressed through the workings of the political process, in which personal philosophies and legally-pursued self-interest are legitimate ingredients. It is not the safety researcher's role to prescribe what is good for society, but to provide reliable information for more informed choices.
Flawed approaches and techniques
Although mentioned briefly in the introduction, the misunderstanding
and misapplication of statistics seems one of the most pervasive flaws in
research on traffic safety (and other fields).
The impression is inescapable that there is a large army of so-called
researchers, and what is much worse, an army of teachers of
researchers, who think the goal is to test if data reject non-quantitative
hypotheses. Unbelievable numbers
of studies end by concluding that A is (or is not) statistically significantly
different from B.
It is difficult to see of what possible use any such conclusion can be, given that, with the possible exception of some elementary particles, all A's are different from all B's. If the conclusion is that A is bigger than B, then this is new information, but of extremely limited value; without knowing anything, such a statement has a 50% chance of being correct. The only answer of any real scientific value is how much bigger is A than B. Comments parallel to the above apply equally if the question is, "Does A affect B?" The answer is yes; everything affects everything. When I spit in the ocean it affects the length of the day. The only question of any real interest is "How much?"
Many papers show detailed results of many statistical tests, but display no raw data; in some cases a table of all the data analyzed would consume less page space than is consumed by the statistical test results. Some claim that applying statistical tests is necessarily more objective than displaying data and making reasonable inferences from them. This point loses much of its impact when one realizes that there is a rich variety of statistical tests and transformations to select from, some combination of which can often produce any desired answer. Most of the key discoveries in science pre-date the field of statistics, which originated mainly in the early decades of the twentieth century. The main uncertainty in traffic safety research is usually interpretation; the focus on statistical detail often obscures this. In many cases there are more than enough data to, say, determine a rate to high precision, but the meaning of the rate relative to traffic safety is subject to various alternative interpretations. Statistical procedures are vital to many studies, especially those involving large numbers of variables, but should not be at the core of all studies involving observational or experimental data.
When questions arise in traffic safety there is often a clarion call to collect more data. Such a call is often without regard to how the sought- after data can be used to address the question. In the era before the birth of experimental science, Greek philosophers thought that nature could be understood by pure thought alone, without the need for data. Nowadays there seem to be people who think that it can be understood with data alone, without the need for thought. Over half a century ago Gibson and Crooks [1938 p. 453] wrote, "Accident statistics are now widely publicized." Basically, traffic safety is one of many fields that can be characterized as data rich, understanding poor. The main thing that has been missing from traffic safety research is the appropriate scientific tradition to extract meaning from copious data that already exist; the answers to many key questions are embedded in existing data. Somewhat similar to the call for more data is the call for better driving simulators, as discussed in Chapter 5.
Flow charts appear constantly in the literature, showing boxes with labels such as "deviation from desired direction" and "driver reaction" connected by lines and arrows, relating actions to inputs, and showing all sorts of feedback. While such diagrams have application to engineering systems for which so-called "transfer functions" are known, I have yet to encounter one in traffic which contained any more information than could have been more succinctly stated in words. It is trivial to produce such flow charts if there is no requirement that we know anything about any of the ingredients (including whether or not they exist), or what observed output is being related to what observed input. In a minute one could produce such a chart with any arbitrary level of detail on, say, how one adds sugar to one's coffee.
Creating a scientific literature
All the traditional sciences accumulate knowledge in a clearly identifiable peer-reviewed, or refereed, literature. It is even more crucial for peer-reviewed literature to be at the core of a subject such as traffic safety research, where so much is written for so many diverse purposes. Authors should feel reasonably comfortable (but not complacent) quoting results from such literature without the need to review all papers in detail. To require that all papers be read critically (which would involve similarly examining the work they cite, and so on) before citing would largely deny the field the possibility of advancing much beyond what one human mind can encompass. Work performed too recently to be published in a peer-reviewed article makes a strong claim to be cited because of its currency; however, authors have a greater responsibility to examine unrefereed reports before using them, or even providing them the increased credibility of a citation.
