Better to see Chapter 1 Introduction of Traffic Safety (2004)

Chapter 4.  ENGINEERING, ROADWAY AND ENVIRONMENTAL FACTORS

(From 1991 book Traffic Safety and the Driver)

  

INTRODUCTION

 

            Many factors are associated with every traffic crash.  The word "cause" has largely disappeared from the technical literature on safety, and for good reasons.  Suppose on a dark rainy morning a young man argues with his wife about the purchase of a sofa, leaves the house late for work in a rage, drives his poorly-maintained car too fast on a badly-designed poorly-lit curve, skids, and is killed in a crash with a truck driven by an older driver.  It is of little value to say that the death was "caused" by the car driver's youth or maleness, the truck driver's old age, the car's bald tires, the high cost of sofas, emotional stress, the non-use of a safety belt, inadequate police enforcement, rain, or any other of the many factors which, if different on this particular occasion, would have prevented the death.

            All too often the term "cause" conveys the notion of a single cause, in the deterministic sense in which it is used in the physical sciences or engineering.  The rich variety of individual factors which, if different, would alter the outcome or probability of occurrence of crashes, can be classified into broad categories using different schemes.  Haddon [1972] introduced a 3X3 matrix classification in which all factors are classified either as human, as vehicle and equipment, or as environment.  Each of these is further categorized as pre-crash, crash, or post-crash.  The example which opened this chapter involved factors in more than half of the nine cells in this matrix.  In this book we use a broader classification into road-user and non-road-user factors; this chapter is devoted to non-road-user factors -- those relating to engineering and environment.

 

DIFFICULTY OF IDENTIFYING SPECIFIC FACTORS

 

            One recurrent complexity in attempting to understand traffic safety is that factors interact with each other -- every piece of the traffic system is in some way connected to every other piece.  If drivers know their vehicles are in poor safety condition, they may exercise increased caution.  If a haz­ardous section of roadway is rebuilt to higher safety standards, it is likely that drivers will travel this section faster than before the improvement, or with reduced care.  Differences in crash rates on different types of roadways reflect not only effects due to the roadways as such, but also that different speed limits, driver speed choices, and driver vigilance levels are associated with different types of roadways.  Any observed dependence of crashes as a function of specific engineering and environmental factors possibly reflects large contributions from interactions with other factors.

            This point can be illustrated dramatically by the large differences in fatality rates for different vehicles.  Fig. 4-1 shows fatal crash involvements per million registered cars for many individual car models; the dependent variable (Head Injury Criterion measured in the New Car Assessment Program) is not germane to the present discussion.  The highest rate is more than six times the lowest.  It is implausible that such differences could be due to differences in engineering safety between the cars, especially as all are of similar model year, and all are in conformity with all applicable safety standards.  The major contributor to the wide variation in rates in Fig. 4-1 is that some cars are used by different types of drivers in different ways and in different driving environments than other cars.

----------------------------------------------------------------------------

Fig. 4-1 about here

----------------------------------------------------------------------------

            The pitfalls of examining simple rates and attributing the differences to engineering factors are illustrated further using insurance data [Highway Loss Data Institute 1989].  Table 4-1 shows injury-claim rates, defined here as insurance injury claims per insured car-year, relative to a value of 100 for the average of all cars.  All cars listed in identical ways in more than one of the categories labelled "station-wagon", "four-door" or "two-door" are included; if, say, a different restraint system is indicated for the two-door and four-door versions, this car is not included.  The published data classify cars as small, midsize, or large depending on whether their wheelbase is 99 inches or less, greater than 99 inches but not greater than 109 inches, and greater than 109 inches.

----------------------------------------------------------------------------

Table 4-1 about here

----------------------------------------------------------------------------

            Table 4-1 shows consistently lower injury-claim rates for station wagons than for non-station wagons of the same make and model years.  There are 18 cases of station wagons and four-door cars of the same make; in 17 of these the station wagon has a lower injury-claim rate -- the average of all 18 differences being 17.7%.  In 8 of 9 cases in which there are station wagons and two-door cars of the same make, the station wagons has a lower injury-claim rate -- the average of all 9 differences being 18.4%.  The evidence is conclusive that station wagons have lower injury-claim frequencies than the non-station wagon versions of the same make cars, the quantitative estimate being that the station wagon rate is (18 + 3)% lower.

            The data in Table 4-1 also indicate differences dependent on the number of doors.  For the two large car models available in two-door and four-door versions, the injury-claim rate for the four-door version is less than that for the two-door version.  The average and standard error of the two differences, (15 + 11)%, indicates lower rates for the four-door version (the nominal interpretation is a 91% probability that the four-door rate is really lower than the two-door rate, compared to a 9% probability of the reverse).  For midsize cars, the larger number of cases of four- and two-door versions of the same model give clearer effects.  In nine out of ten cases the four-door version has a lower injury-claim rate than the two-door version, with the average of the ten differences being (4.4 + 1.5)%.  For the small cars there is no consistent difference dependent on number of doors -- the average value of the 15 differences is (-1.9 + 2.4)%, a difference inconsequential in both magnitude and statistical significance.  Thus we find no difference for small cars, a small but reliable difference for midsize cars, and a large but uncertain difference for large cars.  With the exception of small cars, the four-door version has a lower injury-claim rate than the two-door version of the same car.

            The engineering details of the different versions of the same cars, except differences specific to number of doors or whether it is a station wagon, are generally the same.  Such differences are unlikely to have much influence on injury risk, especially in frontal crashes, the most common injury-producing crash impact direction.  The differences do not plausibly imply that cars should be redesigned to more resemble station wagons, or that a couple of extra doors be added to larger two-door cars to increase their crashworthiness (although Kahane [1990] does find higher ejection rates associated with two-door compared to four-door cars).  What the differences across the rows in Table 4-1 are more likely showing is not so much engineering differences, but differences in use and behavior.  There may also be driver age effects; although the data [Highway Loss Data Institute 1989] include a correction for the number of youthful drivers listed in the insurance policy, this cannot capture the dominant influence of driver age, especially as the age of the driver involved in the crash is not used in the correction.

            Much larger than the differences dependent on car body style are those dependent on car size category.  Focusing on station-wagons shows a 31% higher rate for midsized compared to large, and a 71% higher rate for small compared to large.  For four-door cars the corresponding values are 54% and 99%, and for the two-door cars 42% and 61%.  These differences suggest large increases in injury risk with decreasing car size.  Indeed, much of the variation in Table 4-1 within the three car-size categories is due to the wide range of car sizes necessarily included within each of these broad categories.  Station wagons are generally heavier than corresponding four-door cars, which are generally heavier than two-door versions, which probably accounts for some of the differences.

            Car size effects, being among the largest and most consistently observed effects in traffic safety, are discussed in some detail below.  Not only are such effects important in themselves, but they provide additionally a means to disentangle some of the interactions between engineering factors and road-user factors.

 

VEHICLE MASS

 

            In investigating the influence of vehicle size on risk, we must first define "size".  As vehicles are not of identical shape, one linear dimension, such as the frequently used wheelbase, does not necessarily characterize size.  The least satisfactory approach, from a technical point of view, is the one to which the public probably most readily relates, which is to categorize vehicles as subcompact cars, mid-sized cars, etc.  These manufacturer designations are not based exclusively on physical dimensions, so that a new model-year version of a specific vehicle will retain its former designation even if its physical dimensions change.  The data in Table 4-1 are based on wheelbase, an objective physical property of the vehicle.

            For a number of reasons I consider mass (essentially weight) to be the best (but certainly not perfect) indicator of a vehicle's size for most analyses.  There is no one variable which captures the pertinent safety characteristics related to size under all conditions.  Mass is a clearly defined physical property of the entire vehicle; unlike wheelbase, it nearly always changes when any component in the vehicle changes.  Mass is a central physical parameter in crash dynamics, especially in two-vehicle crashes.  If vehicles of identical mass crash into each other head-on, the change in speed sustained by each is identical without regard to any other physical properties of the vehicles.  This does not mean that the forces experienced by the occupants depend only on mass.  These forces are directly influenced by the amount of time the occupant spends changing from the pre-crash speed to the post-crash speed, and by the maximum deceleration forces sustained.  These are greatly influenced by the amount of crush space available, which is related to size rather than mass. Indeed, McCarthy et al. [1989] find that for the same wheelbase, larger mass cars actually have higher risk in single-car crashes, although this is possibly a confounding effect with engine power since a larger engine will make an otherwise identical car heavier.  Apart from such specific considerations, there is a strong tendency for most linear dimensions of similarly shaped vehicles to increase with mass, so that this single quantity captures size effects to a large extent.  For a crash into an immovable barrier, the quantity and characteristics of crushable material and space in front of the occupant are more important than mass.  However, as crashes in the real world nearly all involve objects that will to some extent move, bend, break or distort, increased mass in the vehicle will always reduce the deceleration forces experienced within the vehicle.

            Masses of specific car models are available in many sources, such as Automotive News [1989].  Some 1989 model-year cars at around 900 kg curb mass are the Toyota Tercel (907 kg) and the Honda Civic Hatchback (913 kg).  The Ford Crown Victoria 4-door station wagon (1804 kg), Mercedes Benz 560 SEL Sedan (1851 kg), and Mercury Marquis 4-door Sedan (1707 kg) are close to 1800 kg.  While 900 kg and 1800 kg provide a fairly extreme comparison, there are still many lighter and heavier cars; for example, the Geo Metro 2-door Hatchback (719 kg), the Ford Fiesta (777 kg), the Cadillac Brougham 4-door Sedan (1901 kg) and the Rolls Royce Bentley Continental (2422 kg).  The median car mass is about 1400 kg.  The above illustrative examples are all for 1989 model-year cars -- older model-year cars, especially those prior to model year 1977, tended to be heavier.

 

Analytical relations with car mass  Another important advantage of a physical measure such as mass is that quantitative relations can be derived.  Data are fitted below by a weighted least squares procedure to

 

                      ln y   =   A  +  b m          Eqn 4-1

 

where ln is the natural logarithm (to base e), y is the crash rate, m is the mass, and A and b are parameters derived from the data.  The data are displayed in a linear rather than logarithmic form, so Eqn 4-1 may be conceptualized more conveniently as:

 

                        y  =  a exp (b m)           Eqn 4-2

 

where a = exp (A).  A logarithmic relationship between m and y has a number of advantages over a linear relationship.  First, the logarithmic relation generally provides a better fit to the data.  Second, the logarithmic relation is a simpler and more effective way to compare the rates for cars of different masses because the comparison involves only the slope parameter, b.  The other parameter, a, is merely a scaling factor which cancels when cars are compared, and is, in any event, arbitrary when relative rates are examined.        

 

Illustrative example -- comparing 900 kg car to 1800 kg car  It is even simpler to compare rates for two cars of specific masses -- one a small car and the other a large car.  The specific masses we chose are 900 kg and 1800 kg.  Thus, we compare the rate for a 900 kg car to that for an 1800 kg car, which, from Eqn 4-2, is given by

 

    y(m = 900 kg)/y( m = 1800 kg)  =  exp (900 b)        Eqn 4-3

 

where b is in kg^-1.  Choosing a specific comparison enables us to compare results summarized in the literature using different equations, such as a linear relation between mass and rates.  After deriving the various car mass effects, they are summarized (Table 4-2) in terms of the 900 kg compared to 1800 kg car risk.

