The unformatted htm text below is from
|Copyright © 2004 by Leonard Evans|
|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|
5 Environment, roadway, and vehicle
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.
A new car with antilock brakes, a clear blue sky, dry roadway, no traffic - close to most drivers' notion of ideal driving conditions. It is commonly assumed that the most desirable driving conditions are also the safest. In this chapter we will find that this is often not the case.
The vast majority (84%) of fatal crashes occur on dry roads (Table 5-1). Roadway surface condition and atmospheric conditions are highly related, but not identical. The roadway surface is wet after rain has stopped. It is, in principle, possible to drive in the rain while the roadway surface is dry - but not for long. The data in Table 5-2 compliment those in Table 5-1, showing that an even larger proportion (88%) of fatal crashes occurs under no adverse atmospheric conditions than occurs on dry roads.
Table 5-1. Percent of fatal crashes occurring under different roadway surface conditions.
Effect of snow
Figure 3-18 (p. 57) showed that February consistently had fewer fatal crashes than any other month, while the largest monthly totals were in summer. As weather seems a likely contributor to this, we seek to examine how the variation in fatal crashes differs between states that experience a lot of snow and states that have no snow.
The monthly snowfall in different states is an unsatisfactory measure to investigate how snow affects crash risk because weather conditions averaged over a state may not, even approximately, reflect weather while driving. For example, isolated mountains may receive much snow while little falls on roads.
The following approach was used to investigate the influence of weather. All 50 US states were rank-ordered according to the percent of all their fatal crashes that took place when the roadway surface condition was coded as snow, slush, or ice. Four states (Florida, Georgia, Hawaii, and South Carolina) had no fatal crashes under such roadway conditions in 2001. Seven states (Alaska, Maine, Minnesota, Montana, North Dakota, Vermont, and Wyoming) had more than 10% of their fatal crashes under such roadway conditions. We designate the two groups of states as snow-free states and snow states.
Effect of snow on fatal crashes per day. The variation in the number of fatal crashes per day in each month is plotted in Fig. 5-1 for the snow-free and snow states. The values are relative to the number of fatal crashes averaged over the entire year. If crashes were a Poisson process with a daily rate that was constant throughout the year, the data would distribute randomly around the value one, indicated by the broken line. For the snow-free states this is approximately what is observed. The average rate of crashes per day in any month does not depart from the yearly average by more than 9%.
The pattern is markedly different for the snow states, in which lower rates are observed in the winter (which requires higher rates in the summer so that the overall average is close to one). The February rate is 43% below the yearly average, while the July and August rates are 38% above the yearly average. The February rate is well under half of the July and August rates.
Effect of snow on crashes for same travel distance. Complete data giving the distance of vehicle travel by month in each of the states are not available. What is available are estimates of vehicle travel on rural arterials. This shows substantially larger reductions in travel in the winter months in the snow states than in the snow-free states, so that a portion of the difference in Fig. 5-1 is because of reduced travel due in part to actual or expected unfavorable traveling conditions.
There are insufficient fatal crashes to provide reliable estimates of the distance rate for rural, so we assume that the vehicle distance of travel on all roads is proportional to the travel on rural arterials. This enables us to examine how the distance rate varies by month for states with and without snow (lower plot in Fig. 5-2).
The pattern is noisier, perhaps due to uncertainties in the estimates of travel distance. However, the indication is clear - the risk of a fatal crash for the same distance of travel is lower when the road is snow-covered than when it is not.
Number of vehicles per fatal crash. Table 5-3 shows that the average number of vehicles involved in a fatal crash on snow-covered roads is substantially higher than on dry roads. On snow-covered roads, 55% of fatal crashes are multiple-vehicle compared to 43% on dry roads. As the overall fatality risks are lower in snow, the risk of fatality in a single-vehicle crash will be additionally lower. This provides supporting indirect evidence that fatality risks are lower in snow, due presumably to lower speeds. Such an interpretation finds further support in the finding that when a pedestrian injury crash occurs, the probability that the pedestrian is killed is greatest when the road surface is dry, and least when it is ice covered. When it is snowing pedestrian fatality risk when a crash occurs is half what it is when visibility is clear.
Table 5-3. Average number of vehicles per crash for all fatal crashes.
Data from Ontario, Canada. The variation in the
number of fatalities through-out the year in Ontario, Canada
is similar to that observed for the US snow states, with
fatalities in January, February and March being about half
those in July, August and September. (p. 204,205) The
climate in Ontario is somewhat similar to that in the US
states with most snow. The finding in Fig. 5-3 that the
number of deaths per injury is lowest in the winter is
consistent with the interpretation that snow covered roads
lead to lower speeds and consequent lower fatality risks.