Increasing the importance of peer-reviewed literature is the most effective way to discard the plethora of non-scientific "results" which overwhelm this field. The value of many papers is highly negative; not only do they spread misinformation, but they may oblige competent researchers to squander their time refuting nonsense. Many deserve to be handled with the clarity with which the great physicist Wolfgang Pauli once dealt with a particularly poor paper; he said, "That paper isn't even good enough to be wrong." [Segal 1990].
My advocacy that all published research results be archived in
peer-reviewed literature does not imply that work in peer-reviewed journals is
necessarily of high quality, or even free from grievous error arising from
incompetence, or worse. The
peer-review process is subject to all the frailties to which humans succumb. It is still surprising that the
Journal of Public Health, a peer-reviewed journal, has consistently opposed treating traffic crashes as a public health problem, and has instead advocated that the only countermeasures worthy of consideration are engineering changes to vehicles. This position has been pursued by publishing many conspicuously incorrect "technical" papers masquerading as science, while at the same offering spurious reasons ("subject not important") for refusing to publish technically correct work demonstrating the roles of non-engineering factors, such as alcohol; the journal is, however, willing to publish papers on behavioral interventions provided they "prove" that they do not work! The editor of "Science", the peer-reviewed journal of the American Association for the Advancement of Science, declines to publish any research on traffic safety, yet uses its pages for his own uninformed opinions on the subject; he seems unaware of any relationship between car size and safety [Koshland 1989a]. Hamm [1989; 1990] takes him to task for writing, "Motor vehicle accidents, which kill more people each year than were killed during all the Vietnam War," [Koshland 1989b] by pointing out that an estimated 1.3 million people were killed in that war, 56 000 of them Americans. Even if interpreted to apply only to Americans, the statement is still false; in none of the seven years prior to his writing did traffic fatalities exceed 48 000, which falls short of being larger than 56 000!
Stressing the importance of reviewed literature is not to deny that many important contributions have appeared in non-refereed sources. Some unrefereed literature is very good, and some peer-reviewed literature is very bad, but the average quality, importance, objectivity, and technical correctness of the peer-reviewed literature is substantially higher.
It seems to me that we should strive for the eventual goal that all
work more than a few years old that has not been accepted in a peer-reviewed
journal should be ignored. Meeting
proceedings, such as the many quoted here,
and papers from SAE meetings should be regarded as texts of oral presentations. It is common in the physical sciences for the proceedings of a meeting to be specifically annotated that it should not be cited. Unconstrained oral presentations and discussion should be encouraged as essential for scientific exchange, and a written record can facilitate this process. However, the written record of such activities should not be considered part of the scientific literature. Publication and presentation should be perceived as separate processes. One of the most regrettable practices is that some journals that review general articles publish special issues devoted to papers presented at a conference; the unwary reader may mistakenly think that such papers are reviewed. Peer review means that each paper is reviewed in as much time as it takes (and usually a whole lot more!); it is not possible to review papers for inclusion in a meeting with scheduling constraints. Individual researchers anonymously review papers; it cannot be done by committees.
Science (and scholarship in general) advances mainly because
individuals devote whole careers to fairly narrow specialized areas. Traffic safety has not acquired this
tradition. Although there are
hundreds of traffic safety researchers world-wide, there is not one who has
devoted a substantial portion of a career to such clearly defined areas as,
say, the influence of speed, or the influence of vehicle mass or size. Yet these are areas of substantial
intellectual challenge which could be advanced by applying the methods of
science to organize and interpret diverse evidence available in many
countries, and accumulated over decades.
Greater understanding of such areas is crucial to traffic safety. Such understanding will advance much
rapidly under the sustained effort and intense focus of a few researchers who are attracted by it than by generalists picking it up from time to time, as is our current practice.
