U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-RD-03-042
Date: November 2003
A Review of Pedestrian Safety Research in the United States and Abroad
Part 2. Crashes Involving Pedestrians
The sheer number of pedestrians killed or seriously injured in U.S. traffic collisions each year is an important consideration in the Nation's highway transportation system. The NHTSA estimated that 5,300 pedestrians were killed in the United States during 1997 (Traffic Safety Facts, 1997). NHTSA also reported 77,000 nonfatal injuries. Casualties of this magnitude take on even greater meaning in view of the national policy to encourage increased walking as a matter of both transportation and health.
Limitations of National Pedestrian Crash Data
Pedestrian crash statistics must be considered with two caveats: First, the numbers presented in this report are estimates, aggregated from information compiled in 50 States plus the District of Columbia, or taken from studies that sample pedestrian collisions. Predictably, there is some uncertainty in the reporting processes and expectable variations due to sampling differences. Second, definitions vary among agencies. Thus, fatality estimates from the National Safety Council (NSC) differ somewhat from those of the NHTSA due to their definitions of death. NSC counts a traffic fatality as any crash death that ensues within 1 year after the crash, whereas NHTSA only counts those deaths that occur within 30 days. Other totals would result if, for example, traffic death counts were confined to victims who died at the crash scene.
Another caution pertains to interpreting the following data tables, which may themselves be overly simplistic. A collision may be associated with several factors, therefore a table that displays only two or three factors leaves much unsaid. For example, a table depicting pedestrian crashes by time of occurrence will show that a certain proportion occurs at night. However, not all of the observed trends shown in such a table would be caused by reduced nighttime visibility. Although reduced visibility at night plays an important role in the likelihood of a crash, the nighttime figures are also influenced by the number of people who are exposed to risk at night as opposed to in the daytime, the ages of the pedestrians, and the role of alcohol. Thus,any table of crash data reflects the influence of variables not shown, as well as the influence of the variables that are depicted.
Moreover, many studies do not consider the number of people actually exposed to risk.
Reports reviewed here span the period during which standardized terminology was evolving. Thus, various authors use the term "accident", "collision", or "crash" to describe the phenomena under examination here. While the terms are sometimes used interchangeably, "crash" has become the term preferred by safety organizations.
The Number of Pedestrian Fatalities
The NSC annually publishes an estimate of U.S. pedestrian deaths for the most recent year available, as well as for years as far back as 1927 (Accident Facts, 1995, 1996, 1997 editions). The estimated number of pedestrian fatalities for 1927 was 10,820, as shown in table 4. The pedestrian death count irregularly grew until 1937, when 15,500 pedestrian deaths were reported, the largest number in U.S. history. Thereafter, the number declined. The number of pedestrian fatalities exceeded 10,000 during 1969, 1972, and 1973, and since then decreased to 6,100 in 1996. It should be noted that the NSC estimating procedure was changed in 1987 and the annual number of deaths it reported became substantially lower beginning that year.
It is also of interest to consider pedestrian fatalities as a percentage of total traffic fatalities. During 1927, pedestrian deaths accounted for 41.9 percent of total traffic deaths. That percentage has declined and was 14.1 percent of the total in 1996, a substantial decline from 1927 (figure 5). Increased travel by car is one factor in that change. Evans (1991), among others, has reported that the higher the degree of motorization in a country, the number of motor vehicles per 100,000 population, the lower the proportion of pedestrian deaths to total traffic deaths. The proportion of pedestrian deaths declined as the United States progressively became more motorized. This trend also is seen in less motorized countries (Choueiri et al., 1993).
This influence of car travel was highlighted by the U.S. experience during World War II when there was an interruption in the steady decline in pedestrian fatalities as a percentage of total traffic fatalities. The pedestrian fatalities were 33.9 percent in 1941. During the war years (1942-1945) the total abruptly jumped to values between 38 and 41 percent. In 1947, the figure reversed again to a new low of 32 percent, and after 1949 did not again reach 30 percent (See table 4). The increase during World War II probably reflected the fact that motor vehicle mileage fell during those years because civilian vehicle production gave way to military needs, and the existing civilian motor vehicle fleet amassed drastically fewer miles because of gasoline and tire (that is, rubber) rationing. It is also likely that the characteristics of highway users changed during that period, what with millions of young men away in the military, plus unprecedented numbers of women in the work force. It should be noted that Choueiri et al. (1993) show a greater pedestrian death decline in Europe during recent years than in the United States.
