Better to see Chapter 3 Overview of traffic fatalities


Chapter 15 The dramatic failure of US safety policy of Traffic Safety  (2004)

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


Words only (no formatting, figures, tables, or photographs) from 1991 book



Paperback copy of complete unchanged book available from , list price $29.95


Chapter 3.  AN OVERVIEW OF UNITED STATES TRAFFIC FATALITIES (From 1991 book Traffic Safety and the Driver)


                 The 1988 FARS data code 42 119 fatal crashes in the US in which 47 093 people were killed, for an average of 1.118 fatalities per fatal crash; 70% of those killed were male.  Fig. 3-1 shows a breakdown of these fatalities by different categories of road users.  Percents (of the 47 093 fatalities that occurred in 1988) are shown rather than the numbers because the pattern is fairly stable from year to year.  A year to year decrease (increase) in overall fatalities tends to increase (or decrease) the number of fatalities in each cell in Fig. 3-1 in an approximately proportional way so that the percents shown remain fairly constant from year to year [see also Hedlund 1985; National Highway Traffic Safety Administration 1989 (and earlier years)].  For FARS data from 1975 through 1988 the annual number of fatalities varied from a low of 42 584 in 1983 to a high of 51 093 in 1979.  The pattern in Fig. 3-1 would not persist over long time trends -- in the 1930's about 35% of all traffic fatalities were pedestrians compared to 14.7% in 1988.


Fig. 3-1 about here


                 Another reason why we generally display percents is that in nearly all FARS categories there are cases with some unknown variables (for example, the occupant's age, or the make of the vehicle, is unknown).  When unknown cases are few, distributing them into other categories in proportion to the numbers already in those categories is appropriate.  Only in cases of variables for which there are substantial numbers in the unknown category (such as safety
belt use or blood alcohol concentration) will we need to address this problem more specifically.




                 While every traffic crash involves at least one driver, it is important to note, from Fig. 3-1, that 42% of traffic fatalities are not drivers, with the value being higher for other countries because of a larger proportion of non-occupant fatalities, and higher occupancy rates in vehicles.  The fraction of all traffic fatalities that are drivers depends strongly on sex and age, as shown in Fig. 3-2.  Only a very small fraction of those killed at ages below that required to obtain a driving license are drivers.  The fraction of male fatalities that are drivers exceeds 70% between ages 20 and 40.  The fraction of male traffic fatalities that are drivers exceeds the fraction of female fatalities that are drivers at every age, based on the 1983-1985 FARS data used to produce Fig. 3-2.


Fig. 3-2 about here





                 Fig. 3-1 shows that 56% of all fatalities are occupants of cars.  Of those killed in motorized vehicles, 67% are killed in cars; 64% of all drivers killed are drivers of cars.  Because of the dominant role of car occupants in overall occupant fatalities we examine car occupant fatalities in some detail below.  Parallel information for all occupants is less informative because of the wide range of occupants, from those in motorcycles to those in busses.


Vehicles and objects struck


                 The 26 069 car occupant fatalities in 1988 are shown in Fig. 3-3 according to the other vehicles involved.  The most important factor to note here is that the most common type of crash leading to a car-occupant fatality is a single-car crash; this is at variance with a popular image of a fatal crash most typically involving two cars.  This misconception probably arises because the most common type of crash, without regard to injury level, is indeed a two-car crash.  If we consider only minor property damage crashes, two-car crashes dominate, with rear-end "fender bender" incidents being most common.  Crashes involving cars travelling in congested traffic in the same direction, although numerous, are unlikely to prove fatal.  As we focus on crashes of increasing severity, the role of two-car crashes becomes less dominant, and that of single-car crashes more dominant.


Fig. 3-3 about here


                 Single-car crashes are either non-collisions or crashes into objects (a breakdown is shown in Fig. 3-3).  More than 90% of the non-collision crashes are rollovers; the remainder is the sum of occupants killed due to fire, explosion, immersion in water, falling from the vehicle, injured in non-crashing vehicle, and other types of non-collisions.  The most common struck object leading to car occupant death is a tree, reflecting the large number of trees close to roadways, especially in rural two-lane roads.  The "all other objects" includes highway sign post, boulder, wall, fire hydrant, animal, impact attenuator and many other objects making small contributions to the total.