Governments usually want research that addresses some current crisis, or is "important." Both criteria are flawed. Even if the research is good, by the time it is completed another crisis is in center stage. Importance is not a reason for researching some area, as would be readily realized if the subject were not traffic safety. The most important medical problems are probably that people don't feel good, and they die, yet I am not aware of any project specifically aimed at developing a feel-good pill or an immortality pill. The edifice of science is built from a large number of specific results. Physics does not advance by lots of people addressing such broad questions as what is the nature of matter, but by intense specialization. The choice of which research to undertake is not based on the importance of the problem, but the product of importance times the probability of making useful progress. This is obviously the most difficult problem to address as it is steeped in high uncertainty, but the researcher embedded in a subject is far more likely to make a wise choice than someone less knowledgeable.
Many of the reasons that have prevented traffic safety research from
acquiring more of the values and methods of science are related to the
institutional framework in which it is generally pursued. Much of the work is based on a
government defining a problem and offering contracts for institutions to
complete specific projects. Institutions
grow up dependent on such contracts. The
professionals employed in them spend a good deal of their time preparing
material to bid for contracts. If
successful, there will be a
fixed sum of money and a specified time for completion. The whole thing about research is that you are supposed to be finding out things not previously known. This is difficult enough under the best of conditions, but near impossible to do according to a schedule. Even if the people doing the work are competent to do it, they must share their time between completing the present contract and obtaining the next one. Given that researchers must eat, it is perhaps excusable that the only "conclusion" in some such studies is that more research is necessary. No scientific study would ever come to so vacuous a conclusion, which was obviously known before starting; it implies that nothing was learned in the study, which is often correct but rarely stated.
The situation described above could be ameliorated by creating a half dozen or so well-endowed professorial chairs sprinkled among distinguished universities throughout the world. Appointments to such positions would be based on distinguished careers doing quantitative research, and the holders would be free to pursue safety as a science, letting their curiosity determine their research agenda. Their next year's resources would not be contingent on pleasing somebody by this year's offerings. They could therefore provide more detached council on safety questions than is presently available to policy-making bodies. They could also offer that sense of historical perspective that is presently often lacking, and help policy-makers avoid travelling down identical blind alleys every ten years or so.
The cost of such an undertaking would be only a modest portion of what is currently spent. While it could not bear fruit for at least five years (more likely ten, and possibly never), such an investment has a far greater chance of increasing traffic safety knowledge than spending enormously larger amounts on the types of things that have been supported traditionally.
REFERENCES (CHAPTER 14)
Adams, J.G.U. Transportation planning: vision and practice. London, UK: Routledge and Kegan Paul; 1981.
Adams, J.G.U. Risk and freedom - the record of road safety regulations. Nottingham, UK: Bottesford Press; 1985.
Broad, W.; Wade, N. Betrayers of the truth. New York, NY: Simon and Schuster; 1983.
Broughton, J. Predictive models of road accident fatalities. Traffic Engineering and Control 29:296-300; 1988.
Cerrelli, E. Older drivers, the age factor in traffic safety. Washington, DC: National Highway Traffic Safety Administration, report DOT HS 807 402; February 1989.
Evans, L. Post symposium reflections (comments on problems and issues in traffic safety research). In: Evans, L.; Schwing, R.C., editors. Human behavior and traffic safety. New York, NY: Plenum Press, p. 525-529; 1985.
Evans, L. Older driver involvement in fatal and severe traffic crashes. Journal of Gerontology: Social Sciences 43:S186-S193; 1988a.
Evans, L. The science of traffic safety. Physics Teacher 26:426-431; 1988b.
Evans, L. Comments (on two papers on mandatory safety belt use laws, and reflections on broader issues). In: Graham, J.D., editor. Preventing automobile injury -- new findings from evaluation research. Dover, MA: Auburn House, p. 73-83; 1988c.
Evans, L.; Blumenfeld, D.E. Car occupant life expectancy: car mass and seat belt effects. Risk Analysis 2:259-268; 1982.
Evans, L.; Frick, M.C.; Schwing, R.C. Is it safer to fly or drive? -- a problem in risk communication. Risk Analysis 10:239-246; 1990.
Feinstein, A.R. Scientific standards in epidemiologic studies of the menace of daily life. Science 242:1257-1263; 1988.
Feyerabend, P.K. Against method: outline of an anarchistic theory of knowledge. Atlantic Highlands, NJ: Humanities Press; 1975.