Table 5 shows a comparison of States (and Puerto Rico), using NHTSA data (Traffic Safety Facts, 1995) from highest rank to lowest according to pedestrian deaths per 100,000 population. The table is reproduced here to suggest some of the complexities that underlie pedestrian crash and death figures. The 10 States with the highest pedestrian death rate per 100,000 population vary significantly: two are urban and densely populated (DC and Puerto Rico), 3 are sparsely populated desert states, and 3 are among the 11 most populous states (California, Florida, and North Carolina). This suggests that the complex of factors producing high pedestrian death rates may differ from one State to another. Also, none of these rates consider pedestrian exposure or vehicle exposure, the proportion of pedestrians by age, or many other possible contributing factors.
At the other end of the spectrum, the 10 States with the lowest pedestrian death rate per 100,000 population also differ. The lowest 10 include rural States, cold weather States, industrial heartland states, and the agricultural midwest. This may be indicative of lower amounts of walking by pedestrians and/or less pedestrian interaction with high volumes of motor vehicles.
When Do Pedestrian Collisions Occur?
A. TIME OF DAY
Collisions can and do occur at any time, but there are trends involving the time of occurrence. Table 6 shows how pedestrian crashes, injuries, and fatalities vary by time of day.
Table 6 can be used to contrast the time of occurrence for "all" pedestrian crashes, fatally injured pedestrians, and pedestrians injured but not killed.
Among the sample of pedestrian collisions (Hunter et al., 1996), 62 percent happened during the day between 6:00 a.m. and 6:00 p.m. Consistent with those daytime numbers are other findings, based on data from urban areas. These show general agreement that the peak time for pedestrian crashes is between 3 p.m. and 6 p.m. This peak represents about 30 to 40 percent of the collisions (Fatal Accident Reporting System, 1990; Knoblauch, 1977; Davis and Huelke, 1969; Cove, 1990) and the proportion decreases on either side of this period. Smaller secondary peaks from 7 a.m. to 9 a.m. and Noon to 1 p.m. were reported by Smeed (1968).
NHTSA figures (Traffic Safety Facts, 1995) show that the distribution by time of occurrence for pedestrian injuriesis somewhat similar to that of all collisions (about 63 percent of injuries concentrated from 6:00 a.m. to 6:00 p.m., compared with 61 percent for all collisions). This was also reported by Cove (1990), showing pedestrian injury crashes to have a major peak between 3 p.m. and 7 p.m. and a minor peak between 7 a.m. and 9 a.m., based on data from the National Accident Sampling System (see figure 7).
The time distribution of pedestrian fatalities stands in sharp contrast to the above. Among pedestrian fatalities, 62 percent happened at night (table 6). Fatal pedestrian crashes peak in the evening hours (Cove, 1990), between 5 and 11 p.m., including one minor peak from midnight to 2 a.m. (See figure 8).
This trend in fatalities could be partly associated with rural pedestrian crashes involving high-speed vehicles and pedestrians walking along a dark road or in some cases lying unconscious (sleeping) in the road. In fact, in North Carolina, 10 percent of all pedestrian fatalities involve a pedestrian lying in the road (North Carolina Traffic Accident Facts, 1990). Thus, the role of alcohol among adults is thought to be involved in nighttime pedestrian fatalities. Fatality figures suggest that the victim profile would vary by time of day. Alcohol is a greater factor at night, but children are less likely to be involved at night.
Both the severity of pedestrian injury and the type of pedestrian crash vary by time of day. Data from six states, reported by Hunter et al. (1996), illustrate this point. Based on a sample of all pedestrian crashes, table 7 shows a breakdown of pedestrian crash types according to four different light conditions (i.e., related to time of day), distributed as follows:
These four categories create a rough scale from the greatest amount of light (daylight) to the least (dark, street not lighted), thus partly related to time of day, and table 7 shows the crash types that are most frequent under each light condition. Though daylight hours account for about 61 percent of all crashes in Hunter's study, more than 70 percent of some types of crashes occur in the daytime (e.g., 73 percent of pedestrian crashes related to buses occur in daylight).
Although unlit rural streets and roads might be considered a considerable hazard for pedestrians, the data in table 7 show that these conditions account for a much smaller percent of all crashes than, for example, walking along a roadway in daylight.
The nighttime association with fatal crashes is emphasized in a study of the effects of daylight savings time (DST) (Ferguson et al., 1995). This study asserts that if DST were retained year-round, approximately 900 additional traffic deaths would be avoided (727 pedestrians and 174 motor vehicle occupants, say the authors) because the DST clock setting results in more daylight at the end of the day, when fatal crashes are more likely, an advantage that is apparently not fully offset by the comparably less daylight during the morning hours.