Table 5-4 shows that more than half of fatal crashes occur in daylight, while 29% occur in darkness (no daylight or artificial light). Thus the most common weather and lighting conditions in which fatal crashes occur are daylight, no precipitation or other environmental factors degrading visibility, and on a dry pavement. For every person killed while traveling in the dark while it is snowing, 87 are killed traveling in daylight under no adverse atmospheric conditions. For every person killed while traveling in the dark while it is raining, 19 are killed traveling in daylight under no adverse atmospheric conditions.
The 6.35 million kilometers of roads in the US (Table 5-5) exceeds 150 times the earth's circumference, or 15 times the distance between the earth and the moon; the overview presented in Table 5-5 is based on various US Department of Transportation data sources. - These roads vary greatly in characteristics, quality, and use; 35% of the total length is unpaved. Local roads are supported mainly by local taxes (city, township, state), whereas Interstate system roads are supported mainly by Federal taxes (in some cases augmented by user tolls). The Interstate system originated in the 1950s to achieve the national defense goals of connecting the nation with efficient road transportation and facilitating speedy exit from cities. It is built to the highest design standards, with traffic in opposite directions separated by wide medians, or when space is restricted, as in urban areas, by physical barriers. Arterials include many state freeways supported mainly by state taxes that are similar in physical characteristics to the Interstate system, but are not part of it. While Interstates represent only 1.2% of all roads (by length), they carry 24% of the traffic and account for 13% of the traffic fatalities. The distance fatality rate on the Interstate system is less than half that on the non-Interstate system (5.0 fatalities per billion km compared to 10.8 fatalities per billion km).
On average, fatal crashes occur at a rate of one per 168 km of roadway per year, making them rare events on any particular section of roadway notwithstanding the 2001 total of 37,795 fatal crashes. The annual total of
16.35 million crashes of all types (p 9) gives an average of 2.6 crashes per km of roadway per year. Crashes are of course not distributed even approximately randomly over the length of the road system.
Rural compared to urban
Rural roads account for 78% of all roads (by length), and for 60% of traffic fatalities. Yet only 40% of travel is on rural roads. The distance fatality rate on urban roads is 57% lower than the rural rate. Differences in physical characteristics contribute to this difference. A much larger contribution is due to differences in travel speed. Speed limits and congestion reduce travel speeds.
The average daily traffic (the number of vehicles traveling in either direction passing the same location on a road) on urban roads is more than five times the rural rate. During the morning and late afternoon peak travel periods, urban traffic is normally constrained to travel slower than the speeds drivers would choose, and often slower than posted speed limits (which are generally lower on urban than on rural roads).
For the Interstate system, physical characteristics are similar on rural and urban roads. Such physical differences as do exist between the urban and rural Interstates tend to generate a safety disadvantage for the urban roads (more lanes, more exits generating more merging, overtaking, and lane changing). Yet the rural Interstate fatality rate is 89% higher than the urban rate. As physical features of the roadway cannot explain this, it is most plausibly attributed to speed. Speed limits are, on average, higher on the rural system than on the urban system. However, from 1974 to 1987 speed limits were identical on rural and urban Interstates, as required by the nationwide maximum speed limit of 55 mph. Yet fatality rates were still lower on the urban sections. For example, for 1985 the urban Interstate fatality rate was 32% lower than the rural rate, a substantial difference, but less than the 48% difference in 2001. This lower rate in 1985 is due to the safety benefits of congestion.
Fatality rates for different roadway classifications
For rural roads, the fatality rate on the Interstate is 65% lower than on local roads. Local roads include many that are two-lane on which vehicles may approach each other at legal closing speeds in excess of 200 km/h and
pass separated by only a few meters. A head-on crash at such relative speeds will likely be fatal. Improper overtaking or loss of control on curves can produce such crashes. When multiple-vehicle crashes do occur on freeways they usually involve vehicles traveling in the same direction, for which relative speeds are much less. Freeways also largely eliminate some of the most hazardous types of crashes that occur on local roads, including striking trees near the roadway, intersection crashes, and impacts with pedestrians.
Speeds are, on average, lower on non-Interstate roads than on the Interstates, yet fatality rates are more than twice as high on the lower-speed non-Interstates. This could be interpreted to mean that replacing all roads by Interstates would reduce the nation's traffic deaths by more than half. While this is neither feasible nor desirable for a host of reasons, it does underline the important role of the physical characteristics of the roadway in safety.
Many safety interventions, such as speed limits, reduce mobility. However, replacing higher fatality rate roads by lower fatality rate freeways increases both mobility and safety. There are, however, interactive effects that reduce the safety benefits. Enhanced mobility generates more travel. Providing additional roadway capacity to alleviate congestion leads not only to higher speeds, but also to additional traffic.
Keeping roadways in good condition increases mobility, but not necessarily safety. If a speed bump enhances safety, why would a pothole not enhance safety?
