Home

MACHINARY INSURANCE POLICY

Enjoy web design
SYNTHESIS OF SAFETY RESEARCH RELATED TO SPEED
AND SPEED LIMITS

INTRODUCTION

This document provides a review of safety research related to speed and speed management. This review builds upon
a similar synthesis prepared in 1982. This synthesis highlights the relationships among vehicle speed and safety;
factors influencing speeds; and the effects on speed and crashes of speed limits, speed enforcement, traffic calming
and other engineering measures intended to manage speed.

Despite the substantial social and technological changes that have occurred since the original speed synthesis was
published, vehicle speed remains an important public policy, engineering, and traffic safety issue. Speed is cited as a
related factor in 30 percent of fatal crashes and 12 percent of all crashes (Bowie and Walz, 1994). Based on on-scene
investigations of over 2,000 crashes in Indiana by teams of trained technicians, excessive speed for conditions was
identified as the second most frequent causal factor out of approximately 50 driver, vehicle, and environmental factors
(Treat et al., 1977).

Excessive vehicle speed reduces a driver's ability to negotiate curves or maneuver around obstacles in the roadway,
extends the distance necessary for a vehicle to stop, and increases the distance a vehicle travels while the driver
reacts to a hazard.

The following pages present the results of a systematic review of the literature concerning safety research related to
speed and speed management. Initial listings of citations were generated using multiple keyword filters on several
bibliographic databases. The most productive databases were those of the National Technical Information Service
(NTIS), the Knight-Ridder Transportation Resources Index, and the Transportation Research Information Service (TRIS).
The initial inventory of approximately 700 citations was supplemented by searches of the Institute of Transportation
Engineers (ITE) index and more than 100 items that either predated the on-line data bases or otherwise were known to
be pertinent.

SPEED-SAFETY RELATIONSHIPS

Speed is the quintessential traffic safety issue, probably due to the clearly perceived relationship between vehicle
velocity and human capabilities and limitations. Even inexperienced drivers usually recognize the merit of reducing
their speed in uncertain or hazardous conditions to provide additional time for decision-making and action; driving
experience affirms this natural tendency for self-preservation. Good judgment, however, is not uniformly applied by the
operators of motor vehicles, nor are skills and abilities possessed in equal measure by all drivers. For these reasons,
vehicle speed could be related to traffic safety in two ways: (1) the greater a vehicle's velocity the less time available for
the operator to react to a hazard or for other motorists, bicyclists, or pedestrians to react to the vehicle; and (2) the
physical relationship of mass and speed to energy. If the first relationship exists, it would be expressed in the relative
incidence of crashes at different speeds. If the second relationship exists, it would be expressed in the relative severity
of crashes at different speeds. Research concerning these relationships is reviewed in the following paragraphs.

Speed and the Incidence of Crashes
In a landmark study of speed and crashes involving 10,000 drivers on 600 miles (970 kilometers) of rural highways,
Solomon (1964) found a relationship between vehicle speed and crash incidence that is illustrated by a U-shaped
curve. Crash rates were lowest for travel speeds near the mean speed of traffic, and increased with greater deviations
above and below the mean. The estimated travel speed from the accident records were compared to the speeds
measured at representative sites within each study section. The comparisons showed that crash-involved drivers were
over-represented in both high- and low- speed categories of the speed distribution.

Crash-involvement rates decreased with increasing speeds up to 65 mi/h (105 km/h), then increased at higher speeds.
Further, Solomon reported that the results of his study showed that "low speed drivers are more likely to be involved in
accidents than relatively high speed drivers." Cirillo (1968) in a similar analysis of 2,000 vehicles involved in daytime
crashes on interstate freeways confirmed Solomon's results, extending the U-shaped curve to interstate freeways, as
illustrated in figure 1. The analysis was limited to crashes involving two or more vehicles traveling in the same direction.

In these studies, the speeds of crashed vehicles were obtained
from police reports, driver's reports, or third party estimates -
sources that are subject to error and unknown reliability.
Another serious challenge to the internal validity of results is
that many of the crashes involving slow speed likely involved
vehicles that were stopping or slowing to turn or just entering
the road. Whereas, the speed data were collected at locations
within the study sections that were representative of the
average speed for the entire section but away from
intersections, driveways, and other locations having a major
effect on speed. These problems would tend to overstate the
risk of vehicles traveling at slower speeds.