 

Two-car crashes

 

            The large role of mass in two-car crashes has long been recognized.  Using state injury data, Joksch [1976a], and Campbell and Reinfurt [1973] find that when cars of dissimilar mass crash into each other, the risk of injury is substantially greater in the smaller car.  FARS provides more extensive data than were available when these important earlier studies were performed.

            Fig. 4-2 shows the number of drivers killed in cars in one mass category as a result of crashing into cars in another mass category, based on FARS data for 1975 through 1980.  For example, there are 578 drivers in cars in the lightest of the six mass categories killed in crashes with cars in the heaviest category; there are 35 drivers in heaviest cars killed in crashes with the lightest cars.  All these drivers were killed in the same set of crashes -- those between lightest and heaviest cars. The ratio, 578/35 = 16.51, shows that when cars in the lightest and heaviest categories crash into each other, the driver in the lighter car is about 17 times as likely to die as the driver in the heavier car.  Extending the analysis to trucks shows that drivers in small cars are about 50 times as likely to die as drivers in large trucks when these vehicles crash into each other.

----------------------------------------------------------------------------

Fig. 4-2 about here

----------------------------------------------------------------------------

            To answer the more general question of how the risk of death when one pair of cars crash compares to the risk when some other pair of cars crash requires knowing the rates at which the different pairs crash.  In other words, we need to estimate exposure for involvement in two-car crashes.  This can be done using an approach, to be described in the next section, in which the numbers of pedestrians killed in crashes involving cars in the six mass categories are used to estimate exposure.  In this way, the relative risk of driver death for two-car crashes involving any pair of cars is calculated, as shown in Fig. 4-3.  All values are expressed relative to an arbitrary value of unity for the driver at lowest risk -- namely one in the heaviest car crashing into the lightest car.  A driver in a car in the lightest category crashing into another in the lightest category is 7.04 times as likely to be killed as is a driver in a car in the heaviest category crashing into a car in the lightest category, but 7.04/16.51 = 0.43 times as likely to be killed as a driver in a car in the lightest category crashing into one in the heaviest.

----------------------------------------------------------------------------

Fig. 4-3 about here

----------------------------------------------------------------------------

            The data in Fig. 4-3 show that as a car's mass increases, fatality risk in that car decreases, but the risk in the other involved car increases.  Thus drivers transferring to larger cars reduce their personal fatality risk, but impose increased fatality risk on occupants of other cars.  This raises the question of whether a driver transferring to a larger car generates a net increase or decrease in fatality risk.  The data in Fig. 4-3 can be used to address this question.  Consider an illustrative example of a driver who transfers from an m₃ to an m₄ category car.  If the m₃ car is in collision with, say, an m₁ car, the risk to the m₃ driver is 3.50 and that to the m₁ driver 15.15, for a combined risk of 18.65.  If the m₃ car becomes an m₄ car, the corresponding risks become 2.14 and 16.05, for a total risk of 18.19, which is slightly less than the prior 18.65.  Applying the comparison for each category of car into which the m₃, and later m₄, car might crash reveals that in six out of six cases the net fatality risk is lower when the driver transfers to the heavier car.  Applying the same comparison for all five cases of a driver switching to a car in the next highest mass category shows that in 26 out of 30 cases a net reduction in fatality risk results.  If we do not restrict the mass increase to one mass category, but allow all possibilities (for example, a m₁ driver might switch to a m₆ category car), we find that 78 of 90 comparisons show a net fatality decrease.

            Joksch [1983] fits to state injury and fatality data a relationship which can be expressed as

 

            y =  exp (- 0.001 102 M₁  +  0.000 441 M₂) , Eqn 4-4

 

where y is the relative risk of an injury to an occupant in a car of mass M₁ when it crashes with a car of mass M₂ (masses in kg).  If we add the risk to the driver in car M₁ to the risk to the driver in car M₂ we obtain an expression for the combined risk to both drivers.  Increasing the value of M₁ decreases the combined risk except when cars of over 1300 kg are replaced by yet heavier ones.  Eqn 4-4 and Fig. 4-3 consistently indicate that substituting a heavier for a lighter car nearly always reduces the system-wide harm from two-car crashes, assuming that driver behavior remains unchanged.  The exceptions are likely spurious, especially as the regions where the inversions occur are different for Eqn 4-4 and Fig. 4-3.

            Fig. 4-4 shows information parallel to that in Fig. 4-3, but for head-on crashes only (principal impact point at 12 o'clock).  About 40% of the Fig. 4-2 data are for head-on crashes.  Note that the effect of mass is greater in the head-on than in the all-directions cases. 

----------------------------------------------------------------------------

Fig. 4-4 about here

----------------------------------------------------------------------------

            Of particular interest in Figs 4-3 and 4-4 are the highlighted diagonal elements, which show how the risk of driver death in crashes involving cars of similar mass depends on their common mass.  The data for head-on crashes are plotted in Fig. 4-5, together with data for serious injuries (including fatalities) per registered car from New York [Negri and Riley 1974] and North Carolina [Campbell and Reinfurt 1973].  These data are fitted to Eqn 4-2 to yield b = 0.000 785 kg^-1, with a slightly higher value of 0.000 854 kg^-1 for the all-directions case.  Thus, a driver in a 900 kg car crashing head-on into another 900 kg car is about 2.0 times as likely to be killed as is a driver in an 1800 kg car crashing head-on into another 1800 kg car.  The corresponding value for the all-directions case is 2.2.  Substituting M₁ = M₂ into Eqn 4-4 generates an equation of the same form as Eqn 4-3 with b = -0.000 661 kg ^-1.  Substituting 900 kg and 1800 kg gives that the risk in the small-small crash is 1.8 times the risk in the large-large case.

----------------------------------------------------------------------------

Fig. 4-5 about here

----------------------------------------------------------------------------

            When cars of identical mass crash into each other head-on, mass as such does not influence overall crash dynamics.  The effect here is certainly due to correlates with mass, such as additional crush material and space in the occupant compartment.  The much larger differences in fatality risk to occupants of light and heavy cars when they crash into each other are due mainly to differences in mass as such.

            The above results are interpreted to indicate risk in a crash, given that the crash occurs.  Evans [1982] averaged the relative risk of death in a two-car crash over the mix of cars by mass on the roads (in 1978) and found that the driver of a 900 kg car is about four times as likely to be killed in a "typical" two-car crash as is the driver of a 1800 kg car.  Direct examination of the number of fatalities sustained in two-car crashes per registered car, as discussed below, shows a considerably smaller effect.

            Partyka [1989a] derives relations between occupant fatalities and car mass for different crash types by dividing the number of fatalities from FARS data by the number of registered vehicles as tabulated by R.L. Polk.  She includes a correction reflecting that the same car tends to be coded as about 100 pounds heavier in the registration than in the fatality data [Partyka 1990].  The relationship derived by Partyka [1989a] for cars involved in multiple vehicle crashes (two-car crashes constitute the majority of these) estimates 169.5 occupant fatalities per 900 kg car per year compared to 91.8 for 1800 kg cars, for a ratio of 1.85.  In relating deaths per registered car to the probability of death given a crash, it is more appropriate to concentrate on driver fatalities per crash than occupant fatalities per crash.  The larger occupancy rates associated with larger cars places more occupants at risk, which increases fatality rates.  Car mass effects are generally best discussed in terms of driver risk, with the reasonable assumption that the risks scale proportionately for other occupants.  If a large car is replaced by a smaller car in a specific family context, occupancy will not necessarily decline.  Partyka [1989b] repeated the same analysis for drivers, deriving a relationship for cars involved in multiple vehicle crashes which estimates 113.6 driver fatalities per million 900 kg cars compared to 58.5 for 1800 kg cars, for a ratio of 1.94.

            The results for relative risk in crashes (about a factor of four between 900 kg and 1800 kg cars) and fatalities per registered vehicle (about a factor of two) are not inconsistent -- they measure different phenomena [Evans 1985a].  The relative risk when cars crash into each other depends only on engineering and biomechanics.  The fatalities per vehicle rate depends additionally on crash involvement rates.  Based on an analysis of over 100 000 police-reported two-car crashes in New York State in 1971 and 1972, Evans [1985b] finds that the number of two-car crashes per registered car increased systematically and strongly with car mass; the quantitative specifics of this study may not apply beyond the early 1970's because small cars at that time in the US were of a rather different nature than in later years.  The early 1970's data gave that, on a per registered car basis, 900 kg cars crashed into 900 kg cars at only 30% the rate that 1800 kg cars crashed into 1800 kg cars.  Thus, notwithstanding the greater fatality risk when the two small cars crash into each other, the total number of fatalities per registered car in such crashes is actually estimated to be less because the reduced crash rate more than offsets the increased fatality risk in the crash.

 

Single-car crashes

 

            There is no simple reasoning based on elementary physics why fatality risk should depend on mass in single-car crashes.  Accordingly, effects are expected to be smaller, and exposure estimates are critical.  Some early studies examining quantities such as the number of injuries per crash proved inconclusive.  A major problem here is that a single-car crash is included in a data file only if it is, say, reported to the police.  Yet if the vehicle is drivable after the crash, the driver may decline to inform the authorities.  As larger cars are less likely than smaller cars to be immobilized by the same crash, the same physical characteristics that reduce injury would also deplete the exposure measure.  The same comment applies with greater emphasis to such measures as deaths per injury-producing crash, or serious injuries per minor-injury crash; lower values may mean fewer serious injuries, or more minor injuries.

            While a number of studies of single-car crashes [Partyka 1989a; c; McCarthy et al. 1989; Evans 1982] do show that fatalities per registered vehicle decrease with increasing car mass, such measures reflect crash involvement rates as well as the risk of death in a crash.

 

The pedestrian exposure approach  As the most common object struck in a single-car fatal crash is a tree, let us consider cars hitting trees.  FARS provides the number of drivers by age and sex killed in cars of known make, model year, and mass crashing into trees.  What is not known is how many car/tree crashes occur in which no one is killed, because unless there is a fatality, the crash is not coded in FARS.  We get round this problem by using crashes into pedestrians as a surrogate for crashes into trees; such crashes are coded in FARS provided the pedestrian is killed.  We proceed by assuming that when a car and a pedestrian crash, the probability that the pedestrian is killed is independent of the mass of the car.  This is a reasonable assumption on physical grounds because the lightest car is so much heavier than the heaviest pedestrian that the car trajectory is relatively unaffected by the collision.  As the crash forces on the pedestrian are independent of car mass, the number of pedestrian fatalities is consequently proportional to the number of severe crashes.

            Thus we conclude that the number of pedestrian fatalities associated with cars in some category (say, in the same mass range) is a measure of the exposure of that group of cars to severe crashes in general.  If the category includes more cars, proportionally more pedestrian fatalities will result.  Similarly, if these cars are driven more, are associated with younger drivers, with riskier drivers, with more alcohol abuse, etc., then we would expect all these factors to increase the number of pedestrian fatalities and the number of severe crashes into trees in similar proportions.  When we take the ratio of driver to pedestrian fatalities, all the factors relating to number of cars, use and behavioral factors (that is, exposure) appear as multiplicative factors in both denominator and numerator, and accordingly cancel, leaving the ratio as a measure of driver fatality risk in a single-car crash as a function of the properties of the cars in the particular category chosen.