Gibson, J.J.; Crooks, L.E. A theoretical field-analysis of automobile driving. American Journal of Psychology 5l:453-47l; 1938.
Haight, F. A. The place of safety research in transportation research. Transportation Research 19A:373-376; 1985.
Haight, F.A. International cooperation in traffic safety. IATSS Research -- Journal of International Association of Traffic and Safety Sciences 11:11-16; 1987.
Hamm, R.M. Statistical mortality (Letter to the Editor). Science 245:111-111; 1989.
Hamm, R.M. Cognitive impediments to a statistical consideration of war's consequences. Boulder, CO: Institute of Cognitive Science, U. of Colorado, technical report #90-3; 1990.
Hauer, E. The reign of ignorance in road safety: a case for separating evaluation from implementation. In: Moses, L.N.; Savage, I, editors. Transportation safety in an age of deregulation. Oxford, UK: Oxford University Press, p. 56-69; 1989.
Hutchinson, T.P. Road accident statistics. Adelaide, Australia: Rumsby Scientific Publishing; 1987.
Injury in America: a continuing public health problem. Washington, DC: Committee on Trauma Research, Commission on Life Sciences, National Research Council and Institute of Medicine; 1985.
Kamerud, D.B. Evaluating the new 65 mph speed limit. In: Graham, J.D., editor. Preventing automobile injury -- new findings from evaluation research. Dover, MA: Auburn House, p. 231-256; 1988.
Koshland, D.E., Jr. A tax on sin: the six-cylinder car. Science 243:281-281; 1989a.
Koshland, D.E., Jr. Drunk driving and statistical morality. Science 244:513-513; 1989b.
Kuhn, T.S. The structure of scientific revolutions. Chicago, IL: University of Chicago Press; 1970.
Lave, C. Things won't get a lot worse: the future of U.S. traffic congestion. Paper 890603, presented to the 69th Annual Meeting of the Transportation Research Board, Washington DC; 7-11 January 1990.
Mackay, M. Towards a unified traffic science. IATSS Research -- Journal of International Association of Traffic and Safety Sciences 14(?):??-??;1990. (in press)
Mackie, P.J; Bonsall, P.W. Traveller response to road improvements: implications for user benefits. Traffic Engineering and Control 29:411-416; 1989.
Macrae, N. The next ages of man. London, UK: Economist, p. 5-20; 24 December 1988.
Marchetti, C. Infrastructures for movement. Technological Forecasting and Social Change 32:373-393; 1987.
Marchetti, C. The automobile in a system context: the past 80 years and the next 20 years. Technological Forecasting and Social Change 23:3-23; 1983.
Miller, T.R. 65 mph: does it save time. Association for the Advancement of Automotive Medicine, 33rd Annual Proceedings, Baltimore MD, p. 73-90; 2-4 October 1989.
National Center for Health Statistics. Monthly vital statistics report. Washington, DC: US Department of Health and Human Services, vol. 38, no. 5; 26 September 1989.
National Highway Traffic Safety Administration. National highway safety forecast and assessment: a 1985 traffic safety setting. Washington, DC: Department of Transportation; 1975.
National Safety Council. Accident facts. Chicago, IL; 1989 Edition.
Segal, I.E. The unscientific charm of the big bang. Letter to the Editor, New York Times; 4 May 1990.
Sivak, M. A 1975 forecast of the 1985 traffic safety situation: what did we learn from an inaccurate forecast? In: Rothergatter, J.A.; de Bruin, R.A., editors. Road users and traffic safety. Assen/Maastricht, Netherlands: Van Gorcum, p. 13-25; 1987
Trinca, G.W.; Johnston, I.R.; Campbell, B.J.; Haight, F.A.; Knight, P.R.; Mackay, G.M.; McLean, A.J.; Petrucelli, E. Reducing traffic injury -- a global challenge. Melbourne, Australia: A.H. Massina; 1988.
UMTRI Research Review. Forecast and analysis of the U.S. automotive industry through 1995 - Delphi IV. University of Michigan, Ann Arbor, MI; March-April 1987.