B. DAY OF WEEK
Pedestrian crashes also vary by day of week. Available data indicate that pedestrian crashes are overrepresented on Friday and Saturday and are underrepresented on Sunday. These trends may be related to such factors as: the amount of walking by day of week; less pedestrian interaction with rush-hour traffic; and/or less late-night drinking and walking. A Wayne County, MI, study (Davis and Huelke, 1969) reported that 35 percent of the crashes there occurred on Friday and Saturday. This was especially true for children, with Friday being the highest crash day. Similar patterns were found for urban, suburban, and rural data samples from a number of U.S. cities and counties (Knoblauch, 1977). The highest overrepresentation of pedestrian crashes was on Friday.
Data dealing with pedestrian fatalities also reveal that Friday and Saturday have the greatest percentages of such crashes for both rural and urban areas, with pedestrian fatalities nearly constant for Sunday through Wednesday (see figure 9). Pedestrian crashes resulting in non-fatal injuries were most prevalent on Fridays and lowest on Sundays (Cove, 1990).
C. MONTH OF YEAR
Beyond the trends by hour of day and day of week, there are also differences in U.S. pedestrian crashes by season of year, mediated in part by factors related to pedestrian age. Figure 10 shows that among older pedestrians, more crashes occur during the fall and winter months, whereas among younger pedestrians more occur during the spring and summer (Zegeer et al., 1993).
Consistent with the above, a study conducted in Wayne County, MI (Davis and Huelke, 1969) showed that more pedestrians (13 percent) were killed during December than in any other month. A study of rural and urban data samples of U.S. areas (Knoblauch, 1977) showed December to be the month having the greatest overrepresentation. Nationwide pedestrian fatalities in 1989 were found to be highest in September through January (Figure 11), months with fewer daylight hours and more inclement weather.
D. SUMMARY: WHEN DO PEDESTRIAN COLLISIONS OCCUR?
Who Is Involved in Pedestrian Crashes?
Table 8 shows that the largest percentage of pedestrian deaths is seen for the two age groups that span the 25-44 year age range. A total of 31.8 percent (15.4 percent + 16.4 percent) of pedestrian fatalities occur in this age range, a greater percentage than for the 5 preceding age groups combined. (Note the varying age intervals used in this table prepared by NHTSA [Traffic Accident Facts, 1995.] The 25-34 age category spans 10 years while the next lower category (21-24) only covers 4 years. Thus, the 15.4 percent of fatalities that falls in the 25-34 year category is not as much greater as might appear when compared with the next younger category, since the latter includes data from less than half as many years. Note also that 13.9 percent of fatalities fall in the oldest category [greater than 75 years old]; however, the open-ended age span of this category includes data from more years than the immediately preceding category. Overall, more than twice as many male pedestrians (3742) are killed as female (1727) pedestrians.
The final column shows fatalities expressed as deaths per 100,000 persons in a given age category. The oldest age category stands in greatest contrast to the rest.
The oldest age category does not contrast as sharply to the next younger age groups in terms of injury statistics (non-fatal). This may stem in part from the fact that older pedestrians are more likely to succumb to their injuries than younger adults. Children may be more likely to survive collisions on lower speed neighborhood streets. The difference in injuries between male and female pedestrians is less than that for fatalities. The ratio of male/female pedestrian injuries is 1.4 while the ratio of male/female pedestrian deaths is 2.2.
For "all" pedestrian crashes (Hunter et al., 1996), the distribution by age is more similar to injury crashes than to fatal crashes. Also, the ratio of male to female differs. In fatal crashes, there are more males in every age category, but in "all" crashes, males make up fewer than half of the cases in several age categories, including the 10-20 year age range and the categories that include persons 55 years old and older. In this database, 61 percent of the pedestrians involved in a collision were males. In short, samples of fatal pedestrian crashes are more likely to involve males and ages 25-54 compared to nonfatal pedestrian crashes.
Table 9 shows a relationship of crash type to age of the collision-involved pedestrian. Certain crash types are overrepresented in certain age groups. Among the youngest age group shown here, 0-9 years old, the intersection dash and the midblock dash account for 41 and 55 percent, respectively, even though only 19 percent of all crashes affect children in this age group. For the 10-14-year-old category, the crash types most overrepresented are bus related (24 percent) and the intersection dash (23 percent), compared with 11 percent overall. For the oldest age group, backing vehicles seem to constitute the greatest risk (19 percent).