  Fig. 4-6 shows the ratio of drivers killed in single-car crashes to pedestrians killed in crashes involving similar mass cars, separated into three driver age groups.  The similarity of the three plots supports the interpretation that it is mainly car-mass effects that are shown, because there is no hint of the large age effects discussed in Chapter 2.

----------------------------------------------------------------------------

Fig. 4-6 about here

----------------------------------------------------------------------------

            The same approach may be applied using motorcyclist fatalities rather than pedestrian fatalities to estimate exposure.  Also, the fatality risk to drivers in cars crashing with large trucks may depend on car mass in a similar manner to that for single-car crashes.  Thus we have two types of crashes in which drivers are killed (single-car and car-truck) evaluated by two exposure measures (pedestrian and motorcyclist fatalities) and three driver age categories, generating in all twelve separate relationships with car mass.  The close quantitative agreement found [Evans 1984a] between all 12 plots indicates highly robust phenomena.  In addition, the number of drivers killed in single-car crashes divided by the number killed in car-truck crashes is independent of mass, as is the number of pedestrian fatalities divided by the number of motorcyclist fatalities.  All these results support the interpretation given that the ratio of driver deaths to pedestrian (or motorcyclist) deaths reflects the physical effect of car mass, essentially independent of driver behavior effects.

            There are sources of error.  While pedestrian fatality risk does not depend on car mass, it may depend on the details of hood geometry and other car details that might be correlated with car mass.  There are some indications that smaller cars may be less likely to injure pedestrians [McLean 1972] and others that the reverse may be so [Mackay 1985, p. 353].  The driver  and pedestrian fatalities could take place at different times and places in ways that could invalidate the method.  However, repeating the analyses for data divided by time of the day, and by rural versus urban crashes, provided essentially similar relations with car mass [Evans 1984a].  Larger size, as such, does of course increase the probability that the car strikes the pedestrian; but the same reasoning applies equally to striking the tree.

            The same pedestrian exposure approach is applied for drivers coded in the FARS data as using safety belts [Evans 1985c].  Because of fewer data, drivers killed in either single-car or car-truck crashes are included in the numerator, and pedestrian and motorcyclist fatalities are included in the denominator.  The result for unbelted drivers is shown in Fig 4-7 computed in the same way.  Note how similar the values of b are for each case, showing that the influence of car mass on fatality risk is essentially the same for belted and unbelted occupants.  The difference between these curves, although  related to belt effectiveness, should not be interpreted to measure belt effectiveness because of large reporting biases, which are assumed to not depend on car mass.  As belts are not exclusively a vehicle factor -- they only have influence when worn -- we discuss them, together with related passive restraint systems, in Chapter 9.

            The data in Fig 4-7 indicate that in single-car crashes a driver in a 900 kg car is 2.4 times as likely to be killed as is a driver in a 1800 kg car, assuming both drivers are unbelted.  When both drivers are belted, the corresponding ratio is 2.3.  Subject to the assumptions of the pedestrian exposure approach, this result flows from differences in the cars and is unrelated to driver behavior.  If, given a pedestrian crash, the pedestrian is more likely to die if the car is large, then the mass effects estimated will be too large.

            We compare the number of deaths per registered car in the same way as previously done for two-car crashes.  Partyka's [1989a] equation for occupants estimates 112.6 fatalities per million 900 kg cars per year compared to 79.3 for 1800 kg cars, for a ratio of 1.42, which may be compared to the value 1.48 reported by Evans [1982].  Partyka's [1989b] equation for drivers estimates 79.3 fatalities per million 900 kg cars compared to 53.9 for 1800 kg cars, for a ratio of 1.47, which may be compared to the value 1.68 reported by Evans [1982].  The higher values in the earlier study reflect the discrepancy in the mass definitions which are corrected for in the later estimates.

            As with two-car crashes, the finding that the estimate for the physical effect is larger than that for the fatalities per registered vehicle does not imply inconsistency, but rather points to differences in crash rates (and other factors) dependent on car size.  Such effects are directly measured [Evans 1984b] using data for all crashes from North Carolina (1979), New York State (1971 and 1972) and Michigan (1976).  For each state the data are partitioned into three driver age categories, thus providing nine crash involvement rate versus car mass relationships; all nine indicate higher crash rates for larger cars.  The average result is that, when driven by drivers of the same age, 900 kg cars are involved in 28% fewer crashes per unit distance of travel than 1800 kg cars.  Let us assume that this result, based on all crashes, most of which are minor property damage crashes involving two cars, applies also to single-car crashes.  If one takes the physical effect as being a 2.4 ratio, the expected number of fatalities per registered car is given by 2.4 X (1-0.28) = 1.72, which may be compared to the directly measured 1.47 ratio.  There is direct observational evidence that larger cars are associated with higher levels of driver risk-taking, as indicated by higher travel speeds and closer following headways [Wasielewski and Evans 1985].

            It is not quite correct to assume that the risk of death in a crash times the number of crashes per unit distance of travel gives the number of crashes per registered vehicle because a number of other factors are also related to car mass.  Although I am unaware of quantitative estimates, it is recognized that smaller cars tend to travel less and be more associated with urban driving than larger cars; both these effects will reduce the relative number of driver fatalities in smaller cars.  On the other hand, small cars are more likely to be driven by younger drivers [Evans 1985d] who have higher crash rates, but the effect of this will be somewhat lessened by the higher survivability of the younger drivers.

 

Rollover  Partyka [1989a; c] partitions single-vehicle crashes into those involving rollover and those not.  The regression equation for crashes with rollover estimates 39.5 driver fatalities per million 900 kg cars compared to 21.4 for 1800 kg cars, for a ratio of 1.84.  The effect for crashes not involving rollover is smaller (37.2 for 900 kg compared to 32.5 at 1800 kg for a ratio of 1.15) but still statistically significantly different from zero (the corresponding difference for occupants rather than drivers is not statistically different from zero [Partyka and Boehly 1989]).  For the 900 kg car, almost half of the single-car fatalities involve rollover.  Further evidence of the relationship between mass and rollover is provided in findings that the fraction of fatalities that would be prevented by eliminating ejection is larger for smaller cars [Evans and Frick 1989a].

            Rollover risk does not flow from mass as such.  One vehicle manufactured from materials a constant fraction more dense than another vehicle would have the same rollover resistance under the same conditions.  Likewise, a vehicle that is the same shape, but smaller, than another would similarly have the same rollover resistance.  However, small vehicles are not scaled down versions of larger ones because of the need to accommodate their human occupants.  Thus, as vehicles become smaller, the height of the center of gravity above the ground must remain relatively constant, whereas the track width, the distance between the paths of the left and right wheels, decreases.  Simple physics indicates that rollover resistance should be related to the ratio of the track width to the height of center of gravity above the ground.  Stonex [1962] defined a "stability factor" as half this ratio, namely, the distance from the point under the vehicle's center of gravity to a line along which the wheels travel, divided by the height of the center of gravity above the ground.  Mengert et al. [1989] find that the fraction of all single-vehicle crashes that are rollovers is related strongly to this stability factor, with observed values of the fraction of single-vehicle crashes that are rollovers varying from below 10% for vehicles with a stability factor of over 1.4 to over 30% for vehicles with stability factors under 1.1.  Robertson [1989] finds that rollover crashes per registered vehicle also increases with decreasing stability factor.

            For vehicles of fixed mass, those with lower stability factors have lesser rollover resistance; for the same stability factor, heavier vehicles provide more crash protection.  As for the case of car mass in two-vehicle crashes, it is not surprising that vehicles obey the laws of physics.  One interesting difference is that there has been advocacy for regulation to restrict the height of the center of gravity above the ground for a vehicle of given track width [Robertson 1989], yet none to place lower limits on car mass.  The mass and center of gravity of a vehicle are chosen based on many considerations and constraints.  Knowing that a higher center of gravity or a lighter vehicle implies a reduction in safety is hardly sufficient reason to justify an arbitrary cut-off criterion.

 

Summary of vehicle-mass effects

 

            The above vehicle mass effects are summarized in Table 4-2 in terms of the example of a small 900 kg car and a large 1800 kg car.  The results for these cars crashing into each other are based on simple interpolations from the data in Figs 4-3 and 4-4 [Evans 1986].  The result that for all crashes the fatality risk in a 900 kg car is 2.8 times what it is in a 1800 kg car is a simple weighting (by 0.75 and 0.25) of the single-car and two-car effects to reflect the relative occurrence of non-two-car crashes and two-car crashes [Evans 1989].  The value of 1.7 for driver fatalities per registered car [Partyka 1989b] is not materially different from the value of 1.8 for occupants (drivers would have been a more appropriate choice, as discussed above) used in Schwing, Evans, and Schreck [1983], which is based on Evans [1982]. 

            In order to be in compliance with Corporate Average Fuel Economy (CAFE) standards mandated by the Energy Policy Conservation Act of 1975, the average mass of US vehicles declined.  Using some of the relations in Table 4-2, Crandall and Graham [1989] estimated that as a result of these mass reductions, 1989 model-year cars would be responsible for 2200 to 3900 additional fatalities in the ten years following their introduction.

----------------------------------------------------------------------------

Table 4-2 about here

----------------------------------------------------------------------------

 

OTHER VEHICLE CHARACTERISTICS

 

            Many vehicle modifications aimed at increasing safety have been introduced over the years.  In many cases these were first introduced by one vehicle manufacturer, but were later required by Federal Motor Vehicle Safety Standards (FMVSS), often in modified form, to apply to all vehicles sold in the US.  The effectiveness of specific FMVSS standards in reducing fatalities can be estimated by comparing fatality rates for vehicles of model year prior to introducing the standard to the corresponding fatality rates after introduction.  The estimates discussed below are all by the National Highway Traffic Safety Administration [Kahane 1984], the agency responsible for the standards.

 

Federal Motor Vehicle Safety Standards (FMVSS)

 

            There are three main categories of standards that apply to 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 immediate post-crash considerations.  The National Traffic and Motor Vehicle Safety Act of 1966 directed that all vehicles manufactured in 1968 or later satisfy a number of these standards; additional standards continue to be promulgated.

            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 impact in a frontal crash, and FMVSS 204 limited the rearward displacement of the steering wheel towards the driver.  These standards, which became effective in 1968, aimed at reducing driver fatality risk in frontal crashes, but were not aimed at changing passenger risk, nor driver risk in non-frontal crashes.  Hence, their effect might be detected in two ways; by comparing right-front passenger to driver fatality risk, or by comparing driver risk in frontal crashes to driver risk in non-frontal crashes.  Kahane [1981; 1982a] used FARS data from 1975 through 1979 to perform such analyses, obtaining estimates of 13% and 11%, respectively.  Although these estimates are not independent, in that each uses the same driver fatalities, the agreement nonetheless suggests a fairly robust effect.  Thus Kahane's [1981] conclusion of a 12.1% effectiveness, with confidence bounds from 8.5% to 15.5%, seems well supported by the data and analysis presented, although Evans and Frick [1989b] suggest that the estimate could be somewhat high because of possible concurrent changes in the control crashes.  A 12.1% reduction in fatality risk in frontal crashes, which in the definition used in Kahane [1981] constitute about 54% of all fatal crashes, implies a net 6.6% reduction in driver fatalities, or, when averaged over all car occupants, a 4.4% reduction (Table 4-3).