To address these concerns, the Research Triangle Institute
(1970) used a combination of trained on-scene crash
investigators and a system of automated continuous speed
monitoring stations using sensors embedded in the roadway
pavement to obtain the speed of crash-involved vehicles and
accurate measurements of traffic speeds at the time of the
crash. Detailed data were collected on 114 crashes involving
216 vehicles on a state highway in Indiana with speed limits of
40 to 65 mi/h (65 - 105km/h). In about nine cases, speeds
could be linked to specific vehicles involved in crashes and
matched well with the estimates of the professional
investigators. More importantly, the investigators recognized
that vehicles slowing to negotiate a turn should be treated differently in the analysis than vehicles moving slowly in the
flow of traffic. The former involves a required slow speed to safely complete an intended maneuver, while the later is
more likely to reflect driver choice or limited ability.

West and Dunn (1971) reported the results of the Research
Triangle Institute studies. Crashes involving turning vehicles
accounted for 44 percent of all crashes observed in the study.
Excluding these crashes from the analysis greatly attenuated
the factors that created the U-shaped curve characteristic of
the earlier studies. Without vehicles slowing to turn, or turning
across traffic, the investigators found the risk of traveling much
slower than average was much less pronounced. Crash risk
was greatest for vehicles traveling more than two standard
deviation above the mean speed. As illustrated in figure 2, the
likelihood of being involved in a crash was extremely flat, with
little difference in crash risk for vehicles traveling within 15
mi/h (25 km/h) of the mean speed of traffic. Even excluding
turning crashes, the crash risk for vehicles traveling much
faster or slower was six times the average rate.

Munden (1967), following a different approach, reported similar
results for drivers in the United Kingdom who habitually drive
at deviant speeds. The speed of selected drivers were
observed and compared to the four preceding and four
following vehicles. For drivers observed more than once, those
traveling more than 1.8 standard deviations above or below the
mean traffic speed had significantly higher crash rates.
However, drivers observed only once did not exhibit a U-shape
relationship.

More recently, Australian researchers, Fildes, Rumbold, and Leening (1991), used self-reported crash data collected at
roadside from motorists whose driving speed had been unobtrusively measured. The researchers found a trend of
increasing crash involvement for speeds above the mean speed in both rural and urban conditions - similar to the
correlations reported in the early studies. However, no relationship between slower speeds and increased crash
involvement was found. In fact, Fildes and Lee (1993) report that the researchers, "...failed to observe any vehicles
traveling at the very slow speeds reported by Solomon on rural highways."

Figure 3 illustrates the speed-crash relationships identified by
Fildes et al, for the two rural and two urban sites used in their
study. The relationships are presented along with the U-shaped
curves derived from the early research on this topic. Some of
the difference between the results can be attributed to changes
in driver behavior (e.g., far less "drinking and driving" now than
in the 1950s and 1960s) and safety improvements in road and
vehicle design during the nearly half-century since the early
data were collected.

Harkey, Robertson, and Davis (1990)recently replicated the
U-shape relationship between speed and crashes on urban
roads. The researcher compared the police-estimated travel
speed of 532 vehicles involved in crashes over a 3-year period
to 24-hr speed data collected on the same section of
non-55-mi/h roads in mostly built-up areas of Colorado and
North Carolina. To partial address the concerns of earlier
studies and make the crash and speed data more comparable,
their analysis was limited to non-intersection, non-alcohol, and
weekday crashes. However, the estimated travel speeds of the
vehicles before the crash are questionable.

In defense of the early studies, it is important to note that the
researchers emphasized speed variance, rather than absolute
speed, as the primary culprit in the incidence of crashes;
speed variation is defined as a vehicle's deviation from the
mean speed of free-flowing traffic. Hauer's (1971) theoretical analysis of overtakings demonstrated that the number
vehicle interactions in terms of passing or being passed is a U-shaped curve with a minimum at the median speed. The
number of vehicles that a driver catches up with and overtakes increases with speed and the number of times a driver
is passed by others decreases with speed. Thus, the increased risk of crash involvement is a result of potential
conflicts from faster traffic catching up with and passing slower vehicles. The slower motorists go relative to the median
speed, the more overtakings and potential inter-vehicle conflicts encountered. This is illustrated in figure 4, which
compares the relative overtaking rates for a 100-km/h road with a standard deviation of 10 percent with the crash risk
form various studies. Hauer claimed "the indiscriminate public crusade against speeding should be replaced by a
balanced approach emphasizing the dangers of both fast and slow driving."