Special mention should be made of children in pedestrian crashes because of their particular vulnerability. One study (Dunne et al., 1992) asserts that parents overestimate their children's ability to handle street crossings. They overestimate what the children know and how well they will perform. The discrepancy between expectations and performance is greatest for the younger children (5 years old).
A German study reported that a significant portion of pedestrians younger than 6 years old and involved in a collision on the way to a playground were not accompanied by an adult (Kloeckner, et al., 1989). Injury severity was also greater in crashes involving unaccompanied children. Another study showed that children struck in a pedestrian collision that caused injury showed slightly poorer performance on a laboratory vigilance test than did a closely matched control group who were not in a collision (Pless et al., 1995).
Other factors of a more global, societal nature also are associated with a higher likelihood of a child's involvement in a pedestrian collision. One study reports that children are four-to-five times more likely to be in a pedestrian collision in poor neighborhoods compared with well-to-do neighborhoods (Cagley, 1992). Another study reported the statistical association of lower family cohesion with a higher risk of a child pedestrian crash (Christoffel, 1996).
Children's risk for a pedestrian collision reflect a number of factors, some of which can be addressed by a traffic engineer, while others are more properly addressed by various disciplines.
More males than females are seen in every single age category for fatal crashes (table 8). The ratio of male to female deaths varies from 3.6 to 1 in the 21 to 24 age group, down to 1.3 to 1 in the oldest age group. Despite the greater numbers of females compared to males in the 65+ age group, males slightly outnumber females in pedestrian fatalities; indicating that if the male and female populations in this age group were equal, the ratio of male to female fatalities would be even higher. Even in the youngest age group, pedestrians less than 5 years of age, the ratio of deaths among boys compared with girls is 1.7 to 1. In one study, boys were found to be involved in about twice as many pedestrian crashes as girls from 5 to 7 (AAA Pedestrian Safety Report 1984, 1977). Such differences between males and females have been observed for pedestrians as young as 2 years old. In an analysis of all pedestrian collisions in North Carolina for 1993, 1994, and 1995, crash records of 1,336 pedestrians were reviewed for those 12 years old or younger (Campbell, 1996). In this group, 61 percent were male. Even among 3-year-olds in pedestrian crashes, 62 percent were male; among 2-year-olds, 66 percent were males. Only for the 2-year-old and younger category (54 percent male) did the proportion approach the baseline proportion of males seen in birth statistics. This indicates the possibility of fundamental differences between the behavior of young boys and girls, and/or in the way they are supervised, even as toddlers.
The trends for pedestrian crashes by gender are somewhat different when looking at non-fatal injuries (table 8). The heavy preponderance of males is not seen in each and every category. Also, the 5-to-9 and 10-to-15 age categories account for a larger proportion of total injuries than any other age categories, contrary to the trends seen when only fatalities are considered. For the sample of all crashes, female involvement is relatively greater. Males are overrepresented, and the degree of overrepresentation is greater in fatal crashes than in non-fatal crashes.
Studies have reported that alcohol impairment is a major problem for pedestrians and drivers. Most states have a statutory definition of driver impairment of 0.08 or 0.10 blood-alcohol concentration (BAC). While it is illegal to drive a motor vehicle with BACs of 0.10, there is no such prohibition for walking.
One study reported that from 1980 and 1989, between 37 and 44 percent of fatally injured pedestrians had BACs of 0.10 percent or greater (Alcohol Fatality Facts, 1990). These percentages were slightly lower than for fatalities involving passengers, vehicle operators, and motorcycle operators. Among adults pedestrians killed in 1989 nighttime collisions with motor vehicles, 59 percent had BACs of 0.10 percent or greater, while only 31 percent had no alcohol in their blood.
One Phoenix study showed that 29 percent of all pedestrians involved in collisions had been drinking, while only 4.4 percent of the drivers had been drinking before the collision (another 13 percent of these drivers had an "unknown" physical condition, likely due to "Hit and Run" [Cynecki, 1998]. Considering that many pedestrians are younger than legal drinking age, but most of the drivers are of legal drinking age, this difference is substantial.
The percentage of fatally injured pedestrians with high BACs (0.10 percent or more) did not decline during the 1980s. This contrasts a 20-percent decrease in high BACs for drivers during that period (Alcohol Fatality Facts, 1990; Williams and Lund, 1990).