The central impetus for much of traffic engineering is the efficient operation of the traffic system, with safety an important consideration. The decision to, say, replace a stop sign with a more expensive traffic light is based mainly on traffic volume and other operational considerations. It is surprisingly difficult to determine how such changes influence safety. An intrinsic problem is that, despite the more than 16 million crashes per year in the US,6 crashes are rare events at any particular site. This makes it difficult to get enough before and after data to support reliable conclusions. Other relevant factors also change between the before and after periods.
Selecting sites for treatment because they have unusually high crash experience encounters another technical problem - regression to the mean.7 To illustrate the concept, assume that all intersections have identical crash risks. Then, as discussed in Chapter 1, all will not have the same number of crashes in some before period because of randomness. The intersections with the largest number of before crashes are not expected to have a number different from the average in any after period. If some treatment that has no influence on crash risk is applied to the high-crash risk intersections, a reduction will still likely
be observed due purely to randomness. It would be incorrect to attribute it to
All intersections do not have equal risk, but it is difficult to infer from counting crashes how much of an above average number is due to random fluctuations, and how much is due to the risk at the intersection really being higher. Techniques have been developed to address these problems, and the influence of many traffic engineering and roadway features has been reliably estimated.7 As this is an extensive subject with an extensive literature venturing beyond the scope of this book, we limit the present discussion to one item that is in a state of flux in the US.
Roundabouts. Roundabouts are a form of intersection control in widespread use throughout the world, being particularly common in Britain. They are relatively uncommon in North America. As roundabouts require more land than traffic lights, one might have expected them to be more common in North America than in much higher population density Britain. The higher use of roundabouts in Britain is, like the choice of left or right-side driving, a manifestation of independent evolution of traffic practice in different continents. This may perhaps be like the evolution of polar bears in the Arctic and penguins in Antarctica - either creature would likely find the opposite polar region just as congenial.
Some of the early roundabouts in the United States were of poor design, which discouraged expanded deployment. The term modern roundabouts refers to more recent US roundabouts of improved design. , As the schematic in Fig. 5-4 indicates, even the simplest roundabout controlling the right-angle intersection of two roads, each with only one lane in either direction, requires many design decisions. The goals of mobility and safety are often in conflict, so that trade-offs are required. For example, a smaller angle between the approach lane and the circulatory roadway will enable vehicles to maintain higher speeds, but an angle approaching closer to 90 degrees will reduce speeds and lead to greater safety, especially for any pedestrians present. The complexity increases as the number of intersecting roads, and the number of lanes per road, increases. Efforts are underway in the US to further refine design guidelines for improved operation and safety with the possibility of more extensive use of roundabouts in the US.
Roundabouts essentially eliminate the most severe type of intersection crash - one vehicle striking another on the side. One study found that existing modern roundabouts in the US compared to traffic signals reduced crashes by 38%, injury crashes by 76%, and fatal and incapacitating injury crashes by about 90%. These findings, which are in line with experience in other countries, imply that increased use of roundabouts in North America will substantially reduce intersection casualties.
Vehicle factors fall into two broad categories - those aimed at reducing harm when crashes occur, and those aimed at reducing the risk of crashing. Chapter 4 was devoted to the vehicle factor that has the greatest influence on the risk of death when a crash occurs, namely, the mass (and size) of the vehicle. Another set of vehicle factors, occupant protection devices such as safety belts and airbags, will be dealt with in Chapters 11 and 12. Here we consider just one device aimed at reducing the risk of crashing, especially in adverse weather conditions, and then discuss a number of safety standards.
Antilock braking systems (ABS)
Antilock braking systems (ABS) use electronic controls to maintain wheel rotation under hard braking that would otherwise lock a vehicle's wheels. Keeping the wheels rotating increases vehicle stability, especially
when tire/roadway friction is reduced or varying, as when the roadway is wet. Anyone who informally investigates the emergency-braking performance of ABS on a snow-covered deserted parking lot will soon be impressed by the effectiveness of the technology. Systematic test-track evaluations have convincingly demonstrated the technical advantages of ABS under a wide variety of conditions. - However, the influence of ABS on crash risk cannot be investigated on test tracks (or in parking lots) - this requires data from
A number of studies used the crash experience of seven General Motors cars (Chevrolet Cavalier, Chevrolet Beretta, Chevrolet Corsica, Chevrolet Lumina APV, Pontiac Sunbird, Pontiac Trans Sport, and Oldsmobile Silhouette). These were relatively unchanged between model year 1991 and model year 1992, except that none of the 1991 models had ABS while all the 1992 models did. Thus, comparing the crash experience of the 1992 and 1991 models is equivalent to comparing the experience of vehicles with and without ABS.