If conflicts created by large differences in travel speeds were a major factor in the likelihood of crashes, then one might
expect to find a large number of crashes involving two or more vehicles traveling in the same direction. Cerrilli (1997)
found less than one-third of all crashes and 5 percent of all fatal crashes in 1996 involved two or more vehicle traveling
in the same direction. Many of these likely occurred as a consequence of a vehicle slowing or stopping for cause (i.e.,
to make an intended maneuver or avoid striking a stopped vehicle or other hazard) and being struck from behind by a
vehicle following too closely or going too fast for the driver to stop in time to avoid the collision. By far, the predominant
crash type on rural roads is a single vehicle running off the road.

In a review of the issues associated with speed and traffic safety, Fildes and Lee (1993) reported that little research
was conducted concerning the relationship between speed and crash involvement during the 1970s and 1980s. Lave
(1985) revived the issue of speed variance as a contributor to crashes, suggesting that raising the speed limit would
result in fewer crashes in situations where variance was reduced by the higher limit. Lave concluded that "speed limits
designed to reduce the fatality rate should concentrate on reducing variance. This means taking action against slow
drivers as well as fast ones."

Similarly, Garber and Gadiraju (1988) reported that crash rates increased with increasing variance on all types of
roadways and that speeds were higher on roads with higher design speeds, irrespective of the posted speed limits.
They reported minimal variance when the posted speed limit was fewer than 16 km/h (10 mi/h) below the design speed
of the road. In the analysis, the researchers combined data from different road types (e.g., rural two-lane, urban
freeway, and rural freeway) which could lead to spurious results.

Speed And The Severity Of Crashes
The relationship between vehicle speed and crash severity is unequivocal and based on the laws of physics. The
kinetic energy of a moving vehicle is a function of its mass and velocity squared. Kinetic energy is dissipated in a
collision by friction, heat, and the deformation of mass. Generally, the more kinetic energy to be dissipated in a
collision, the greater the potential for injury to vehicle occupants. Because kinetic energy is determined by the square
of the vehicle's speed, rather than by speed alone, the probability of injury, and the severity of injuries that occur in a
crash, increase exponentially with vehicle speed. For example, a 30-percent increase in speed (e.g., from 50 to 65
mi/h [80 to 105 km/h]) results in a 69-percent increase in the kinetic energy of a vehicle.

The relationship between travel speed and the severity of injuries sustained in a crash was examined by Solomon
(1964), who reported an increase in crash severity with increasing vehicle speeds on rural roads. From an analysis of
10,000 crashes, Solomon concluded that crash severity increased rapidly at speeds in excess of 60 mi/h (96 km/h),
and the probability of fatal injuries increased sharply above 70 mi/h (112 km/h).

Bowie and Waltz (1994), in an analysis of tow-away crashes reported in the National Accident Sampling System over
a 7-year period, found that the chance of being injured in a crash depended on the change in speed at impact (delta V)
. As shown in table 1, the risk of a moderate or more serious injury was less than 5 percent when delta V was less
than 10 mi/h (16 km/h) and increased to more than 50 percent when delta V exceeds 30 mi/h (48 km/h).

Table 1. Injuries per 100 Occupants by Change in Speed (deltaV) at Impact
delta V
mi/h
Moderate Injury
AIS 2+
Serious Injury
AIS 3+
delta V
km/h
1-10
4.5
1.0
1-16
11-20
10.6
2.6
17-32
21-30
29.2
11.1
33-48
31-40
53.4
27.9
49-64
41-50
67.2
40.6
65-80
50+
69.3
54.3
80+

Joksch (1993) found that the risk of a car driver being killed in a
crash increased with the change in speed to the fourth power as
shown in figure 5. The risk of a fatality begins to rise when the
change in speed at moment of impact exceeds 30 mi/h (48
km/h) and is more than 50 percent likely to be fatal when the
change exceeds 60 mi/h (96 km/h). The probability of death
from an impact speed of 50 mi/h (80 km/h) is 15 times the
probability of death from an impact speed of 25 mi/h (40 km/h).

The fatality risk curve from an earlier study by O'Day and Flora
(1982) is also shown for comparison. The shift in the curve to
the right can be explained in part by improvements in vehicle
crashworthiness, seat-belt use, and emergency medical care
over time. (See TRB, 184; Evans, 1991; Zador and
Ciccone,1991; and FORS, 1992).