----------------------------------------------------------------------------

Table 4-3 about here

----------------------------------------------------------------------------

            Kahane [1988] estimates that improvements in instrument panels in the 1965-1975 era reduced fatality risk by about 13% for unrestrained front passengers in frontal crashes, or a 7% reduction for all crashes.  As right-front plus center-front passengers constitute 24% of all car-occupant fatalities (Fig. 3-4), this reduces car-occupant fatalities by about 1.7%. 

            Side door beams (FMVSS 214) are estimated by Kahane [1982b] to prevent 480 fatalities, or, based on the number in 1980, about 1.7% of car-occupant fatalities.  Head restraints for drivers and right-front passengers are estimated to reduce overall injury risk in rear impacts by 12% [Kahane 1982c].  Let us make the very approximate assumption that this applies also to fatalities, about 3% of which result from rear (principal impact point 6 o'clock) impact, so we obtain a net reduction in outboard-front occupant fatalities of 0.36%, or 0.33% of all occupant fatalities.  Adhesive bonding is estimated [Kahane 1985] to halve windshield bond separation and occupant ejection through the windshield, thereby preventing 105 fatalities.  As an approximation, we express this as an average risk reduction of 0.39% for all occupants.  Kahane [1990] estimates that improved door locks and door retention components (FMVSS 206) and improved roof crush resistance (FMVSS 216) reduced occupant fatalities by 400 and 110 per year, or 1.5% and 0.43%, respectively.  One measure aimed at crash prevention rather than occupant protection, namely dual master brake cylinders, is estimated [Kahane 1983] to prevent 260 fatalities, or 0.9%.

 

Combined effect of all standards

 

            Although summing the last column in Table 4-3 gives an approximate estimate of the total reduction in occupant fatalities from the combined effects of all the changes listed, from a formal point of view, it is an inappropriate calculation which would lead to serious errors if the percent reductions were larger.  If, say, better brakes prevent a crash, then the contribution of the (say) energy absorbing steering column must not be included for that crash which did not occur.  The application of two measures which successively reduce something by 50% do not eliminate it, but reduce it by 75%.

            In order to address the combined effects of all the measures, one must compute the effect for each occupant in the vehicle, and then compute a weighted sum over all occupants.  Seven of the eight items in Table 4-3 reduce driver fatalities.  Their combined effect is to reduce driver fatalities by

 

 =  1  - (1-0.066)(1-0.017)(1-0.015)(1-0.0043)(1-0.0039)(1-0.0036)(1-0.009)

 

                    =   11.43% .         Eqn 4-5

 

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% (Fig. 3-4).  By weighting each reduction by the corresponding occupancy, the combined effect of the six measures in Table 4-3 is estimated to reduce car occupant fatalities by 10.9%.

            Table 4-3 does not include all changes which may have reduced car-occupant fatalities, but only those for which fatality reductions have been estimated.  Other FMVSS standards may be associated with fatality reductions too small to have been measured, yet important in terms of total numbers.  Much engineering attention has been focused on energy management during crashes, with finite element techniques being applied to design vehicle structure in front of the occupants to crush in ways that transmit the least damaging forces to occupants.  The requirements that the forces on belted anthropomorphic dummies in barrier crash tests (FMVSS 208) be within specified limits has helped stimulate developments in these areas, although the performance on such tests as such has not been found to be related to field fatality rates [Grush, Marsh, and South 1983].  Many changes addressing specific injury modes in specific crashes have been made based on engineering judgment.  If such a change prevents, say, 10 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 all the effects missed, this implies that vehicle changes have reduced occupant fatality risk by 16%.  Many automotive engineers consider that the cumulative effect of vehicle changes from the early 1960's to the present have reduced car-occupant risk by somewhere in the range 10% to 20%, a range consistent with the above discussion.

 

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 calender year for FARS data is 1975, by which time the newest cars unaffected by changes incorporated in FMVSS standards, namely 1966 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 new cars by larger amounts than are expected to be associated with vehicle design standards.

            Robertson [1981] estimates 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.  Robertson [1981, p. 820] concludes, "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). 

            The same 1975-1978 FARS data used in Robertson's analysis show 4665 motorcyclists (3975 drivers and 690 passengers, more wearing helmets than not wearing) killed in crashes involving model year 1968 or later cars; these fatalities constitute less than a third of all motorcyclist fatalities -- the most common crash mode involves no vehicle other than the motorcycle.  Robertson's conclusion thus implies that without the regulations, motorcyclist fatalities in crashes with MY 1968 or later cars would have been about 6665 instead of 4665; a 30% reduction in motorcyclist fatalities is therefore attributed to changes in these cars.  An inference that car safety standards had reduced motorcyclist fatality risk by 30% should have immediately brought into question the face-validity of the model fitted to the data.

            Orr [1984] addresses the validity of the model directly by performing additional multivariate analyses on Robertson's data and concludes that most of the reduction in fatalities is more appropriately attributed to car age; older cars, which are driven by younger drivers, have substantially higher crash-involvement rates than newer cars.  Orr estimates that the total fatality reductions attributable to vehicle changes are in the range zero to 9200, less than one quarter those found by Robertson.  Lower crash-involvement rates for newer, post-regulation, cars provides a more plausible explanation for their substantially lower involvement rates in pedestrian, pedalcycle and motorcycle fatality crashes than do federal regulations aimed mainly at protecting car occupants.  Orr [1984] also points out the inappropriate inclusion of trucks, which were not subject to regulatory changes, and other methodological problems.  In response, Robertson [1984] applies a different model, this time reporting even larger effects;  for example, fatality reductions of 15 311 in 1979 and 15 909 in 1980.  Orr [1985] responds by claiming that the new estimate contains the same basic flaw as the original -- it is primarily a car-age effect that has nothing to do with regulation.  Indeed, Robertson's [1984, p. 1392] own Figure 1 showing fatality rate versus model year shows no indication of any decline from MY 1966 to MY 1968 sufficiently in excess of the trend to generate the claimed reductions.  Robertson [1985] responds citing, "irrefutable evidence of the effectiveness of seat belts, energy absorbing steering mechanisms, etc."  Given that safety belts were preventing less than one thousand fatalities in 1980 [Partyka 1988; Evans 1987], and that Kahane [1981] estimates that all cars having energy absorbing columns and limited column displacement would reduce fatalities by 1300, it is difficult to see how the "etc." is going to account for the more than 13 000 additional fatalities alleged to have been prevented! 

            Simple graphical presentations by Adams [1985a] further demonstrate the absence of any large change in occupant fatality risk coincident with the introduction of the safety standards.  Applying the pedestrian exposure approach to compare post-1968 and pre-1966 MY cars shows rates incompatible with any change of the magnitude claimed by Robertson.

            Another attempt to estimate the aggregate effect of federal regulations on fatalities is that of Peltzman [1975], who uses data from 1947 to 1965 to project fatality rates for the first seven years of federal safety standards, and then compares these estimates to actual values.  He concludes that the net effect of the standards is essentially zero; a small reduction in deaths to car occupants is balanced by a corresponding increase in deaths to non-occupants.  Unlike the Robertson [1981; 1984] estimates, a mechanism is in this case offered to explain the alleged effect.  The explanation is that safer vehicles increase driver risk-taking, thereby reducing, but not eliminating, the benefits to car occupants, but increasing the risk to non-occupants.  Chapter 11 is devoted to discussing such behavioral feedback responses to safety measures -- suffice it to comment here that the changes associated with the safety standards are so invisible to most drivers that any large behavior response to them is unlikely.  Peltzman's [1975] paper has been much discussed in the literature [Joksch 1976b; 1976c; Robertson 1977; Peltzman 1976; Graham 1984; Crandall and Graham 1984; Graham and Garber 1984; Zlatoper 1984], and the same, or similar, data have been shown to lead to a wide variety of conclusions in the hands of others.  This illustrates what seems to me to be an intrinsic problems with complicated multivariate analyses.  There are so many choices of variables and of transformations at the discretion of the analyst that the detached reader has rarely 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 abstract issues as whether to use the logarithm or the square of the dependent variable in the model specification.

            The two estimates discussed above, one that the safety standards prevented zero deaths and the other that they prevented over 37 000 deaths, both probably reflect the triumph of zeal over science, or perhaps even common sense.  An enthusiastic editorial in the American Journal of Public Health [Yankauer 1981] states, "It is good to know that in 1975-1978, the automobile safety standards laid down by the federal government some years earlier resulted in the saving of 37,000 lives."  There has been no subsequent editorial to set the record straight.

 

EFFECTS OF ROADWAY

 

            Table 4-4 shows the number of fatalities (all road users), and the rates per unit distance of vehicle travel for different types of roads.  The distinction between federal-aid and non-federal aid is based on the manner in which the roads are financed, rather than strictly on physical characteristics.  The same road can be changed from one classification to another administratively, without any physical changes in the road.  Generally, the more major the road, the more likely it is to receive federal aid.  The classifications for the roadways are the same as those coded in FARS data, the source of the fatality data.  The Interstate category represents the most homogeneous physical system, with all Interstate roads being limited access freeways, with at least two lanes of traffic in each direction being well separated from each other.

----------------------------------------------------------------------------

Table 4-4 about here

----------------------------------------------------------------------------

            For all roadways for which a comparison can be made, rural fatality rates are substantially higher than urban rates.  By far the lowest fatality rate, 5.5 deaths per billion km, occurs on the urban Interstate system.  The rate on the rural system, 9.7 deaths per billion km, is 76% higher.  This difference is primarily due to different use patterns on the two types of roads.  Much of the travel on the urban system is high flow commuting travel under congested conditions which constrain speed.  Alcohol is not prevalent during commuting hours.  In contrast, lower traffic densities and less commuting traffic on the rural system permit higher speeds.  The 65 mph speed limit on some portions of the rural system compared to 55 mph on the urban also contributes to the difference; however, from 1974 to 1986, when the maximum speed limit throughout the entire system was 55 mph, the rural rate still exceeded the urban rate by substantial amounts, though less than the proportionate difference for 1988.

            After controlling for the urban compared to rural difference, the differences in fatality rates are much more linked to the physical nature of the roadway.  The much higher fatality rates on two-lane roads are associated with head-on crashes involving cars travelling in the opposite direction, with striking trees and other objects close to the trafficway, with intersection crashes, and with pedestrian impacts.  All these crash-types are absent from Interstate (or other) freeways.  Each of the non-Interstate categories in Table 4-4 contains a mix of types of roadways, some with lower and some with higher fatality rates than the average.