A study of motor vehicle fatalities in North Carolina between 1972 and 1989 (figure 13) showed that between 42 and 61 percent of pedestrian fatalities involved pedestrians under the influence of alcohol (i.e., BAC of 0.10 percent or greater). This compared with the range of 53 to 64 percent of drivers of single-vehicle collisions who were under the influence. Among 176 fatally injured pedestrians tested in 1989 (Rehabilitation Alcohol Test, 1990), the following distribution was reported:
Alcohol and age are associated in pedestrian crashes (Zegeer et al., 1993). The North Carolina data shown in figure 14 are the distribution by age of pedestrians in all reported crashes involving use of alcohol.
The FARS data address alcohol use among fatally injured pedestrians, broken down by pedestrian age. Alcohol is most often present for pedestrians 25-44 years old, with the proportion declining for older pedestrian age categories. These findings indicate that part of the overall age distribution in pedestrian crashes is related to alcohol use, although the involvement of the elderly in fatal pedestrian crashes seems likely to be attributable to factors other than alcohol.
As might be expected, pedestrian sobriety is also related to crash type (Hunter, et al., 1996). As seen in Table 10, 15.4 percent of all pedestrians in a collision were reported to be using alcohol or drugs before the collision. For certain crash types, however, the proportion is higher. For the crash type of walking along the roadway, the alcohol percentage for pedestrians is nearly double (29.6 percent). Adding to the significance of that figure is the fact that for drivers, the "walking along the road" crash type is likewise more frequent for drivers using alcohol or drugs (Hunter, et al., 1996).
A 1992 study by Dickman and Cope (1992) reported on 4,329 injured pedestrians, of whom 2,109 (49 percent) had been drinking. The blood alcohol levels were:
A substantial proportion of those who had been drinking had seriously elevated BACs. Not only is alcohol use by pedestrians a significant factor associated with involvement in a crash, but it is also reported that excessive amounts of alcohol significantly increased the likelihood that death will result from the collision (Miles-Doan, 1996).
D. VEHICLE TYPE
Another factor related to the "who" of pedestrian crashes is vehicle type. In a NHTSA report (Traffic Safety Facts, 1995, table 93), pedestrian collision percentages with regard to vehicle type are:
Light trucks as well as the "other" vehicle classes are more often represented in fatal pedestrian crashes than in injury crashes.
E. SUMMARY: WHO IS INVOLVED IN PEDESTRIAN COLLISIONS?
Where Do Pedestrian Collisions Occur?
A. RURAL VERSUS URBAN
The rural/urban distribution of pedestrian crashes is given in table 11, based on estimates by the NSC (Accident Facts, 1994). Of the estimated 71,200 pedestrian collisions in the United States in 1993, 75 percent occurred in urban areas, where pedestrian traffic is much higher than in rural areas. The table also shows that rural areas account for only 23.1 percent of nonfatal injury pedestrian crashes, but 45.2 percent of fatal pedestrian crashes. The overrepresentation of fatalities in rural areas is most likely because of the higher speeds and more severe crash types in rural areas.
B. LAND USE
Several studies that categorized pedestrian crashes by land use are summarized below:
Thirteen major cities (Snyder and Knoblauch, 1971)
Sample - 2,100 pedestrian crashes
Wayne County, MI (Davis and Huelke, 1969)
Sample - 268 Fatal pedestrian crashes
Rural and suburban area sample (Knoblauch, 1977)
Tucson, AZ (Nizlek, 1984)
The four studies vary in definition of categories and outcome, but they are similar in the prominence of pedestrian crashes in residential and commercial areas where most pedestrian exposure occurs.
C. TRAFFIC CONTROL BY CRASH TYPE
Another way of describing pedestrian crashes is by the nature of any traffic control present at the pedestrian crash site. Table 12 is based on the aforementioned 1992 sample of 4,329 pedestrian crashes of all kinds (Diekman and Cope, 1992), and the following overall distribution is seen with respect to traffic controls:
Specific types of crashes markedly depart from the above distribution. Thus, whereas 74 percent of pedestrian crashes happen where there is no traffic control, the value is 96 percent for crashes involving walking along the roadway. On the other hand, only 15.5 percent of pedestrian crashes that involve turning at an intersection occur where there is no traffic control (as compared with 74 percent overall). Overall, 7 percent of pedestrian crashes occur where a stop sign is the traffic control, but for vehicles turning at an intersection, the value is higher at 20 percent. As a final example, 17.3 percent of all pedestrian crashes occur at signalized locations, but the value is 63 percent for crashes involving vehicles turning at an intersection. Thus, a strong association with crash type accompanies the overall distribution of pedestrian crashes with respect to traffic control types.