The relationship between impact speed and crash severity is
particularly critical for pedestrians, the most vulnerable road
users. In a recent review of the issues, the European Transport
Safety Council (1995) report that only 5 percent of pedestrians
died when struck by a vehicle traveling at 20 mi/h (32 km/h);
however, the proportion of fatalities increased to 45 percent at
30 mi/h (48 km/h) and to 85 percent at 40 mi/h (64 km/h).

Kloeden et al. (1997) compared the estimated speeds of over
150 cars involved in non-alcohol related injury crashes in 60
km/h speed zones in Australia with the free speed of cars measured at the same location at the same time of day and
day of week. The pre-crash traveling speeds were based on detailed investigations of each crash scene and
computer-aided crash reconstruction. The average and median speed of traffic was about 60 km/h ( 37 mi/h). As
shown in figure 6, the risk of being involved in an injury crash was lowest for vehicles traveling near or below the median
speed and increased exponentially at higher speeds. Vehicles exceeding the 90th percentile speed or traveling more
than 7 km/h faster (4 mi/h) than the speed limit and median speed had above average injury crash involvement rates.
Nearly 25 percent of the cars involved in injury crashes were traveling faster than 72 km/h (45 mi/h) compared to only 2
percent of free flow traffic.

Clearly, a research or engineering approach to speed management that ignores the injury consequences of vehicle
speed could lead to unintended results.

FACTORS INFLUENCING SPEED

In most of the crashes involving a slow-speed vehicle, the operator of
the slow-speed vehicle is either preparing for or in the process of a
maneuver that required a slow speed for safe execution (e.g., turning,
crossing, entering, or exiting). The current discussion focuses on the
conditions in which driving speed is a matter of individual choice.

Many different factors can influence the speed at which a motorist
chooses to drive. Speed choice can be influenced by driver age,
gender, attitude, and the perceived risks of law enforcement or crash.
Speed choice also is influenced by situational factors, such as
weather, road or vehicle characteristics, speed zoning, speed
adaptation, impairment, or simply "running late." These and other
factors are addressed in the following paragraphs.

Driver Attitudes and Behavior
Solomon (1964) identified the driver and vehicle characteristics
associated with speeding on rural highways during the late 1950s. He
reported higher mean speeds for young drivers, out of state vehicles,
buses, and late model passenger vehicles, especially high-performance
models. Other early studies linked driving speed to age, trip length, and presence or absence of passengers. More
recently, Fildes et al. (1991) unobtrusively measured the speeds of vehicles on urban and rural road segments in
Victoria, Australia, then stopped a sample of the vehicles to interview the drivers. The researchers found that younger
drivers, drivers without passengers, drivers of newer cars, drivers traveling for business purposes, and high mileage
drivers were more likely to drive faster than average and exceed the speed limit.

Mustyn and Sheppard (1980) found more than 75 percent of drivers claiming they drive at a speed that traffic and road
conditions permit, regardless of the posted speed limit. Although the motorists who were interviewed tended to
consider speeding to be one of the primary causes of crashes, they did not consider driving 10 mi/h (16 km/h) over the
limit to be particularly wrong. However, most of those interviewed considered driving 20 mi/h (32 km/h) over the limit to
be a serious offense.

Of all drivers involved in fatal crashes, young males are the most likely to have speed as a collision factor. In 1995,
nearly 40 percent of the fatal crashing involving male drivers 15 to 20 years old were speed related (NHTSA 1995). The
relative proportion of speed-related crashes to all crashes decreases with increasing driver age.

A recent study of the behavioral cues associated with driving while intoxicated (DWI) found that drivers who were
exceeding speed limits by 10 mi/h (16 km/h) or more were DWI (BAC>0.08) only in 9 percent of all nighttime
enforcement stops, but those driving more than 10 mi/h (16 km/h) under the limit were found to be DWI in 48 percent of
the stops (Stuster, 1997); driving under the speed limit does not include maneuvers that require slow speed. A previous
study of motorcyclist DWI detection found that 10 percent of speeding motorcyclists have BACs of 0.08 or greater
(Stuster, 1993). These probabilities of DWI are low compared to other behaviors, such as weaving, turning with a wide
radius, or drifting during a curve (all with probabilities of DWI greater than 50 percent).