            The highest fatality rate in Table 4-4 is 825% above the lowest.  If one confines the comparison to rural travel only, then the highest rate is 425% above the lowest; for urban only, the highest rate is 120% above the lowest.  If the rate on the urban Interstate system applied for all travel, then the total fatalities would be 18 019 instead of 47 093, a reduction of 62%.  If all urban and all rural travel were at the same fatality rate as the corresponding Interstate rate, then fatalities would be 23 491, a reduction of 50%.

            These calculations bring out the enormous influence of roadway on safety.  They do not imply that if all roadways were upgraded to Interstate standards, the calculated reductions would occur.  To start with, a world without local streets would be hard to imagine.  More specifically, the upgrading of a roadway does not simply substitute a new lower fatality rate for a prior higher fatality rate.  Upgrading roads reduces congestion and delay, which, in time, generates increased travel [Mackie and Bonsall 1989], with consequent higher than calculated fatalities if the fatality rate remains unchanged.  Although interactive effects unquestionably occur and are substantial, the differences between fatality rates on different types of roads are so great that it is beyond reasonable doubt that replacing, say, a well-travelled rural two-lane road with a limited access freeway will reduce traffic deaths and injuries.

 

Traffic engineering

 

            Traffic engineering changes, such as installing traffic lights or stop signs, have many motivations in addition to safety.  Surprisingly, the influence of such devices on safety is not all that clearly established [Persaud 1988; Hauer 1987; 1988].  There is a general theme in the traffic engineering literature that traffic control devices enhance safety, but definitive evidence is difficult to generate.  An intrinsic problem is that at any particular site, traffic crashes are rare events, so that it is extremely difficult to get enough "before" and "after" data to support reliable conclusions.  Sites selected for treatment generally have much higher than average crash rates; because of "regression to the mean", these rates would tend to be lower in subsequent years regardless of treatment [Hauer 1980], thus adding to the problem of satisfactorily evaluating the safety effect of treatments.  One change for which there is clear evidence of large reductions in crash rates is replacing two-way by four-way stop signs [Hauer 1985].

 

ENVIRONMENTAL FACTORS

 

Weather - variations by season

 

            Fig. 4-8 shows the number of fatalities per calender month for the six years 1983-1988.  A clear cyclical pattern is apparent.  Total fatalities typically peak in August or September, and have minima in February.  Pedestrian fatalities are greatest in December, and least in January.  Motorcycle drivers contribute to the cyclical pattern for drivers -- the pattern for car drivers is more like that for pedestrians than for all drivers.

----------------------------------------------------------------------------

Fig. 4-8 about here

----------------------------------------------------------------------------

            Part of the reason for the cyclical pattern is, of course, that travel is greater in the summer months.  Fig. 4-9 shows the data in Fig. 4-8 for all fatalities divided by total distance of travel.  Nominally, this normalization should remove effects due to different amounts of travel, and also from different numbers of days in the month.  Note that while total traffic fatalities were on an upward trend in the six years (Fig. 4-8), the trend for the fatality rate was downwards. 

----------------------------------------------------------------------------

Fig. 4-9 about here

----------------------------------------------------------------------------

            In order to illustrate better the cyclical behavior of fatalities, Fig. 4-10 shows the fatality rate for each month relative to the yearly average; the average for the 12 values for each year is one.  A highly regular pattern is apparent.  The lowest fatality rates occur consistently in the winter months, notwithstanding the increased adverse factors of darkness, snow and ice.  The highest rates occur in the summer and fall months.

----------------------------------------------------------------------------

Fig. 4-10 about here

----------------------------------------------------------------------------

            FARS data for 1988 show that 8.3% of fatal crashes occurred in rain, 1.9% in snow, and 1.5% in fog or other unusual atmospheric conditions; thus the vast majority (88.3%) of fatal crashes occurred in the absence of any adverse atmospheric conditions.  In addition, 83.6% of fatal crashes occurred on dry roadway surfaces; the remainder are 12.8% on wet surfaces, 1.6% on snow, 1.8% on ice, and 0.3% on sand or other surface.  A study of injury-producing crashes in Leeds, UK, finds that 81% occurred under fine weather conditions, 69% occurred in daylight, and 61% occurred when the road was dry [Carsten, Tight, and Southwell 1989].  Information on the amount of travel as a function of roadway surface or atmospheric conditions is unavailable, so it is not possible to determine directly how these factors influence fatality risk.

            Adams [1985b] presents data on how injuries and fatalities in Ontario, Canada depend on the month of the year.  There, where the winter is much more severe than the average for the US, the seasonal variation is greater.  The numbers of fatalities in the summer months are about 100% higher than in the winter months (compared to 30% for the United States).  Adams [1985b] also plots the ratio of fatality to injury crashes for Ontario as a function of month, with the results shown in Fig. 4-11.  The data in Adams [1985b] are further tabulated by road surface condition, as reproduced here in Table 4-5.  Mueller, Rivara, and Bergman [1987] find that, given a pedestrian injury crash, the probability that the pedestrian is killed is greatest when the road surface is dry, and least when ice covered; when the visibility is clear, pedestrian fatality risk in a crash is twice what is when it is snowing.

----------------------------------------------------------------------------

Fig. 4-11 about here

----------------------------------------------------------------------------

----------------------------------------------------------------------------

Table 4-5 about here

----------------------------------------------------------------------------

            The above data associate highest fatality rates with dry roadway surfaces  and favorable atmospheric conditions.  There are insufficient exposure data to determine to what extent crash rates depend on roadway surface, but they almost certainly are higher with reduced roadway friction and visibility.  Driver responses to inclement weather and slippery roadways, especially reduced speeds, lead to more but less severe crashes, with consequent reductions in fatalities.  Thus through road-user responses, to be discussed more fully in Chapter 11, environmental factors reduce mobility but, in terms of fatalities, actually increase safety.

 

Darkness - variations by day and hour

 

            Schwing and Kamerud [1988] examine traffic fatality risk for each of the 7 X 24 = 168 hours in the week (Fig. 4-12).  The top graph shows the distribution of fatalities to occupants of cars and light trucks, based on FARS data for 1983.  If fatalities were equally likely at all times, then 1/168, or 0.595% (indicated by a dotted line), would occur in each hour of the week.  Fatalities are in fact distributed in a far from uniform manner, with particularly large peaks on late Friday/early Saturday and late Saturday/early Sunday.  Other days have lesser peaks in the afternoon and around midnight.  Fatalities are lowest at between 4:00 a.m. and 5:00 a.m. on weekdays, and 8:00 a.m. and 9:00 a.m. on weekends.

----------------------------------------------------------------------------

Fig. 4-12 about here

----------------------------------------------------------------------------

            The center graph shows the distribution of travel (total distance travelled by vehicle occupants).  This is derived from a survey [Horowitz 1986] of the travel activities in 2000 households during a one week period.  Each household member recorded the starting time, duration, purpose, whether a driver or passenger, etc. for each trip taken.  The cycles in the two curves are essentially out of phase -- the greatest numbers of fatalities tend to occur at times associated with the least amount of travel.

            The bottom graph shows the fatalities per unit distance of travel relative to a value of unity for the average rate over the entire week.  The variations are so large that the data are shown in a log representation.  The 168 values of relative risk vary from a low of 0.32 (between 10:00 a.m. and 11:00 a.m. on Sunday morning) to a high of 43 (between 3:00 a.m. and 4:00 a.m. on Sunday morning); the ratio between these indicates that the most dangerous hour is 134 times as dangerous as the safest hour.

            Many factors contribute to the wide variation in risk.  Another study [Mortimer and Fell 1988] finds that drivers in different age categories, including older than 65 years, have higher crash rates between midnight and 6:00 a.m. than during daytime periods.  As alcohol is less of a factor for older drivers this finding could point to darkness as a possible factor.  However, the data in Fig. 4-12 show that weekend rates at 3:00 to 4:00 a.m. are well over ten times greater than at the similarly dark period 10:00 to 11:00 p.m., so that darkness is not the main contributor to the pattern.  Local minima in risk occur during the morning and afternoon rush periods, in part because congestion, through speed reduction, reduces fatality, though not crash, risk.

            The risk pattern refers to the relative risk sustained by the mix of drivers, by age, purpose of trip, alcohol use, etc. who are driving at the indicated times; it does not indicate how the risk varies for an individual driver.  Let us imagine a specific individual driver with unvarying sober careful behavior, and ignore possible influences from effects of darkness, fatigue, etc.  The crashes that contribute to the peaks are nearly all single-vehicle crashes.  There is no reason why our hypothetical driver should dramatically increase single-vehicle crash risk just because it is 3:00 a.m. on Sunday morning.  It seems plausible to presume that for this driver, the single-vehicle crash risk would remain constant, and low.  His risk of being killed in a multiple-vehicle crash is proportional to the probability that another vehicle strikes his.  This can be assumed to be proportional to the probability that the other vehicle is involved in any type of serious crash, which, if measured by fatalities, is given by the top graph in Fig. 4-12.  If we assume that our hypothetical driver is so careful that the risk of a single-vehicle crash is zero, then his fatality risk depends on time and day in a similar manner to the top graph in Fig. 4-12.  Thus his fatality risk at 3:00 a.m. on Sunday is about three times its average value provided he maintains his own driving constant and careful.

            What Fig. 4-12 shows with dramatic clarity is large variation in risk in a system in which the engineering is largely constant.  Environmental factors may contribute to the variation, but cannot come close to explaining it all. There can be little doubt that the main contributor to the risk pattern is road-user rather than engineering in origin, with such factors as alcohol (Fig. 7-5) and youthful driving playing crucial roles. 

 

STUDIES TO IDENTIFY FACTOR CONTRIBUTIONS DIRECTLY

 

            In the 1970's two major studies, one in the US and one in the UK, were performed to identify factors associated with a large sample of crashes.  The US study was performed by Indiana University, and is often referred to as the "Tri-level study" because crashes were examined in one of three levels of depth, depending mainly on their severity; the study has been described in many detailed reports, with Treat [1980] providing a succinct description of the methods and results.  The British study was performed by the Transport and Road Research Laboratory, and is described by Sabey and Taylor [1980] and Sabey and Staughton [1975].  In both studies a team of multi-disciplinary experts conducted a detailed post-crash examination of crashes satisfying specified selection criteria.  The crash site was examined for physical evidence, the vehicles involved were examined by an engineer, and the participants in the crash were interviewed in depth.  Based on such information, factors contributing to the crash were identified.

            Rumar [1985] elegantly summarized the results from both studies in one figure reproduced as Fig. 4-13.  The interpretation is that, for example, in the US study, the vehicle is identified as the sole factor in 2% of crashes, the interaction between vehicle and road user in 6% of the crashes, the interaction between vehicle, road user, and environment in 3% of crashes, and the interaction between vehicle and road in 1%; the corresponding values for the UK study are 2%, 4%, 1%, and 1%, respectively.

----------------------------------------------------------------------------

Fig. 4-13 about here

----------------------------------------------------------------------------

            The studies were performed independently  -- indeed it appears that neither study group was aware of the activities of the other.  The results are remarkably consistent.  Each finds that the non-road-user factors of vehicle and environment are rarely the sole factors associated with a crash -- the British study finding 5% of crashes not linked to the road user, and the US study 6%.  In both studies, when only one factor is identified, it is overwhelmingly the road user (65% in the British study, 57% in the US study).  The British study finds that road user factors are present as sole or contributory factors in 95% of crashes, the US study 94%.