Fatalities according to seating position


                 Car occupants killed are shown in Fig. 3-4 distributed according to the seat in which they were travelling.  This figure, in common with others above, uses 1988 FARS data. It therefore reflects the mix of cars by type and model year on the roads in 1988.  Only occupants coded as occupying one of the six seats are included in the analysis -- about 0.3% are coded as being in unknown seats.  The distributions in Fig. 3-4 are important in addressing questions of occupant protection.  A device that reduces driver fatality risk by 1% would prevent a larger number of fatalities than one which reduces center-rear passenger fatality risk by 60%.


Fig. 3-4 about here



Relative fatality risk in different seats


                 The data in Fig. 3-4 do not enable us to address the relative risk of sitting in different seats, because a greater number of fatalities in one seat occurs mainly because of a higher occupancy rate in that seat.  However, even if we could correct for different occupancy rates, many other factors that affect fatality risk would still make it difficult to isolate the influence of the seating position, as such.  Cars with only one occupant are involved in crashes of different types and severity than those with more than one occupant.  Occupants in different seats have different distributions by sex and age, factors that have a substantial influence on fatality risk (Chapter 2).  We thus encounter an example of the problem of exposure, referred to in
Chapter 1, which arises when attempts are made to make inferences from field data.

                 This problem is addressed [Evans and Frick 1988a] by selecting from 1975-1985 FARS data cars in which there were drivers and passengers in specified seats.  In order to avoid confounding sex and age effects, only cases in which the driver and passenger are of the same sex, and have ages the same to within three years, are included.  Also, occupants coded as using any restraint system, or who were less than 16 years old, are excluded from the analysis.  From data restricted in this way, the ratio R


            Number of passenger fatalities in specified seat       

      R  =  ────────────────────────────────────────────────                          Eqn 3-1

                    Number of driver fatalities


is computed.  Because all the occupants in Eqn 3-1 are killed in crashes in which the other occupant is also present in the car, and both occupants are of the same sex and similar age, R provides a remarkably assumption-free estimate of the difference in risk due to differences in the physical environment of the different seating positions, essentially free from confounding effects due to occupant characteristics being correlated with the use of different seats. 

                 Raw data and computed values of R are shown in Fig. 3-5; because all values are relative to the driver, there is no computed relative risk for the driver, for whom, by definition, R = 1.  For cars containing a driver and a right-front passenger, there are 15 880 right-front passenger fatalities compared to 15 793 driver fatalities, for a right-front passenger relative fatality risk of R = 1.006 + 0.011.  The error is computed assuming that the fatalities arise from a Poisson process.  Thus the finding is that, to high precision, there is no difference in fatality risk to drivers and right-front passengers.  The center-front seat R = 0.78 + 0.04 indicates that this
position is associated with a (22 + 4)% lower fatality risk than the outboard (driver or right-front passenger) front positions.  The outboard-rear seats have a composite R = 0.739 + 0.015.  That is, for unrestrained occupants in outboard seating positions, rear seats are associated with a fatality risk (26.1 + 1.5)% lower than for front seats.


Fig. 3-5 about here



Seating position and direction of impact


                 The FARS data contain a principal impact point variable, defined as the impact that is judged to have produced the greatest personal injury or property damage for a particular vehicle.  The impact refers to the location on the vehicle sustaining damage, so that principal impact point at 12 o'clock means that the damage is in the center-front of the vehicle.  The actual direction of impact cannot be inferred from the damage alone -- this would require a detailed post-crash investigation. The center front could be damaged by, say, an oblique impact into a tree.  However, principal impact point 12 o'clock may be approximately interpreted as indicating, on average at least, head-on impacts.