SPEED LIMIT BY CRASH TYPE
Table 13 shows pedestrian crashes according to the speed limit posted where the collision took place. The posted speed limit is certainly not a perfect indication of the actual speed at which a collision may occur. The overall distribution is:
Some specific crash types markedly differ from this breakdown. Pedestrian crashes involving a backing vehicle are higher on roads with low speed limits than overall (50 percent versus 27 percent). In contrast, for pedestrian crashes involving walking in the roadway, sites with higher speed limits account for more than expected (37 percent versus 12 percent), with a correspondingly smaller proportion at the lower speed limit sites (14 percent versus 27 percent).
E. INTERSECTION VERSUS NON-INTERSECTION
Almost 60 percent of U.S. urban pedestrian crashes occur at places other than intersections. In rural areas, the figure is closer to 67 percent. When considering fatalities, the proportion remains essentially the same in urban areas. In rural areas, approximately 85 percent of the deaths occur at places other than intersections (Accident Facts, 1988).
A 1977 survey of child pedestrian crashes in more than 1,900 cities indicated that approximately 75 percent occurred at non-intersection locations. This increases to between 80 and 90 percent for the 5-year-old-and-under age group (AAA Pedestrian Safety Report, 1984). This high percentage is likely associated in part with the high incidence of young children running into the street at midblock locations.
A 1989 summary of pedestrian injuries and deaths by age was estimated by NHTSA for intersections and nonintersections. As illustrated in figure 16, a majority of crashes involving pedestrians up to age 44 occurs at nonintersections. For ages 45-64, there almost is an equal number of pedestrian crashes at intersections and nonintersections. For pedestrians age 65 and older, the trend reverses, and more are struck at intersections than at nonintersections. Although only limited pedestrian behavioral and survey data are available concerning choice of crossing locations, older pedestrians generally are more likely to cross at intersections than are younger ones (General Estimates System, 1990).
F. SUMMARY: WHERE DO PEDESTRIAN COLLISIONS OCCUR?
How Do Pedestrian Collisions Occur?
A. CONTRIBUTING FACTORS
A considerable number of factors, singly or in combination, increase the likelihood of a pedestrian collision. Table 14 shows the breakdown of a series of contributing factors related to 5,073 pedestrians involved in crashes (Hunter, et al., 1996). These are shown in four groups:
Pedestrian contributing factors
It is noteworthy that for each of the four major data groupings, the largest single category is "None indicated." Thus, despite numerous categories for assigning crashes, an appreciable portion of crashes defy such categorization. (This is all the more noteworthy because a main purpose of Hunter's study was to categorize pedestrian crashes.)
With respect to "Pedestrian Factors", the largest single category is "ran into road." Yet this category accounts for only 15 percent of collisions. Many other contributing factors are cited, but each is identified in only a small percent of cases. The largest specific "Roadway Factor" is "vision obstruction" (11 percent). For drivers, the largest category is "failure to yield right-of-way."
"Hit and run" is a somewhat larger category in the driver group, but such a classification tells nothing about the driver's actions that contributed to the crash, only that unlawful flight took place afterwards. Some characteristics of the hit-and-run situation are presented for fatal crashes in a 1995 study (Solnick and Hemenway, 1995). The authors examined the FARS database for 1989-1991 and listed factors that were over- or underrepresented in fatal hit-and-run situations. Overall, 20 percent of drivers left the scene:
Vehicle factors are cited in only 12 percent of the cases. Because numerous factors can contribute to pedestrian collisions, and because each individual factor accounts for only a small portion of the crash total, a successful pedestrian collision prevention program must pursue many different countermeasures. Only a small portion of total collisions will be reduced by any given countermeasure.
Table 15 shows a breakdown of who was primarily responsible for causing a pedestrian collision according to a study that reported that police ascribe fault to the pedestrian in 43 percent of cases and to drivers in 35 percent (Hunter et al., 1996). These are only the overall values, however. In specific circumstances the results vary, as shown in table 16. For example, though drivers are at fault in about 35 percent of the situations overall, the figure is 88 percent for driver violations at intersections, and 79 percent in situations involving vehicles turning at an intersection. Drivers alone are at fault in as few as 18 percent of crashes involving pedestrians walking along the roadway, but are at least partially at fault in another 35 percent of these collisions.
When the pedestrian alone is at fault (43 percent of cases overall), the situation varies by crash type. When a vehicle is backing, the pedestrian is adjudged at fault only 10 percent of the time, but in an intersection or midblock dash, the pedestrian is adjudged at fault 91 percent of the time.