Two-car crashes were examined using data from five states (Indiana, Missouri, North Carolina, Pennsylvania and Texas). It was found that on wet roads ABS reduced the risk of crashing into a lead vehicle by (32 ± 8)%, but increased the risk of being struck in the rear by (30 ± 14)%. These results provide unmistakable evidence that ABS led to large differences in braking
Another study using the same data compared crash risks under one condition to risks under another condition, as summarized in Table 5-6. The first entry shows that ABS vehicles have (10 ± 3)% fewer crashes on wet roads relative to their experience on dry roads compared to the corresponding ratio for non-ABS vehicles. If one makes the plausible assumption that ABS does not affect crash risk on dry roads, the result implies that ABS reduced crash risk on wet roads by (10 ± 3)%. Likewise, assuming that ABS does not affect crash risk when it is not raining leads to the result that ABS reduced crash risk when it is raining by (12 ± 2)%. These important crash-risk reductions appear to be the result of ABS preventing skidding on slippery surfaces, what the system was designed to do. If ABS has no effect on the dry roads on which 80% of crashes occur, then the net effect on all crashes would be a reduction of about 2%. This is consistent with the findings of the many studies cited in Refs 15 and that find no net reduction in crashes to be associated with ABS. It is extremely unlikely that an effect as small as 2% can be detected in an overall crash rate. Early plans by insurance companies to give premium reductions for ABS were soon abandoned in the face of no demonstrable benefits.
The last entry in Table 5-6 shows that ABS is associated with a (39 * 16)% increase in rollover risk. After the first report that ABS substantially increased rollover risk, a number of studies - found similarly large risk increases, as summarized in Table 5-7. While the results are in some cases based on multiple analyses of the same or overlapping data, they nonetheless paint a picture that leaves little doubt that equipping cars with ABS increases rollover risk. As rollover crashes pose a high fatality risk, it is not surprising that there is no evidence that ABS reduces fatality risk overall.20, Indeed, the data suggest an increase is more likely than a decrease.
Why does ABS increase rollover risk? The finding that ABS increased rollover risk generated much concern by government officials and auto executives, who urgently sought explanations. Surveys showed that many drivers did not know that their vehicles had ABS, and those who did know were often unaware what ABS did, or how they were supposed to use it.16 There seems to be no convincing reason why a lack of knowledge about ABS should increase rollover risk. It is, of course, possible that wheels rotating rather than locking, or other engineering effect could, could promote rollover, but evidence is lacking.
The finding of higher rollover risk for ABS cars is not all that surprising given that one of the earliest technical papers on driving, published in the American Journal of Psychology in 1938, discussed why better breaking does not enhance safety. (We discuss this in detail in Chapter 13). Here we address specifically why ABS does not enhance safety even though it provides demonstrably superior braking.
Evidence the ABS-equipped vehicles are driven at higher average speeds.
I believe that the reason behind these effects is that ABS leads to marginally higher average travel speeds. While there are no rigorous studies showing that ABS vehicles travel at higher speeds than non-ABS vehicles, I am nonetheless confident that this is the explanation because of the combined weight of the following evidence:
1. Data from Oregon for the same previously mentioned seven GM cars showed that the fraction of all traffic violations that were for speeding was greater for the ABS cars.15
2. ABS-equipped taxis in Norway were observed to follow at shorter headways than non-ABS taxis. Other research indicates that tailgaters drive faster. ,
3. The previously mentioned result that on wet roads ABS cars were (30 ± 14)% more likely to be struck in the rear than non-ABS cars is consistent with the interpretation that ABS was facilitating closer following, which led to braking levels that following cars without ABS could not match.
4. A 10% increase in average severity was associated with ABS in Canadian insurance claims, consistent with higher speeds.
5. A test-track experiment suggested that drivers of ABS vehicles choose higher travel speeds.
6. I have asked the following question to many audiences. "Have any of you, under any circumstances, on any occasion, ever driven faster because your vehicle was equipped with ABS"? Generally a few hands are raised. I then ask the parallel question with "slower" replacing "faster". It is rare for anyone to indicate that they ever drove slower because they had ABS. Any instance of faster driving not balanced by one of slower driving implies that average speed is higher with ABS. The accumulated responses provide overwhelming statistical evidence that drivers self-report that, on average, they drive faster when they have ABS.
7. Respondents to a formal written questionnaire indicated they would drive faster if driving an ABS-equipped vehicle.
8. My own average speed is unquestionably higher with ABS. Before I had ABS and snow covered the two-lane road on which I commuted, I chose a speed sufficiently low to preclude much risk of moderate braking. I did not relish the possibility of skidding into oncoming traffic or into the deep drainage ditch on my right. When ABS eliminated these risks, I naturally drove faster. (As stated on p. 359, I have never been in a crash).