                 Fig. 3-6 shows relative risk for the five passenger seating positions versus principal impact point displayed in the same bird's eye view of the vehicle travelling up the page used in Figs 3-4 and 3-5.  All values are relative to a value one for drivers.  Focusing on the right-front passenger data, which is displayed in the right-top circle in Fig. 3-6, we note that when the impact is from the right, the right-front passenger is 2.74 times as likely to die as is the driver (as before, both occupants are of the same sex,
and ages not different by more than three years).  When the impact is from the left, then the right-front passenger is 0.38 times as likely to die as the driver; this could be expressed equivalently by saying that the driver is 1/0.38 = 2.63 times as likely to die as is the right-front passenger.  The essential symmetry (reflected in the closeness of the ratios 2.74 and 2.63) is to be expected on physical grounds, and increases confidence in the estimates.  For principal impact point 12 o'clock the value of R for right-front passengers is R = 0.988 + 0.019.  Thus the similarity of fatality risk to drivers and right-front passengers applies also to the frontal case.  Drivers and right-front passengers are equally likely to be killed from rear impacts (R = 1.00 + 0.10).  In frontal crashes, the advantage of rear compared to corresponding front seating positions is larger than for all crash directions combined.  The general pattern in Fig. 3-6 shows that occupants near the point of impact are at much greater fatality risk than those far from the point of impact.  Although rear occupants are at much greater risk than front occupants in rear impact crashes, such crashes account for less than 5% of all fatalities.  The overall 26% lower fatality risk in rear than in front seats reflects the predominance of frontal fatal crashes.

                 A corresponding phenomenon occurs for motorcyclists, where it is found [Evans and Frick 1988b] that fatality risk in the driver seat is (26 +  2)% greater than that in the passenger seat; for frontal crashes the difference is (40% + 6)%, again demonstrating the greater risk associated with being nearer the impact.  For non-frontal crashes there is no difference between driver and passenger risk (R = 1.01 + 0.04).  In the motorcycle case, the front occupant probably helped cushion the impact for the passenger.

                 An additional finding in Evans and Frick [1988a] study of car occupant risk is that there are 38% more impacts of high severity from the right than from the left, a result possibly reflecting asymmetries resulting from driving
on the right; it would be interesting to see if countries which drive on the left produce an opposite effect.


Fig. 3-6 about here



Passenger compared to driver risk versus model year


                 The finding that the fatality risk to right-front passengers is the same as that to drivers, to within about 1%, is surprising in view of the substantial differences between the physical environment of drivers and right-front passengers, especially the presence of the steering wheel.  To explore this further, the ratio of right-front passenger risk to driver risk is examined versus model year [Evans and Frick 1989a] with the results shown in Fig. 3-7.  Error limits are large for the oldest cars because few remained by the first year, 1975, in which FARS data became available.  Errors are also large for the latest model-year cars shown because the majority of crashes in which these cars are likely to be involved had not yet occurred by 1986, the last FARS year included in the study.  The results suggest that fatality risk to right-front passengers is somewhat higher than that to drivers for early 1960's model-year cars, but reversed for later model years.  This suggests that the various vehicle modifications aimed at reducing occupant risk have generated somewhat larger fatality risk reductions for right-front passengers than for drivers [Kahane 1988].


Fig. 3-7 about here


                 In the past it was thought that the right-front seat had a substantially higher fatality risk than the driver seat.  I suspect that this arose more because of the 25% increased fatality risk associated with being female (Chapter 2); in the period when the term suicide seat was used, a higher proportion of severe crashes would have involved a male driver and a female passenger.  Campbell [1987] finds right-front passengers about 15% more likely to be injured than drivers in data for 1973-1981; I suspect that differential sex effects might have contributed to this.


Fraction of deaths that are ejections


                 Fig. 3-8 shows information on ejection for all six car seating positions, based on 1975-1986 FARS data [Evans and Frick 1989b].  Occupants coded in FARS as totally ejected or partially ejected are included.  According to Clark and Sursi [1985], half of ejection fatalities are through glazing areas.  The ratio, R, of fatality risk if ejected to fatality risk if not ejected is estimated by applying the double pair comparison method described in Chapter 2.  The values obtained are in agreement with similar quantities estimated by others [Huelke and Compton 1983; Sikora 1986; Green et al. 1987].  Basically, an occupant ejected in a crash is three to four times as likely to be killed as an occupant not ejected in a similar crash.