            Carsten, Tight, and Southwell [1989] use a somewhat similar approach to study injury-producing crashes in Leeds, UK.  Their exclusively urban sample of injured road users contained 24% pedestrians or pedalcyclists.  While differences in categorization and method make it impossible to compare directly their results with the two prior studies, their data do suggest that vehicle factors played an even smaller role (about 1%).  They associate perceptual or judgment errors with essentially all of the crashes.

            Many words of caution have been offered [for example, Shinar 1985, p. 166] regarding interpreting the findings of studies such as the three above.  The identification of factors implies only that if that factor, or combination of factors, had been absent on this specific occasion, this particular crash would not have occurred.  Suppose a head-on collision resulting from improper overtaking at too high speed occurred on a dry, well-lit roadway.  It is unlikely that any factors other than road-user factors would be associated with this crash; yet such a crash would not occur on a divided highway.  If a driver's neglect of vehicle upkeep culminated in a tire failure that preceded a crash, it is unlikely that any factors other than vehicle factors would be associated with this crash.  Also, outcomes are not identified in Fig. 4-13; the same crash, with identical factors, might have different outcomes dependent on vehicle size and belt use. 

            Identifying the mix of factors is not the same as identifying the mix of countermeasures.  This point can be illustrated building upon the example, invoked by Haddon [1972], of sending fragile packages in the mail.  If some fraction of these arrived damaged, multi-disciplinary investigation would doubtless discover that in almost all cases the damage resulted from improper handling by postal employees.  However, this does not logically imply that the most effective remedy is to attempt to upgrade the handling practices of the employees by training and motivation.  Better packaging might achieve larger benefits at less cost.  No general principal can be inferred from this example; in each case countermeasures must be evaluated in terms of potential effectiveness, and benefit and cost information not suggested in Fig. 4-13.

            The consistent findings in the two studies of the dominant role of road- user factors does complement other results derived in this chapter showing that attempts to measure the influence of non-road-user factors often encounter much larger influences from road-user factors.  The recurrent central role of the road-user as a factor in traffic safety leads us to treat this in detail in the next two chapters.

 

CONCLUSIONS

 

            Traffic crashes are best examined in terms of factors, which, if different, would have altered the probability of occurrence or severity of outcome of the crash; it is generally less illuminating to focus on causes, especially single causes.  Efforts to identify the influence of specific engineering and environmental factors are often difficult because of large confounding influences from road-user factors.  For example, station wagons have about 18% fewer injury claims per insured vehicle than non-station wagon versions of the same model cars; larger model four-door cars tend to have lower injury claim rates than two-door versions of the same models.  Such differences are due more to different use and behavior patterns than to differences in occupant protection in crashes.

            One vehicle characteristic which does exercise a major influence on occupant protection in crashes is vehicle size, which is best represented by vehicle mass, or weight.  When 900 kg and 1800 kg cars crash into each other, the risk of death in the small car is about 13 times what it is in the large car.  A driver in a 900 kg car crashing into another 900 kg car is about twice as likely to be killed as a driver in an 1800 kg car crashing into another 1800 kg car.  In a single-car crash, the risk in a 900 kg car is about 2.4 times that in an 1800 kg car.  Fatalities per registered car show less strong mass effects because smaller cars are involved in fewer severe crashes.  Driver fatalities per registered car decreases steeply with increasing car mass for single-car rollover crashes, and weakly for non-rollover crashes.

            Fatality reductions have been associated with vehicle changes related to various Federal Motor Vehicle Safety Standards (FMVSS).  The largest, a driver fatality reduction of 6.6%, is found for the energy absorbing steering column and associated changes in the maximum rearward displacement in the steering column.  The combined effect of all quantitatively evaluated FMVSS standards is to reduce car-occupant fatalities by 11%.  Taking into account other changes, it seems that present vehicles have occupant fatality risks lower by about 15% to 20% compared to pre-regulation cars.

            The magnitudes of the differences in fatality rates between different types of roads, in some cases over a factor of six, show conclusively that, say, replacing a stretch of rural two-lane roadway by a divided freeway will substantially reduce casualties.  Fatalities per unit distance of travel are lower in the winter months, and fatality risk is less on wet and snow-covered roads than on dry roads.  The effect of inclement weather is more to reduce mobility by deterring travel or reducing speeds than to change safety.  Fatality rates are dramatically higher at night than in the day, an effect largely due to road-user characteristics such as alcohol use.  Multi-disciplinary post-crash investigations in the US and UK identify road-user characteristics as factors in 94% and 95% of crashes, respectively, while only 6% and 5% are associated only with the non-user factors of environment and roadway.  Nearly all attempts to examine engineering and environmental factors encounter larger driver behavior influences.

 

REFERENCES (CHAPTER 4)

 

Adams, J.G.U.  Risk and freedom - the record of road safety regulations. Nottingham, UK: Bottesford Press; 1985a.

Adams, J.G.U.  Smeed's law, seat belts and the emperor's new clothes.  In: Evans, L.; Schwing, R.C., editors. Human behavior and traffic safety. New York, NY: Plenum Press, p. 193-238; 1985b.

Automotive News. 1989 market data book issue. Detroit, MI: Crain Communications; 31 May 31 1989.

Campbell, B.J.; Reinfurt, D.W.  The relationship between driver crash injury and passenger car weight. Chapel Hill, NC: Highway Safety Research Center, University of North Carolina; 1973.

Carsten, O.M.J.; Tight, M.R.; Southwell, M.T.  Urban accidents: why do they happen?  Basingstoke, UK: AA Foundation for Road Safety Research; 1989.

Crandall, R.W.; Graham, J.D.  Automobile safety regulation and off-setting behavior; some new empirical estimates. American Economic Association Proceedings 74:328-330; 1984.

Crandall, R.W.; Graham, J.D.  The effect of fuel economy standards on automobile safety. Journal of Law and Economics 32:97-118; 1989.

Evans, L.  Car mass and likelihood of occupant fatality. SAE paper 820807. Warrendale, PA: Society of Automotive Engineers; 1982.

Evans, L.  Driver fatalities versus car mass using a new exposure approach. Accident Analysis and Prevention 16:19-36; 1984a.

Evans, L.  Accident involvement rate and car size. Accident Analysis and Prevention 16:387-405; 1984b.

Evans, L.  Driver behavior revealed in relations involving car mass. In: Evans, L.; Schwing, R.C., editors. Human behavior and traffic safety. New York, NY: Plenum Press, p. 337-352; 1985a.

Evans, L.  Involvement rate in two-car crashes versus driver age and car mass of each involved car. Accident Analysis and Prevention 17:155-170; 1985b.

Evans, L.  Fatality risk for belted drivers versus car mass. Accident Analysis and Prevention 17:251-271; 1985c.

Evans, L.  Driver age, car mass and accident exposure -- a synthesis of available data. Accident Analysis and Prevention 17:439-448; 1985d.

Evans, L.  Car size and safety: results from analyzing U.S. accident data. Proceedings of the Tenth International Technical Conference on Experimental Safety Vehicles, Oxford, UK; 1-4 July 1985.  National Highway Traffic Safety Administration, report DOT HS 806 916. Washington, DC, p. 548-556; February 1986.

Evans, L.  Estimating fatality reductions from increased safety belt use. Risk Analysis 7:49-57; 1987.

Evans, L.  Passive compared to active approaches to reducing occupant fatalities. Paper No. ESV 89-5B-0-005, presented to the Twelfth International Technical Conference on Experimental Safety Vehicles, Gothenburg, Sweden; 29 May - 1 June 1989. To be published in proceedings of the meeting.

Evans, L.; Frick, M.C.  Potential fatality reductions through eliminating occupant ejection from cars. Accident Analysis and Prevention 21:169-182; 1989a.

Evans, L.; Frick, M.C.  Relative fatality risk in different seating positions versus car model year. Accident Analysis and Prevention  21:581-587; 1989b.

Evans, L.; Wasielewski, P.F.  Serious or fatal driver injury rate versus car mass in head-on crashes between cars of similar mass. Accident Analysis and Prevention 19:119-131; 1987.

Federal Highway Administration.  Fatal and injury accident rates on public roads in the United States. Publication FHWA-SA-90-???, Washington, DC; 1990 (in press).

Graham, J.D.  Technology, behavior and safety an empirical study of occupant-protection regulation. Policy Sciences 17:141-151; 1984.

Graham, J.D.; Garber, S.  Evaluating the effects of automobile safety regulation. Journal of Policy Analysis and Management 3:206-224; 1984.

Grush, E.S.; Marsh, J.C.; South, N.E.  Comparison of high speed crash test results with fatality rates. American Association for Automotive Medicine, 27th Annual Proceedings, San Antonio, TX, p. 189-206; 3-6 October 1983.

Haddon, W. Jr.  A logical framework for categorizing highway safety phenomena and activity. Journal of Trauma 12:193-207; 1972.

Hauer, E.  Selection for treatment as a source of bias in before-and-after studies. Traffic Engineering and Control 20:418-421; 1980.

Hauer, E.  Review of published evidence on the safety effect of conversion from two-way to four-way stop sign control. University of Toronto, Department of Civil Engineering, publication ISBN: 0-7727 7069 7; 1985.

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.

Hauer, E.  A case for science-based safety design and management. In Stammer, R.E., editor. Highway safety: at the crossroads. Washington, DC: American Society of Civil Engineers, p. 241-267; 1988.

Highway Loss Data Institute.  Insurance injury report: passenger cars, vans, pickups, and utility vehicles. Arlington, VA: Research report HLDI I88-1; September 1989.

Horowitz, A.D.  Automobile usage: a factbook on trips and weekly travel. Warren, MI: General Motors Research Laboratories, research publication GMR-5351; 2 April 1986.

Joksch, H.C.  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; January 1976a.

Joksch, H.C.  Critique of Sam Peltzman's study:  The effects of automobile safety regulation. Accident Analysis and Prevention 8:l29-l37; 1976b.

Joksch, H.C.  The effects of automobile safety regulation:  Comments on Peltzman's reply. Accident Analysis and Prevention 8: 2l3-2l4; 1976c.

Joksch, H.C.  Light-weight car safety analysis, phase II, part II: occupant fatality and injury risk in relation to car weight. Performed under contract CEM-8102C1160, Center for the Environment and Man, Hartford, CT; June 1983.

Kahane, C.J.  An evaluation of Federal Motor Vehicle Safety Standards for passenger car steering assemblies: Standard 203 - impact protection for the driver; Standard 204 - rearward column displacement. Washington, DC: National Highway Traffic Safety Administration, report DOT HS-805 705; January 1981.

Kahane, C.J.  Evaluation of current energy-absorbing steering assemblies. SAE paper 820473. Warrendale, PA: Society of Automotive Engineers; 1982a. (Also included in: Occupant crash interaction with the steering system. SAE special publication SP-507, p. 45-49; 1982a).