The risk of rollover is expected to depend very steeply on speed, so that even small speed increases generate large increases in rollover risk. While injury risk from, say, hitting a tree depends strongly on speed, it does so as a continuous function of speed. Rollover is more of a trigger, or threshold, phenomenon. A small increment of speed can make the difference between no incident of any type and a fatal rollover crash. Fatality risk, averaged over all crashes, increases as the fourth power of travel speed, so rollover fatality risk is expected to increase even more steeply than this. It is plausible that an undetectable small increase in travel speed associated with ABS would lead to the observed increases in rollover risk.
Is ABS a desirable technology? It is hard to imagine reasons why a driver would not prefer a vehicle with ABS over one without ABS, other factors being equal. Better braking is a desirable vehicle feature. However, it can be used to increase mobility or safety. Mobility may be increased by traveling at higher speeds, maintaining cruising speeds longer before beginning to stop, or by completing severe-weather journeys that would be abandoned if ABS were unavailable. We have treated this subject in some detail because it illustrates many themes that are central to traffic safety. It particularly provides guidance regarding the likely performance of proposed new technologies.
Major public efforts to incorporate advanced technologies
ABS is an example of applying advanced technology in an effort to enhance vehicle safety. It is the example that has had by far the most thorough evaluation. Many ABS effects in traffic are well established in multiple studies. The motivation and funding for ABS technology came from the auto industry, and led to a product that found widespread customer demand, notwithstanding its lack of any demonstrated overall safety benefit.
Major public funds have been spent in attempts to increase the use of modern technology to improve the performance of vehicles and the systems in which they operate. In the US, part of the Intermodal Surface Transportation Efficiency Act of 1991 authorized substantial expenditures to support Intelligent Vehicle Highway Systems (IVHS). The goal was to use a range of smart car and smart highway technologies to improve the safety, efficiency, and environmental friendliness of the highway system. This concept later evolved into Intelligent Transportation Systems (ITS), with stated aims of applying electronic, computer, and communication technology to vehicles and roadways to increase safety, reduce congestion, enhance mobility, minimize environmental impact, increase energy efficiency, and promote economic productivity for a healthier economy. In Europe Programme for European Traffic with Highest Efficiency and Unprecedented Safety (PROMETHEUS) was generously funded.
These programs generated massive activity, committees, bureaucracy, and extensive documentation on process, coordination, goals, needs, problem statements, definitions, and so on. Claims were made that reducing congestion would enhance safety, whereas the opposite is to be expected. It was likewise claimed that improved driver information would reduce congestion, without realizing that such a claim implied rejecting Wardrop's celebrated 1952 principle, "journey times on all routes actually used are equal." This principle is at the core of traffic assignment modeling and has been accepted for decades as a good approximation when traffic is in a normal, or equilibrium, state. If the equilibrium is disturbed by, for example, a crash, then information can be helpful in selecting an alternate route. However, a few drivers receiving such information by traditional radio broadcasting will soon establish a new equilibrium, restoring a new parity between routes. Under the umbrella of these government-supported programs some technologies, such as on-board vehicle navigation, in-vehicle telephones, Internet, and on-board television reception, advanced. However, the advances were underway and would have occurred through the same competitive forces that produced earlier automotive innovation. I have been unable to discover in the mountain of literature any understandable summary of what was accomplished at such vast public expense. An approximate estimate of the expense is also unavailable because each item is a not too clearly identified portion of some larger transporting allocation.
The situation brings to mind comments of Ezra Hauer, the distinguished Professor Emeritus of Transportation at Toronto University.
Those amongst you who have attempted a critical review of the literature will attest that many of the research reports found will be quickly discarded. They will be found deficient in method, too small to draw conclusions from, inconclusive, obsolete, of obscure message, biased, or otherwise fatally flawed. In the end one is left with very few studies that are not obviously unreliable and the results of which do not contradict each other. That this is not an exaggeration but the actual state of affairs I know from rich personal experience and from noting the experience of many others. The obvious question is why so much effort, by so many, on so many subjects has produced so little light? Why so much that has been published is unsound, inconclusive and generally of little practical use. The answer becomes obvious if one recognizes how much is produced and published by one-day wonders, by itinerant and untrained researchers without experience, here today - gone tomorrow. How much research has been ill conceived by those who set the question to be answered, who provided the money, who approved the research method, who accepted the product and who published the results.
US Federal Motor Vehicle Safety Standards (FMVSS)
Since the beginning of motorization auto manufacturers developed and marketed vehicle features aimed at increasing safety, including automatic windshield wipers (1921), four-wheel brake systems (1924), interior sun visors (1931), electric turn signals (1938), door safety latches (1955), etc. The National Motor Vehicle Safety Act of 1966 empowered NHTSA to develop Federal Motor Vehicle Safety Standards (FMVSS). All vehicles manufactured in 1968 or
later had to meet a number of these in order to be offered for sale. New standards continue to be promulgated and those in place continue to be refined. In some cases standards required all vehicles to include features already available on some, while in other cases completely new performance standards were specified.