Fig. 3-8 about here


                 Close to a quarter of fatalities for each seating position are ejected occupants.  If these occupants had not been ejected, then a fraction 1/R of them would still have been killed.  Thus, the percent of fatalities preventable by eliminating ejection, F, is readily computed as f(1 - 1/R),
where f is the percent of fatalities that are ejectees.  For all seating positions, eliminating ejection would prevent about 18% of fatalities.  These results are relevant to the effectiveness of safety belts, which essentially prevent ejection.

                 The values of the quantities displayed in Fig. 3-8 vary with occupant age as shown in Fig. 3-9.  Eliminating ejection probably provides a lesser reduction in fatality risk to older occupants because of their higher fatality risk when not ejected.  If ejection does occur, fatality risk is so high that the greater survivability of the young has a discounted value (it would be of no value in completely unsurvivable crashes), whereas when occupants remain in the vehicle, the greater survivability of the younger occupant translates into a real reduction in fatality risk.  The lower rate of rollover crashes of older drivers discussed in the next section also contributes to declines in f and F with age.


Fig. 3-9 about here



Impact direction and age


                 Fig. 3-10 shows the distribution of car-driver fatalities into principal impact categories [Evans 1989].  The definitions used are, for front, 12 o'clock; for near front, 10, 11, 1 and 2 o'clock; for side, 9 and 3; and for rear, 4 to 8.  The non-collisions in the top/non-collision category are essentially all rollovers, although other rare events such as drownings and fire are included.  The data in Fig. 3-10 are such that for each driver age, all five components sum to 100%.  The most noticeable characteristic is that given a fatality, it is increasingly less likely to be a rollover crash as
drivers age.  Also note that the fraction of deaths that are side impacts increases with driver age, supporting other studies [Viano et al. 1990] that side impact plays an increasingly important relative role in older-driver safety.


Fig. 3-10 about here




                 Because there are about 45 000 fatalities per year in the US one occasionally hears comments to the effect that each day there are 120 deaths on our roads, or each hour there are five, or, in extreme form, every ten minutes someone is killed in traffic somewhere in the US.  Fatalities do not occur in the regular pattern implicit in such comments, as the distribution of the number fatalities per day in Fig. 3-11 shows.  This is based on the 653 804 fatalities for which the day of death is coded (there are another 124, or 0.02%, with uncoded day), occurring in the 5114 days in FARS 1975 - 1988.  The average is 127.8 fatalities per day.


Fig. 3-11 about here


                 The least number of fatalities in a day, 47, occurred on Monday 10 January 1977.  The largest number, 319, occurred on Friday 21 December 1979.  If one does not restrict the comparison to legal days (ending at midnight), then the smallest number of fatalities recorded in a 24 hour period is 39, from 7:01 a.m. on Monday 10 January 1977 to 7:00 a.m. on the following Tuesday.  The largest number recorded in a 24 hour period is 400, from 1:46 p.m. on Friday 21 December 1979 to 1:45 p.m. the next day.  The concurrence of weekend and
Christmas celebrations, and the consequent alcohol consumption, undoubtedly contributed to this highest value, more than ten times the smallest value.  Rank ordering fatalities per (legal) day shows that the 31 days with the largest numbers of fatalities (ranging from 249 to 319) are all either Fridays or Saturdays (28 of the 31 are Saturdays).  None of the 77 lowest fatality days (from 47 to 65 fatalities) contained a Friday or a Saturday.  The number of fatalities on Mondays, Tuesdays and Wednesdays average about 54% of the Saturday total, and on Thursdays 60% of the Saturday total.  The other two week-end days, Friday and Sunday, have about 80% as many fatalities as Saturday, the Sunday total being concentrated on the first few hours of the day (that is late on "Saturday"). 