B. CRASH TYPES AND CAUSAL BEHAVIOR
Several major U.S. studies of pedestrian behavior were based on field observations, on interviews with pedestrian safety professionals, and on data from collision reports (AAA Pedestrian Safety Report, 1984; Rehabilitation Alcohol, 1990; Accident Facts, 1988). These studies categorized crash types for urban, rural, and freeway locations as shown in Tables 17, 18, and 19 (Model Pedestrian User's Manual, 1987 edition, Knoblauch, 1975; Knoblauch, 1977; and Knoblauch, 1978). The objective of these studies was to identify crash causes and to develop countermeasures.
Another source of information is a study of freeway pedestrian crashes (Knoblauch et al., 1978), which examined driver and pedestrian activities leading to freeway pedestrian crashes. Table 20 gives the percentages of various driver activities preceding the crash, such as going straight, driving off the road, etc. The percentage distribution of pedestrian activities, such as running across the freeway, standing next to a disabled vehicle, etc., is in table 21.
A 1980 study by Habib identified causal factors related to pedestrian crashes at intersection crosswalks and recommended possible solutions. While 51.4 percent of such pedestrian crashes in Habib's study involved a through vehicle, left-turn vehicle maneuvers were involved twice as often as right-turn crash maneuvers (24.8 percent vs. 13.1 percent). The left-turn maneuver was nearly four times as hazardous as the through-movement in terms of collisions and exposure. Also, driver error was found to increase when the left-turn movement was made as compared with right-turn maneuvers. Factors identified as contributing to the left-turn crashes with pedestrians include driver visibility problems, poor driver habits, and signal location.
Obviously, a left-turning driver has a complex task. Oncoming traffic must be monitored to identify a safe gap to permit the turn. Traffic coming up from behind is also factor in the safety of the left-turning driver (risk of turning vehicle's being struck from rear). Add to this the necessity to monitor pedestrian traffic at the left of the driver's field of view and it is easy to realize that the left turn is a high demand situation for the driver. Solutions proposed by Habib include changes in vehicle design to improve driver visibility, location of an additional signal mounted on the left far-side of the sidewalk, improved crosswalk illumination, and driver education concerning the problem.
A study of the causes of pedestrian collisions in Arizona by Matthias and Stonex (1985) found that urban pedestrian collisions and fatalities tended to occur on wide, high-speed arterial streets. Causes of approximately half of the pedestrian collisions were failure to yield by the driver or pedestrian and failure to use the crosswalk. The authors concluded that there was little in the way of engineering countermeasures that would be useful. They indicated that public education, particularly for children under 14 years of age, appeared to be the most useful countermeasure.
Many factors characterize pedestrian crashes. Some lend themselves to traffic engineering intervention to improve safety, but others are associated with general societal characteristics and generally beyond the traffic engineer's influence. For example, the 1990 Nationwide Personal Transportation Survey (Antonakos, 1995) characterized pedestrians as: slightly lower in socioeconomic status; less likely to be employed; less likely to own a motor vehicle; and less likely to be licensed to drive a motor vehicle.
Still, almost no one can avoid occasional pedestrian status.
C. SUMMARY: HOW DO PEDESTRIAN CRASHES OCCUR?
How Severe Are Pedestrian Collisions?
Although many pedestrians are killed in motor vehicle collisions, most pedestrian collisions do not result in fatal injuries, in spite of pedestrians' vulnerability in a collision with a 908-kg or 2000-lb rigid motor vehicle. Data from NHTSA (Traffic Safety Facts, 1995) show that in 1994 there were:
Other data sources indicate that, in some pedestrian crashes, the police report indicates that the pedestrian is not injured. The injury scale shown here is one used by many police departments. The five-category scale is shown below along with the injury distribution reported by Hunter et al., 1996, who reported on 5,073 pedestrian crashes.
Table 22 shows how this injury distribution differs according to type of crash. Note the crash type categories that indicate most and least severe injuries. As to the fatal category, the greatest risk is “Walking along the roadway” (13.3 percent versus 6.1 percent fatal overall). At the other end of the scale, the crash types in which the largest number of pedestrians who escape with minor injuries or no injuries (that is, “C” injuries plus “No injuries”) are those involving a backing vehicle, 42 percent compared with 31 percent overall. Speed is a likely factor in these differences. Pedestrian crashes resulting in little or no injury are more likely to go unreported than those in which the pedestrian was seriously injured or killed.
Figure 20 shows that injury severity is also related to age (Zegeer et al., 1993). Older persons succumb to injuries that a younger person might survive. Fatal injuries among the youngest age group also are elevated relative to the lowest point of the curve. Fatalities among those 0-4 years are greater than for the next two older age groups.