There are three main categories of standards for cars. Those numbered in the 100's apply to crash avoidance, those numbered in the 200's apply to occupant protection, given that a crash occurs, and those in the 300's apply to the post-crash period. The standard generating by far the most comment and controversy is FMVSS 208, which deals with occupant protection (safety belts, airbags). This is treated in Chapters 11 and 12.
Here we mention only standards (other than FMVSS 208) that have led to a measured change in fatality risk. The effectiveness of a standard can be estimated by comparing the fatality rate of a vehicle satisfying the standard with the rate of pre-standard model year versions of the same vehicle. The estimates discussed below are all from NHTSA. The largest effects are associated with the earliest standards introduced. As evaluation must use data from periods close to the introduction of the standard, the estimates are necessarily from earlier periods. The calculations below use the mix of vehicles, crashes, etc. applying at the times of the studies.
The largest fatality reductions are from the combined effects of FMVSS 203 and FMVSS 204. FMVSS 203 required energy-absorbing steering columns designed to cushion the driver's chest in a frontal crash. FMVSS 204 limited the rearward displacement of the steering wheel towards the driver. The energy-absorbing, or collapsible, steering column which meets these standards was first introduced by General Motors in 1966. It involves dividing the steering column into two separate sections, and joining them with a sleeve made from relatively thin sheet metal. When a driver's chest impacts the steering wheel, the sleeve crumbles thereby reducing the forces on the driver's chest.
The collapsible steering column was designed to reduce driver fatality risk in frontal crashes, but should not affect passenger risk or driver risk in non-frontal crashes. Hence, effectiveness could be estimated by comparing right-front passenger to driver fatality risk, or by comparing driver risk in frontal crashes to driver risk in non-frontal crashes. Applying each method to 1975-1979 FARS data produced estimates of 13% and 11%. , Although these estimates are not independent in that each uses the same driver fatalities, the agreement nonetheless suggests a fairly robust effect. The average, (12.1 ± 1.8)%, applies to frontal crashes which accounted for 54% of driver fatalities. Thus the device provides a net 6.6% reduction in driver fatality risk or, when averaged over all car occupants, a 4.4% reduction (Table 5-8).
The 6.6% reduction in driver fatality risk from this very simple, reliable, and inexpensive device that hurts no one is two thirds that of the 10% reduction from airbags (Chapters 11 and 12). The collapsible steering column, unlike the airbag, is a passive device in that the user does not have to know anything or take steps, such as avoiding being close to it. Indeed, few users even know it is there.
Improvements in instrument panels from 1965 to 1975 are estimated to reduce fatality risk by about 13% for unrestrained front passengers in frontal crashes,
or by 7% for all crashes. As right-front plus center-front passengers constituted 24% of car-occupant fatalities, this reduces unrestrained car-occupant fatalities by about 1.7%.
Side door beams (FMVSS 214) are estimated to have reduced risk by about 1.7%. Improved door locks and door retention components (FMVSS 206) and improved roof crush resistance (FMVSS 216) are estimated to reduce fatality risk by 1.5% and 0.43%. Adhesive bonding is estimated to halve windshield bond separation and occupant ejection through the windshield, thereby preventing 105 fatalities annually. As an approximation, we express this as an average risk reduction of 0.39% for all occupants.
Head restraints for drivers and right-front passengers are estimated to reduce overall injury risk in rear impacts by 12%. Let us make the very approximate assumption that this applies also to fatalities, about 3% of which result from rear impact (principal impact point 6 o'clock), so we obtain a net reduction in outboard-front occupant fatalities of 0.36%, or 0.33% of all occupant fatalities.
One measure aimed at crash prevention rather than occupant protection, namely dual master brake cylinders, is estimated to prevent 260 fatalities, or 0.9%.
Combined effect of all standards. Although summing contributions in Table 5-8 gives an approximate estimate of the total reduction in risk from the combined effects of all the individual reductions, this is an incorrect calculation. The application of two measures that reduce risk by 50% does not eliminate risk, but reduces it by 75%. There are no logical problems with three measures that each reduce risk by 50%. If better brakes prevent a crash, then the contribution of the energy absorbing steering column must not be included for the crash that did not occur.
Seven of the eight items in Table 5-8 apply to drivers. Their combined effect is a reduction in driver fatality risk computed as
Applying similar reasoning gives that the combined effects of all the standards reduce fatalities to right-front, center-front and all rear passengers by 11.8%, 11.5%, and 4.8%. These passengers are 23%, 1%, and 8%, respectively, of all occupant fatalities, drivers being the remaining 68%. By weighting each reduction by the corresponding occupancy, the combined effect of the six measures in Table 5-8 is estimated to reduce car occupant fatalities by 10.9%.