Driving with excessive speed is a risk-taking behavior that often is found in association with other risk-taking behaviors.
For example, in 1995, only 37 percent of passenger vehicle drivers under 21 years old who were involved in fatal
crashes related to speed were wearing safety belts at the time of the crash. In contrast, 56 percent of drivers in the
same age group were properly restrained when speed was not a factor. For drivers 21 years and older, the percentage
of drivers involved in speed-related fatal crashes who were using restraints at the time of the crash was 34 percent, but
62 percent of drivers were restrained in fatal crashes that were not speed related.

Road Characteristics
Road characteristics contribute to the speeds at which drivers operate their vehicles. Warren (1982) reported the most
significant characteristics to be curvature, grade, length of grade, number of lanes, surface condition, sight distance,
lateral clearance, number of intersections, and built-up areas near the roadway. Tignor and Warren (1990) reported that
the number of access points and nearby commercial development are the factors that have the greatest influence on
vehicle speeds. In contrast, Fildes et al. (1987, 1989) found road width and number of lanes to have the greatest
influence on speed choice.

More recently, the European Transport Safety Council (1995) reported that width, gradient, alignment, and layout, and
the consistency of these variables, are the determinants of speed choice on a particular stretch of road. Road
characteristics determine what is physically possible for a vehicle, but they also influence "...what seems appropriate
to a driver." In this regard, individual perceptions of appropriate speed are influenced by the maintenance condition of
the road. For example, Cooper et al. (1980) found that average vehicle speeds increased by 1.6 mi/h (2 km/h) after
resurfacing major roads in the United Kingdom; no change in traffic speed was found in locations where surface
unevenness remained the same after resurfacing. Parker (1997) found no change in speeds on two rural highways and
a 3 mi/h (5 km/h) increase on two urban streets that were resurfaced and had the speed limit raised. It was not
possible to determine if the speed change was due to the higher speed limit or the resurfacing.

Roadway surroundings, especially proximity of tall objects to the road, also can influence the speeds at which
motorists choose to drive. Designing roadway features to influence driver perceptions of appropriate speeds is a
subject that will be addressed briefly in a subsequent section of this report.

The theory of speed adaptation predicts that apparent vehicle speed is influenced by the speed and duration of recent
travel in the vehicle. This adaptation to vehicular speed is the combined result of the visual, auditory, and proprioceptive
feedback associated with various rates of travel. Speed adaptation is a commonly experienced phenomenon that
results in an under estimation of speed after encountering a reduced-speed zone (Schmidt and Tiffin, 1969; Mathews,
1978). In short, according to the speed adaptation hypothesis, the perceived speed of one's own vehicle will be lower
than the actual speed if the driver has recently been operating the vehicle at a higher speed.

Several studies have explored the speed adaptation hypothesis. For example, Denton (1976) found that drivers who
had traveled at 70 mi/h (113 km/h) for three minutes tended to drive 5 to 15 mi/h (8 to 24 km/h) faster in a 30 mi/h (48
km/h) zone than drivers who had not previously driven at the faster speed. Casey and Lund (1987) found a lesser, but
more persistent, effect when drivers made the transition from 55 mi/h to 35 mi/h zones (88.5 to 56.3 km/h). Vehicle
speeds on streets and roadways leading from highways and freeways were greater than the speeds approaching the
highways and freeways, even though the posted speed limits are the same.

The review of speed-related issues prepared by Fildes and Lee (1993) for the Australian Federal Office of Road Safety
describes the cognitive aspects of speed perception. In particular, the authors summarize how the visual pattern that is
presented to a moving observer creates a blur of increasing magnitude at greater deviations from the fixation point. This
"retinal streaming" provides cues that are used to help estimate speed. Human capabilities, however, are limited in this
regard. Most research on the topic has found that drivers underestimate their speeds, especially at the medium and
high speed ranges. Further, research has found perceptual limitations that contribute to drivers underestimating the
curvature of an approaching bend (Shinar, 1977). Brummelaar (1983) and Fildes (1986) identified road curve features
that influence a driver's perception of curvature.

Environmental Conditions
Weather conditions influence the vehicle speed selected by most drivers. For example, reduced visibility due to fog
caused a 6 mi/h (10 km/h) decline in mean speeds on a freeway in Minnesota (CRC, 1995). Greater reductions in
speed can be observed under extreme conditions (Schwab, 1992). Although drivers reduce their speeds during poor
environmental conditions, this reduction is often accompanied by higher variation in speeds. Liang et al. (1998) in an
analysis of speeds on a rural freeway in Idaho found the standard deviation of speed doubles during fog events and
triples during snow. The researchers also found that drivers reduce their speeds a
Mail me