B. SUMMARY: HOW SEVERE ARE PEDESTRIAN COLLISIONS?
Summary of Pedestrian Collision Statistics
Exposure-Based Hazard Index
Several studies have addressed the need for a measure of the relative exposure to hazards based on pedestrian crash statistics (Knoblauch, 1977; Lea et al., 1978; Goodwin and Hutchinson, 1977).
In 1977, Knoblauch developed a hazard index, defined as the ratio of the frequency with which any particular attribute was present in the crash sample to the frequency with which it was present in the general population at the site (base rate), at approximately the same time of day. Tables 23 and 24 show the relative hazards of pedestrian and vehicle actions derived in this manner. Table 23 shows that while crossing at a location other than an intersection was the most frequent action identified in the pedestrian crash sample, when compared with non-involved pedestrian actions at the site, it is a substantially less hazardous action than, for example, standing in the roadway. Similarly, out-of-control and backing vehicles are shown to be extremely hazardous to the pedestrian compared with turning vehicles. This procedure did not take collision severity into account when defining a hazard index (fortunately, these collisions are of low severity). The hazard index was computed based on the percent of pedestrian crashes divided by the relative exposure. Thus, for example, in table 24, the hazard index for "vehicle straight ahead" was relatively safe (hazard index of 0.9), even though 77.2 percent of pedestrian crashes involved a vehicle traveling straight ahead. This is because 85.1 percent of the vehicle exposure involved the straight ahead vehicle movements. Thus, the ratio of crash percentage divided by exposure percentage = (77.2) ÷ (85.1) = 0.9
In a later study, additional hazard relationships were updated, as shown in figure 22 (Knoblauch et al., 1987). Samples of pedestrian crashes and exposure were used to develop hazard scores for various pedestrian and vehicle characteristics. Scores of +1 or higher represent higher-than-average and -l- or less, a safer-than-average level of hazard. Pedestrians aged 1 to 4 years old had the highest hazard scores (+8.3); pedestrians aged 5-9 (+4.0), 10-14 (+1.2), and 60 and older (+1.7) had higher-than-average hazard scores. Running is more hazardous for pedestrians than walking (+4.7 vs. -1.9). Walking against a traffic signal has a hazard score of 5.1 compared with a score of -1.8 for crossing with the signal, while a right-turn-on-red maneuver by a motor vehicle was the most hazardous (score of +3.2) vehicle maneuver. Motorcycles and buses are associated with higher hazards to pedestrians (+3.3 and +2.9, respectively) than other vehicle types.
Conflict Analysis Hazard Formula
Conflict analysis has been used in a number of pedestrian crash studies to determine the hazard level as a basis for developing countermeasures (Knoblauch, 1977; Petzold, 1977; Zegeer et al., 1980; Husband and Sobey, 1978; Cynecki, 1980).
In a Rochester, MI, study, pedestrian conflicts were defined for school zones, based on observations at 10 school sites (Zegeer et al., 1980). These conflicts and events are:
The number of pedestrians crossing the street within the school zone, where pedestrians could be exposed to approaching vehicles, was also counted. The authors selected five conflict variables:
S - Severe conflicts
M - Moderate conflicts
R - Routine conflicts
J - Jaywalkers
C - Legal street crossings
Their relative contributions to hazard was determined by establishing weightings for an index using a delphi procedure. The result was the following formula for a subjective danger index (DI):
DI = 7.4 S + 2.8 M + l.0 R + 0.7 J + 0.2 C
This model was then proposed for use as a ranking tool for identifying high hazard school-zone sites and for guiding the selection of countermeasures.
A conflict analysis technique was developed by Cynecki (1980) for use in identifying hazardous pedestrian crossing locations. A total of 13 types of pedestrian conflicts were defined with assigned severity levels. The technique was tested at five locations and used to select pedestrian crash countermeasures. The author recommended further investigation of relationships among pedestrian conflicts and vehicle-pedestrian crashes.
Another study that attempted to determine the relationship between pedestrian-vehicle conflicts and pedestrian crashes based on a predictive model used discriminate analysis to develop crash-group models for the cities of Washington, DC and Seattle, WA (Davis et al., 1989). These models were used to predict intersection groups expected to have, for example, 0, 1, 2, or 3 or more pedestrian crashes. The crash groups were defined based on conflicts, as well as such exposure measures as pedestrian volume, vehicle volume, number of lanes, and type of traffic control. The models were considered particularly useful in setting priorities for hazardous locations and for evaluating various traffic control strategies. The model was developed using a subset of the crash data, and then was at least partly validated by being applied to the remaining crash data.