Table 5-8 does not include all changes that may have reduced car-occupant fatalities, but only those for which fatality reductions have been quantitatively estimated. Other FMVSS standards may be associated with fatality reductions too small to have been measured, yet important in terms of total numbers. If a change prevents, say, 20 fatalities per year, it is exceedingly unlikely that it will be detected in field data. If the 10.9% estimate is increased by half of its estimated value to capture the missed effects, this implies that vehicle changes have reduced occupant fatality risk by about 16%. Many automotive engineers consider that the cumulative effect of vehicle changes have reduced car-occupant risk somewhere in the range 10% to 25%, a range consistent with the above discussion.
It is to be expected that the largest fatality reductions are associated with the earliest standards, as the most fruitful opportunities are naturally addressed first. While it is in principle possible to always keep generating some increment of increased safety in vehicles, the law of diminishing returns soon sets in.
The discussion has attributed the risk reductions to the standards. As many of the changes would (and in fact did) occur without the standards, and vehicle safety improved prior to the standards, only some unknown fraction of the reductions can be attributed directly to the standards.
Attempts to estimate aggregate effects of FMVSS standards directly. Rather than estimating the aggregate effect of the standards by combining contributions from specific standards, it would be desirable to examine the overall effect by a more general change in fatalities from pre-regulation to post-regulation vehicles. Such a task is rendered difficult because the earliest calendar year for FARS data is 1975, by which time the newest cars unaffected by FMVSS standards, namely 1967 model-year (MY) cars, were already eight years old. Thus any study using FARS data must necessarily focus on very old cars, which have use and ownership patterns that differ from those for newer cars by larger amounts than are expected to be associated with vehicle design standards.
One much cited study estimated the combined effects from all changes by comparing fatalities per unit distance of travel for pre-1964 MY cars, 1964-1967 MY cars, and 1968-1977 MY cars, as estimated by applying multivariate analysis to 1975-1978 FARS data. The study concluded, "The numbers of deaths avoided by the federal safety standards amount to 26,500 occupants, 7,600 pedestrians, 1,000 pedalcyclists and 2,000 motorcyclists - for a total of about 37,000 people who would have died without the standards in those years" (the four years 1975-1978). Another much cited study applying multivariate analysis to similar data finds no net reduction associated with the standards. I have previously discussed these studies in greater detail. (p 81-83)
Such disparate conclusions from the same data illustrate what seems to be an intrinsic problem with complicated multivariate analyses. There are so many choices of variables and of transformations at the discretion of the analyst that the detached reader rarely has any way of knowing whether the analysis is performed to discover new information or to buttress prior beliefs. The reader cannot generally check the calculation, nor get a clear sense of the origin of the claimed effects. Differences in interpretation often do not flow from, say, different assumptions that can be discussed in terms of plausibility, but from such arcane issues as whether to use the logarithm or the reciprocal of the dependent variable in specifying the model. Often effects are reported that exceed reasonable physical explanations, or are inconsistent with straightforward observations. Claims of large benefits based on multivariate analyses seem so often to represent the triumph of zeal over science, or even over common sense.
A simple examination of the trends in Figs 3-3 through 3-5 (p. 38-40) is sufficient to refute any claim that a dramatic reduction in fatalities followed the introduction of vehicles satisfying safety standards in the late 1960s. Extravagant claims of benefits from US vehicle regulation continue to be made - and refuted by data. In Chapter 15 we discuss the role that inflated claims of benefits from vehicle factors played in the dramatic failure of US safety policy.
Summary and conclusions (see printed text)
References for Chapter 5 - 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.
 Severy DM, Harrison MB, Blaisdell
DM. Smaller vehicle versus larger vehicle collisions. Paper
number 710861. SAE Transactions. 1971; 80: 2929-58.
 Campbell BJ, Reinfurt DW. The
relationship between driver crash injury and passenger car
weight. Chapel Hill, NC: Highway Safety Research Center,
University of North Carolina; 1973.
 ONeill B, Joksch H, Haddon W Jr.
Empirical relationships between car size, car weight and
crash injuries in car-to-car crashes. Proceedings of the
Fifth International Technical Conference on Experimental
Safety Vehicles, London, UK, p 362-368. Washington, DC:
National Highway Traffic Safety Administration, 4-7 June
 Joksch HC. Analysis of the future
effects of the fuel shortage and increased small car usage
upon traffic deaths and injuries. Report DOT-TSC-OST-75-21.
Washington, DC: US Department of Transportation; January
 Grime G, Hutchinson TP. Vehicle
mass and driver injury. Ergonomics. 1979; 22: 93-104.