Kahane, C.J.  An evaluation of side structure improvements in response to Federal Motor Vehicle Safety Standard 212. Washington, DC: National Highway Traffic Safety Administration, report DOT HS 806 314; November 1982b.

Kahane, C.J.  An evaluation of head restraints -- Federal Motor Vehicle Safety Standard 202. Washington, DC: National Highway Traffic Safety Administration, report DOT HS-806 108; February 1982c.

Kahane, C.J.  A preliminary evaluation of two braking improvements for passenger cars -- dual master cylinders and front disc brakes. Washington, DC: National Highway Traffic Safety Administration, report DOT HS-806 359; February 1983.

Kahane, C.J.  The National Highway Traffic Safety Administration's evaluations of Federal Motor Vehicle Safety Standards. SAE paper 840902. Warrendale, PA: Society of Automotive Engineers; 1984.

Kahane, C.J.  An evaluation of windshield glazing and installation methods for passenger cars. Washington, DC: National Highway Traffic Safety Administration, report DOT HS-806 693; February 1985.

Kahane, C.J.  An evaluation of occupant protection in frontal interior impact for unrestrained front seat occupants of cars and light trucks. Washington, DC: National Highway Traffic Safety Administration, report DOT HS 807 203: January 1988.

Kahane, C.J.  An evaluation of door locks and roof crush resistance of passenger cars -- Federal Motor Vehicle Safety Standards 206 and 216. Washington, DC: National Highway Traffic Safety Administration, report DOT HS 807 489; November 1989.

Mackay, M.  Comment on p. 353 of Evans, L; Schwing, R.C., editors. Human behavior and traffic safety. New York, NY: Plenum Press; 1985.

Mackie, P.J; Bonsall, P.W.  Traveller response to road improvements: implications for user benefits. Traffic Engineering and Control 29:411-416; 1989.

McCarthy, R.L.  An examination of the relationship between vehicle mass, wheelbase and safety. Paper presented to the Winter Annual Meeting of the American Society of Mechanical Engineers, San Fransciso, CA; 12-15 December 1989.

McLean, A.J.  Car shape and pedestrian injury. In: National Road Safety Symposium, Canberra, Australia, p. 179-192; March 1972.

Mengert, P.; Salvatore, S.; DiSario, R.; Walter, R.  Statistical estimation of rollover risk. Cambridge, MA: National Highway Traffic Safety Administration, Transportation Systems Center, report DOT-HS-807-446/DOT-TSC-NHTSA-89-3; August 1989. 

Mortimer, R.G.; Fell, J.C.  Older drivers: their night fatal crash involvement and risk. Association for the Advancement of Automotive Medicine, 32nd Annual Proceedings, Seattle, WA, p. 327-206; 12-14 September 1988.

Mueller, B.A.; Rivara, F.P.; Bergman, A.B.  Factors associated with pedestrian-vehicle collision injuries and fatalities. Western Journal of Medicine 146:243-245; 1987.

National Highway Traffic Safety Administration.  Fatal Accident Reporting System 1988. Document DOT HS 807 507. Washington, DC; December 1989.

Negri, D.B.; Riley, R.K.  Two car collision study II. Report DOT-HS-245-2-478-4. Albany, NY: State of New York, Department of Motor Vehicles; June 1974.

Orr, L.D.  The effectiveness of automobile safety regulation: evidence from the FARS data. American Journal of Public Health 74:1384-1389; 1984.

Orr, L.D.  Auto safety regulation variable: a reply to Robertson. American Journal of Public Health 75:789-790; 1985.

Partyka, S.C.  Lives saved by seat belts from 1983 through 1987. National Highway Traffic Safety Administration.  Washington, DC; June 1988.

Partyka, S.C.  Registration-based fatality rates by car size from 1978 through 1987. In: Papers on car size -- safety and trends, National Highway Traffic Safety Administration, report DOT HS 807 444, p. 45-72; June 1989a.

Partyka, S.C.  Recomputation of results in Partyka [1985a] for drivers only rather than all occupants. Private communication 1989b.

Partyka, S.C. Differences in reported car weight between fatality and registration data files. Accident Analysis and Prevention 22:161-166; 1990.

Partyka, S.C.; Boehly, W.A.  Passenger car weight and injury severity in single vehicle nonrollover crashes. Paper ESV 89-2B-0-005, presented to the Twelfth International Technical Conference on Experimental Safety Vehicles, Gothenburg, Sweden; 29 May-1 June 1989. To be published in proceedings of the meeting.

Peltzman, S.  The effects of automobile safety regulation. Journal of Political Economy 83:677-725; 1975.

Peltzman, S.  The effects of automobile safety regulation: reply. Accident Analysis and Prevention 8:139-142; 1976.

Persaud, B.N.  Do traffic signals affect safety?  Some methodological issues. Paper 870610, presented to the 67th Annual Meeting of the Transportation Research Board, Washington, DC; 11-14 January 1988.

Robertson, L.S.  A critical analysis of Peltzman's "The effects of automobile safety regulation."  Journal of Economic Issues ll:586-600; 1977.

Robertson, L.S.  Automobile safety regulations and death reductions in the United States. American Journal of Public Health 7l:8l8-822; 1981.

Robertson, L.S.  Automobile safety regulation: rebuttal and new data. American Journal of Public Health 74:1390-1394; 1984.

Robertson, L.S.  Rejoinder from Robertson (to Orr 1985). American Journal of Public Health 75:790-790; 1985.

Robertson, L.S.  Risk of fatal rollover in utility vehicles relative to static stability. American Journal of Public Health 79:300-303; 1989.

Rumar, K.  The role of perceptual and cognitive filters in observed behavior. In: Evans, L; Schwing, R.C., editors. Human behavior and traffic safety. New York, NY: Plenum Press, p. 151-165; 1985.

Sabey, B.E.; Staughton, G.C.  Interacting roles of road environment, vehicle and road user in accidents. Presented to the Fifth International Conference of the International Association for Accident and Traffic Medicine, London, UK; 1975.

Sabey, B.E.; Taylor, H.  The known risks we run: the highway. In: Schwing, R.C.; Albers, W.A., editors. Societal risk assessment -- how safe is safe enough? New York, NY: Plenum Press, p. 43-63; 1980.

Schwing, R.C.; Evans, L.; Schreck, R.M.  Uncertainties in diesel engine health effects (a comment on two papers). Risk Analysis 3:129-131; 1983.

Schwing, R.C; Kamerud, D.B.  The distribution of risks: vehicle occupant fatalities and time of the week. Risk Analysis 8:127-133; 1988.

Shinar, D.  Comment on p. 166-167 of Evans, L; Schwing, R.C., editors. Human behavior and traffic safety. New York, NY: Plenum Press; 1985.

Stonex, K.A.  Vehicle aspects of the single-car accident problem. Second Regional Conference on Single-car Accidents, Flint, MI; October 1962.

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

Wasielewski, P.F.; Evans, L.  Do drivers of small cars take less risk in everyday driving? Risk Analysis 5:25-32; 1985.

Yankauer, A.  Deregulation and the right to life. American Journal of Public Health 71:797-798; 1981.

Zlatoper, T.J.  Regression analysis of time series data on motor vehicle deaths in the United States. Journal of Transport Economics 18:263-274; 1984.

Table 4-1            Injury claims per insured vehicle, relative to a value of 100 for all cars.  Based on data from the Highway Data Loss Institute [1989].

 ÚÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ¿

 ³       ³       ³                          ³                           ³

 ³ Car   ³ Model ³                          ³ Relative claim frequency  ³

 ³       ³       ³        Make              ÃÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄ´

 ³ size  ³ years ³                          ³ Stn wgn ³ 4-door ³ 2-door ³

 ³       ³       ³                          ³         ³        ³        ³

 ÃÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄ´

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Chevrolet Caprice        ³    58   ³   64   ³   67   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Mercury Grand Marquis    ³    58   ³   64   ³        ³

 ³       ³       ³                          ³         ³        ³        ³

 ³ LARGE ³ 86-88 ³ Ford Crown Victoria      ³    58   ³   63   ³        ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Buick Electra            ³    63   ³   67   ³        ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 87-88 ³ Buick LeSabre            ³         ³   68   ³   92   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³       ³ AVERAGE FOR LARGE        ³    59.2 ³   65.2 ³   79.5 ³

 ³       ³       ³                          ³         ³        ³        ³

 ÃÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄ´

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Oldsmobile Cutlass Ciera ³    70   ³   89   ³   96   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Buick Century            ³    71   ³   90   ³   93   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Chevrolet Celebrity      ³    84   ³   95   ³  100   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Plymouth Reliant         ³    92   ³  110   ³  102   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Chevrolet Cavalier       ³   106   ³  129   ³  136   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Dodge Aries              ³   108   ³  100   ³  104   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 87-88 ³ Toyota Camry             ³    59   ³   84   ³        ³

 ³       ³       ³                          ³         ³        ³        ³

 ³ MID-  ³ 86-88 ³ Volvo 240                ³    60   ³   90   ³        ³

 ³       ³       ³                          ³         ³        ³        ³

 ³ SIZE  ³ 86-88 ³ Mercury Sable            ³    67   ³   85   ³        ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Ford Taurus              ³    67   ³   86   ³        ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Pontiac 6000             ³    71   ³   94   ³        ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Chrysler LeBaron         ³    78   ³   90   ³        ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 87-88 ³ Oldsmobile Calais        ³         ³  105   ³  110   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 87-88 ³ Pontiac Grand Am         ³         ³  114   ³  120   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Buick Skyhawk            ³         ³  120   ³  129   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Oldsmobile Firenza       ³         ³  122   ³  137   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³       ³ AVERAGE FOR MIDSIZE      ³   77.7  ³  100.2 ³  112.7 ³

 ³       ³       ³                          ³         ³        ³        ³

 ÃÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄ´

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Subaru DL/GL             ³  100    ³  139   ³  133   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 87-88 ³ Nissan Sentra            ³  102    ³  153   ³  151   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Saab 900                 ³         ³   77   ³   83   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Volkswagen Golf          ³         ³  100   ³   96   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Acura Integra            ³         ³  102   ³  102   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³  88   ³ Honda Civic              ³         ³  111   ³   97   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 87-88 ³ Dodge Shadow             ³         ³  112   ³  126   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³ SMALL ³ 87-88 ³ Plymouth Sundance        ³         ³  112   ³  121   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Mazda 323                ³         ³  119   ³  106   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 87-88 ³ Volkswagen Fox           ³         ³  124   ³  122   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Dodge Colt               ³         ³  143   ³  130   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Plymouth Colt            ³         ³  152   ³  129   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Isuzu I-Mark             ³         ³  164   ³  182   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Chevrolet Sprint         ³         ³  164   ³  178   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³ 86-88 ³ Chevrolet Spectrum       ³         ³  176   ³  166   ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³       ³ AVERAGE FOR SMALL        ³   101.0 ³  129.9 ³  128.1 ³

 ³       ³       ³                          ³         ³        ³        ³

 ÃÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄ´

 ³       ³       ³                          ³         ³        ³        ³

 ³       ³       ³   GRAND AVERAGE          ³    76.2 ³  107.7 ³  118.8 ³

 ³       ³       ³                          ³         ³        ³        ³

 ÀÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÙ

Table 4-2. Summary of car mass effects, expressed as the risk associated with a 900 kg car compared to that associated with a 1800 kg car.