                 The longest fatality-free period is 6 hours 20 minutes; there were no fatalities from 4:09 a.m. to 10:29 a.m. on Sunday 23 January 1977 (fatalities occurred at 4:08 a.m. and 10:30 a.m.).  There have been five fatality free periods longer than five hours, 31 longer than four hours, 221 longer than three hours, 1364 longer than two hours.  It is not possible to determine the highest rate of occurrence of fatal crashes because the time of occurrence of fatal crashes is rarely known to high precision, and in many cases known only approximately.  Hence crashes tend to be coded in accord with the customary practice of estimating uncertain times.  Far more are coded at whole hours than at half hours, followed by 15 minutes before and after the hour, and then in five minute gradations.  The example above of a crash coded at other than a multiple of five minutes past the hour is rare.  The most crashes coded as occurring at the same time is nine, a value that occurs on five occasions, but each is exactly on the hour.  In Chapter 4 the variation of fatalities by month is discussed in the context of weather effects.the 




                 Fig. 3-12 shows the number of traffic fatalities per year in the US since 1900.  The values for 1975 and later are from FARS.  The earlier values are unpublished estimates by the National Highway Traffic Safety Administration.  These values differ from those published by the National Safety Council [1989] in that they have been corrected to reflect the same inclusion criteria as FARS, namely, death resulting from the crash within 30 days of the crash;   The National Safety Council data use a one year inclusion criterion, so that their values are typically four percent higher than those plotted.


Fig. 3-12 about here


                 Also shown in Fig. 3-12 is the number of pedestrian fatalities, which has been declining since peaking in the mid 1930's.  The fraction of all fatalities that are pedestrian fatalities has been declining, a characteristic of developing motorization observed in many countries [Hutchinson 1987].




Campbell, B.J.  Safety belt injury reduction related to crash severity and front seated position. Journal of Trauma 27:733-739; 1987.

Clark C.C.; Sursi P.  Car crash tests of ejection reduction by glass-plastic side glazing. SAE paper 851203. Warrendale, PA: Society of Automotive Engineers; 1975.

Evans, L.  Airbag effectiveness in preventing fatalities predicted according to type of crash, driver age, and blood alcohol concentration. Association for the Advancement of Automotive Medicine, 33rd Annual Proceedings, Baltimore MD, p. 307-322; 2-4 October 1989.

Evans, L.; Frick, M.C.  Seating position in cars and fatality risk. American Journal of Public Health 78:1456-1458; 1988a.

Evans, L.; Frick, M.C.  Helmet effectiveness in preventing motorcycle driver and passenger fatalities. Accident Analysis and Prevention 20:447-458; 1988b.

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

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

Green P.D.; Robertson N.K.B.; Bradford M.A.; Bodiwala G.G.  Car occupant ejection in 919 sampled accidents in the U.K. -- 1983-86. SAE paper 870323. Warrendale, PA: Society of Automotive Engineers; 1987. (Also included in Restraint technologies: front seat occupant protection. SAE special publication SP-690: p. 91-104; 1987).

Hedlund, J.H.  Recent U.S. traffic fatality trends. In: Evans, L.; Schwing, R.C., editors. Human behavior and traffic safety. New York, NY: Plenum Press, p. 7-19; 1985.

Huelke D.F.; Compton C.P.  Injury frequency and severity in rollover car crashes as related to occupant ejection, contacts and roof damage. Accident Analysis and Prevention 15:395-401; 1983.

Hutchinson, T.P.  Road accident statistics. Adelaide, Australia: Rumsby Scientific Publishing; 1987.

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.

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

National Safety Council.  Accident facts. Chicago, IL; 1989 edition (issued annually).

Sikora, J.J.  Relative risk of death for ejected occupants in fatal traffic accidents. Washington, DC: National Highway Traffic Safety Administration, report DOT HS 807 096;  November 1986.

Viano, D.C.; Culver, C.C.; Evans, L.; Frick, M.C; Scott, R.  Involvement of older drivers in multivehicle side-impact crashes. Accident Analysis and Prevention 22:177-199; 1990.