 Insurance Institute for Highway
Safety. Special issue: Driver death rates. Status Report,
Vol. 35, No. 7, August 19, 2000. http://www.hwysafety.org/sr_ddr/sr3507_t1.htm
 Kahane CJ. Vehicle weight,
fatality risk and crash compatibility of model year 1991-99
passenger cars and light trucks. Report DOT HS 809 662.
Washington, DC: US Department of Transportation, National
Highway Traffic Safety Administration; October 2003.
 Evans L, Frick MC. Car size or
car mass – which has greater influence on fatality risk?
Am J Public Health. 1992; 82: 1009-1112.
 Evans L, Frick MC. Mass ratio and
relative driver fatality risk in two-vehicle crashes. Accid
Anal Prev. 1993; 25: 213-224.
 Evans L. Driver injury and
fatality risk in two-car crashes versus mass ratio inferred
using Newtonian Mechanics. Accid Anal Prev. 1994; 26:
 Evans L, Frick MC. Car mass and
fatality risk – has the relationship changed? Am J Public
Health. 1994; 84: 33-36.
 Evans L. Causal influence of car
mass and size on driver fatality risk. Am J Pub Health.
2001; 91: 1076-81.
 Evans L. Safety-belt
effectiveness: The influence of crash severity and selective
recruitment. Accid Anal Prev. 1996; 28: 423-433.
 Joksch HC. Velocity change and
fatality risk in a crash – a rule of thumb. Accid Anal
Prev. 1993; 25: 103-104.
 Toy EL. The distribution of
vehicle mass in the on-road fleet of passenger vehicles. SAE
paper 2004-01-1161. Warrendale, PA: Society of Automotive
 Evans L. Driver fatalities
versus car mass using a new exposure approach. Accid Anal
Prev. 1984; 16: 19-36.
 Joksch H, Massie D, Pichier R.
Vehicle aggressivity: Fleet characterization using traffic
collision data. Report DOT HS 808 679. Washington, DC: US
Department of Transportation, National Highway Traffic
Safety Administration; February 1998.
 Gabler HC, Hollowell WT. The
aggressivity of light trucks and vans in traffic crashes,
SAE paper 980908. Warrendale, PA: Society of Automotive
 Joksch H. Vehicle design versus
aggressivity. Report DOT HS 809 194. Washington, DC: US
Department of Transportation, National Highway Traffic
Safety Administration; April 2000.
 Wood DP, Ydenius A, Adamson D.
Velocity changes, mean accelerations and displacements of
some car types in frontal collisions. Int J Crashworthiness.
2003; 8: 591-603.
 Kahane CJ. Effect of car size on
the frequency and severity of rollover crashes. Proceedings
of the 13th International Technical Conference on
Experimental Safety Vehicles, Paris, France; 4-7 November
1991. Document DOT HS 807 990, Washington DC, Vol. 2, p.
765-770, July 1993.
 Evans L, Wasielewski P. Serious
or fatal driver injury rate versus car mass in head-on
crashes between cars of similar mass. Accid Anal Prev. 1987;
 Ernst E, Bruhning E, Glaeser KP,
Schmidt M. Compatibility problems of small and large
passenger cars in head on collisions. Paper presented to the
13th International Technical Conference on Experimental
Safety Vehicles, Paris, France; 4-7 November 1991.
 Wood DP. Safety and the car size
effect: A fundamental explanation. Accid Anal Prev. 1997;
 Evans L. How to make a car
lighter and safer. SAE paper 2004-01-1172. Warrendale, PA:
Society of Automotive Engineers; 2004.
 Evans L. Traffic Safety and the
Driver. New York, NY: Van Nostrand Reinhold; 1991.
 National Research Council.
Effectiveness and Impact of Corporate Average Fuel Economy
(CAFE) Standards. Washington, DC: National Academy Press;
 National Highway Traffic Safety
Administration. Traffic Safety Facts 2001. Report DOT HS 809
484. Washington, DC: US Department of Transportation;
 Complete details of this and
other calculations available at
 Auto Insurance Rates Results.
 Evans L. The foreign policy of
SUVs. Letter to the Editor, New York Times, 22 October 2002.
 Padmanaban J. Influences of
vehicle size and mass and selected driver factors on odds of
driver fatality. Proceedings of the 47th Annual Meeting of
the Association for the Advancement of Automotive Medicine,
p. 507-524, Lisbon, Portugal; September 22-24, 2003.
 US vehicle population. http://www.autonews.com/files/00regvehiclepop.pdf
 Davis SC. Transportation Energy
Data Book: Edition 21, ORNL-6966. Table 4.9: Light vehicle
market shares by size class, sales periods, 1976-2002. Oak
Ridge, TN: Oak Ridge National Laboratory, US Department of