 

 

 ÚÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄ¿

 ³           ³                         ³                          ³           ³

 ³ Number of ³                         ³       Description        ³ 900 kg to ³

 ³           ³      Quantity           ³                          ³           ³

 ³ cars      ³                         ³           of             ³ 1800 kg   ³

 ³           ³      measured           ³                          ³           ³

 ³ involved  ³                         ³          crash           ³ ratio     ³

 ³           ³                         ³                          ³           ³

 ÃÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄ´

 ³           ³                         ³                          ³           ³

 ³           ³                         ³ Into each other:         ³           ³

 ³           ³ Driver fatalities       ³                          ³           ³

 ³           ³                         ³       - All directions   ³   13      ³

 ³           ³ per crash               ³                          ³           ³

 ³           ³                         ³       - Head-on          ³   14      ³

 ³           ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³ Driver serious injuries ³ Into car of similar mass ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³ (including fatalities)  ³       - All directions   ³   2.2     ³

 ³           ³                         ³                          ³           ³

 ³           ³ per crash               ³       - Head-on          ³   2.0     ³

 ³           ³                         ³                          ³           ³

 ³ TWO-      ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³ Driver fatalities       ³ Into "average" car in    ³           ³

 ³ CAR       ³                         ³                          ³   4       ³

 ³           ³ per crash               ³   1978 car mix           ³           ³

 ³           ³                         ³                          ³           ³

 ³ CRASHES   ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³ Driver fatalities       ³ All driver fatalities    ³           ³

 ³           ³                         ³                          ³   1.9     ³

 ³           ³ per registered car      ³   in two-car crashes     ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³ Police-reported crashes ³                          ³           ³

 ³           ³                         ³ Into car of similar mass ³   0.3     ³

 ³           ³ per registered car      ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ÃÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄ´

 ³           ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³ Driver fatalities       ³ Unbelted drivers         ³   2.4     ³

 ³ SINGLE-   ³                         ³                          ³           ³

 ³           ³    per crash            ³ Belted drivers           ³   2.3     ³

 ³           ³                         ³                          ³           ³

 ³ CAR       ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³                         ³ All single-car crashes   ³   1.5     ³

 ³ CRASHES   ³ Driver fatalities       ³                          ³           ³

 ³           ³                         ³ Rollover only            ³   1.8     ³

 ³           ³ per registered car      ³                          ³           ³

 ³           ³                         ³ Non-rollover only        ³   1.15    ³

 ³           ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³ Police-reported crashes ³ Assumed to be same as    ³           ³

 ³           ³                         ³                          ³  0.72     ³

 ³           ³ per registered car      ³ for all crashes          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ÃÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄ´

 ³           ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³ Driver fatalities       ³                          ³           ³

 ³           ³                         ³ All crashes              ³  2.8      ³

 ³           ³ per crash               ³                          ³           ³

 ³  ALL      ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³ Driver fatalities       ³                          ³           ³

 ³  CAR      ³                         ³ All crashes              ³  1.7      ³

 ³           ³ per registered car      ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³  CRASHES  ³                         ³                          ³           ³

 ³           ³                         ³                          ³           ³

 ³           ³ Police-reported crashes ³                          ³           ³

 ³           ³                         ³ All crashes              ³  0.72     ³

 ³           ³ per registered car      ³                         

Table 4-3.  Fatality reductions estimated by Kahane for various Federal Motor Vehicle Safety Standards (FMVSS).

 

                                                                                                        Fatalities prevented

Description                         FMVSS             Occupants               Protected              Average over

                                                                    protected                 occupants             all occupants

                                                                                                     

Energy absorbing column         203                

Column displacement              204                 Driver                        6.6%                          4.4%

Instrument panels                    201                 Front Passengers       7                                 1.7

Side structure                          214                 All                             1.7                              1.7

Door locks                              206                 All                             1.5                              1.5

Roof crush resistance              216                 All                             0.43                            0.43

Windshield glazing                   212                 All                             0.39                            0.39

Head restraints                        202                 Driver and right-

                                                                     front passenger          0.36                            0.33

Braking improvements             105                 All                             0.9                              0.9

 

Table 4-3.  Fatality reductions estimated by Kahane for various Federal Motor Vehicle Safety Standards (FMVSS).

 

                                               Fatalities prevented

Description             FMVSS  Occupants    Protected   Average over

                               protected    occupants   all occupants

 

Energy absorbing column  203

Column displacement      204     Driver         6.6%         4.4%

Instrument panels        201     Front Psngrs   7            1.7

Side structure           214     All            1.7          1.7

Door locks               206     All            1.5          1.5

Roof crush resistance    216     All            0.43         0.43

Windshield glazing       212     All            0.39         0.39

Head restraints          202     Driver & rt-

                                 front psngr    0.36         0.33

Braking improvements     105     All            0.9          0.9

 

 

ÄÄÄÄÄÄÄÄÄÄÄÄ¿

³                         ³       ³             ³ Fatalities prevented, %   ³

³                         ³       ³ Occupants   ÃÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ´

³     Description         ³ FMVSS ³             ³ Protected ³ Average over  ³

³                         ³       ³ protected   ³           ³               ³

³                         ³       ³             ³ Occupant  ³ all occupants ³

³                         ³       ³             ³           ³               ³

ÃÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ´

³                         ³       ³             ³           ³               ³

³ Energy absorbing column ³  203  ³             ³           ³               ³

³                         ³       ³ Driver      ³   6.6%    ³     4.4%      ³

³ Column displacement     ³  204  ³             ³           ³               ³

³                         ³       ³             ³           ³               ³

³                         ³       ³             ³           ³               ³

³                         ³       ³             ³           ³               ³

³ Instrument panels       ³  201  ³ Front pass  ³   7%      ³     1.7%      ³

³                         ³       ³             ³           ³               ³

³                         ³       ³             ³           ³               ³

³                         ³       ³             ³           ³               ³

³ Side structure          ³  214  ³ All         ³   1.7%    ³     1.7%      ³

³                         ³       ³             ³           ³               ³

³                         ³       ³             ³           ³               ³

³                         ³       ³             ³           ³               ³

³ Door locks              ³  206  ³ All         ³   1.5%    ³     1.5%      ³

³                         ³       ³             ³           ³               ³

³                         ³       ³             ³           ³               ³

³                         ³       ³             ³           ³               ³

³ Roof crush resistance   ³  216  ³ All         ³   0.43%   ³     0.43%     ³

³                         ³       ³             ³           ³               ³

³                         ³       ³             ³           ³               ³

³                         ³       ³             ³           ³               ³

³ Windshield glazing      ³  212  ³ All         ³   0.39%   ³     0.39%     ³

³                         ³       ³             ³           ³               ³

³                         ³       ³             ³           ³               ³

³                         ³       ³ Driver and  ³           ³               ³

³ Head restraints         ³  202  ³             ³   0.36%   ³     0.33%     ³

³                         ³       ³ right front ³           ³               ³

³                         ³       ³             ³           ³               ³

³                         ³       ³             ³           ³               ³

³ Braking improvements    ³  105  ³ All         ³   0.9%    ³     0.9%      ³

³                         ³       ³             ³           ³               ³

ÀÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÙ

 

Table 4-4  Traffic fatalities and fatalities per billion km of travel on various types of roads in 1988.  Data from Table 1 of Federal Highway Administration [1990].

 

ÚÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄ¿

³                              ³      Fatalities       ³  Fats per billion km  ³

³      Highway system          ÃÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄ´

³                              ³ Rural ³ Urban ³ Total ³ Rural ³ Urban ³ Total ³

ÃÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄ´

³                              ³       ³       ³       ³       ³       ³       ³

³    NON-INTERSTATE:-          ³       ³       ³       ³       ³       ³       ³

³                              ³       ³       ³       ³       ³       ³       ³

³ Fed. aid primary (arterial)  ³10 748 ³ 4 252 ³15 000 ³ 21.7  ³  9.7  ³ 16.1  ³

³                              ³       ³       ³       ³       ³       ³       ³

³ Other Federal aid urban      ³   -   ³ 8 726 ³ 8 726 ³  -    ³ 12.2  ³ 12.2  ³

³                              ³       ³       ³       ³       ³       ³       ³

³ Fed. aid secondary collector ³ 6 771 ³   -   ³ 6 771 ³ 24.0  ³  -    ³ 24.0  ³

³                              ³       ³       ³       ³       ³       ³       ³

³ Non-Fed. aid arterial        ³   295 ³   299 ³   594 ³ 50.9  ³  7.2  ³ 12.6  ³

³                              ³       ³       ³       ³       ³       ³       ³

³ Non-Fed. aid collector       ³ 2 214 ³   313 ³ 2 527 ³ 25.0  ³  7.9  ³ 19.7  ³

³                              ³       ³       ³       ³       ³       ³       ³

³ Non-Fed aid local            ³ 4 839 ³ 3 509 ³ 8 348 ³ 32.1  ³ 12.0  ³ 18.8  ³

³                              ³       ³       ³       ³       ³       ³       ³

³   NON-INTERSTATE TOTAL       ³24 867 ³17 099 ³41 966 ³ 24.3  ³ 11.2  ³ 16.4  ³

³                              ³       ³       ³       ³       ³       ³       ³

³   INTERSTATE                 ³ 2 826 ³ 2 301 ³ 5 127 ³  9.7  ³  5.5  ³  7.2  ³

³                              ³       ³       ³       ³       ³       ³       ³

³   ALL ROADS TOTAL            ³27 693 ³19 400 ³47 093 ³ 21.0  ³ 10.0  ³ 14.4  ³

³                              ³       ³       ³       ³       ³       ³       ³

ÀÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÙ

 

Table 4-5  Fatal and personal injury crashes in Ontario, Canada in 1974-1980. From Adams [1985b].

 

     ÚÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄÄÄÄ¿

     ³ Road surface ³ Number of crashes ³ Fatal/injury ³

     ³              ÃÄÄÄÄÄÄÄÂÄÄÄÄÄÄÄÄÄÄÄ´              ³

     ³ condition    ³ Fatal ³  Injury   ³  (percent)   ³

     ÃÄÄÄÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÅÄÄÄÄÄÄÄÄÄÄÄÄÄÄ´

     ³              ³       ³           ³              ³

     ³ Dry          ³  6494 ³ 274 873   ³    2.36      ³

     ³              ³       ³           ³              ³

     ³ Wet          ³  1878 ³ 113 051   ³    1.66      ³

     ³              ³       ³           ³              ³

     ³ Loose snow   ³   214 ³  16 448   ³    1.30      ³

     ³              ³       ³           ³              ³

     ³ Slush        ³   179 ³  11 000   ³    1.63      ³

     ³              ³       ³           ³              ³

     ³ Packed snow  ³   226 ³  12 413   ³    1.82      ³

     ³              ³       ³           ³              ³

     ³ Ice          ³   289 ³  19 446   ³    1.49      ³

     ³              ³       ³           ³              ³

     ÀÄÄÄÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÄÁÄÄÄÄÄÄÄÄÄÄÄÄÄÄÙ