Bicycle Use and Hazard Patterns in the U.S.

Part I.
An Overview of the Bicycle Study
____________________
Gregory B. Rodgers, Ph.D.
Bicycle Project Manager, Directorate for Economic Analysis

Background
Bicycle riding is one of the most popular recreational activities in the United States.
The National Sporting Goods Association (1992) estimates that bicycle riding was the third
leading U.S. recreational activity in 1991, after exercise walking and swimming. In addition,
bicycle riding is an important means of transportation. The Bicycle Institute of America (1993)
estimates that there were about 4.3 million Americans who regularly commuted to work in
1992.
Bicycle riding is also a risky activity, as indicated by the large numbers of injuries and
deaths involving bicycles every year. According to the U.S. Consumer Product Safety
Commission's (CPSC) National Electronic Injury Surveillance System (NEISS), an injury
reporting system that consists of a statistical sample of the nation's hospital emergency rooms,
there have been about one-half million nonfatal bicycle-related injuries treated in hospital
emergency rooms every year since the early 1970s, when NEISS became operational. When
other medically-attended injuries are counted, such as injuries treated in physicians' offices,
there may be on the order of one million medically-attended injuries involving bicycles every
year. In addition, there are as many as 1,000 bicycle-related fatalities annually. The estimated
costs of these injuries and deaths to society are high -- approximately $8 billion annually -- and
suggest that injury reduction strategies with even modest levels of effectiveness could prove to
be cost-effective.
The CPSC has long had an interest in bicycle-related hazards and in promoting bicycle
safety. The agency began development of a mandatory standard for bicycles as one of its first
orders of business in 1973. The bicycle standard, which became effective in 1976,1 set safety
requirements for reflectors, wheels and tires, chains, pedals, braking and steering systems, and
for structural components such as frames and forks. More recently, the Commission has
provided a substantial amount of information on bicycle safety to the public and encourages all
riders to use helmets.

1

See, 16 CFR Part 1512, 41 Federal Register 4144-4154, January 28, 1976, and 16 CFR
Part 1512, 43 Federal Register 60034-46, December 22, 1978.
1

Bicycle safety is also promoted by many other governmental and non-governmental
organizations, and is of considerable interest to the health and safety research community. In
1991, Congress passed the Intermodal Surface Transportation Efficiency Act (ISTEA), an act
that required all states and metropolitan planning organizations incorporate programs and
facilities for bicyclists in their transportation plans. Also in 1991, the Department of
Transportation's (DOT) Appropriations Act instructed DOT to develop a plan to promote
bicycling and walking, and to enhance the safety of these transportation modes.
The interest of the health and safety community in bicycle safety is evidenced by the
large number of professional publications in the safety and medical literature. For the most
part, however, the published literature on bicycle hazards consists of injury analyses carried out
at the level of the individual hospital or in limited geographical areas. While these studies
provide valuable information about injury characteristics in various localities, there has never
been a comprehensive national study of bicycle use and hazard patterns designed to quantify
riding patterns and the rider and environmental factors associated with risk. Moreover, while
injuries resulting from bicycle-motor vehicle collisions have been evaluated extensively (Cross
and Fisher, 1977; Roland et al., 1979), little attention has been given to the analysis of bicyclerelated hazard patterns which do not involve motor vehicles, but which do account for the
great majority of injuries.
The CPSC bicycle project was intended to remedy some of these data deficiencies by
evaluating bicycle use and hazard patterns on a national basis. The remainder of this report
provides an overview of the methodology of the bicycle study, and the study findings.
Data and Methods
The CPSC conducted two nationwide bicycle surveys in 1991. The first, the "injury
survey," was conducted by the CPSC's Directorate for Epidemiology (EP) during calendar year
1991 and gathered information on a sample of 463 bicycle-related (nonfatal) injuries reported
through NEISS. NEISS injury reports were followed up with telephone interviews to collect
information on the characteristics and use patterns of riders with injuries treated in hospital
emergency rooms, the types of injuries suffered, and descriptions of the injury and hazard
scenarios.
EP identified 41 incidents (i.e., injury accidents reported through NEISS) which might
have involved mechanical failure or design problems. These incidents were assigned for on-site
investigations. The Directorate for Engineering Sciences (ES) evaluated these incident
investigations to determine if there were systematic mechanical hazards which might be
addressed by revisions to the existing mandatory standard.
The second survey, the "exposure survey," was a national random-digit-dial telephone
survey that collected information on the characteristics and use patterns of the general
population of bicyclists. The survey was conducted by Abt Associates, Inc., under the

2

direction of the Directorate for Economic Analysis (EC). It resulted in 1,254 completed
interviews with bicyclists from around the nation.
These surveys provided nationally representative samples of injured bicyclists who were
treated in hospital emergency rooms and of the general population of bicyclists. Because they
gathered parallel information on injured and noninjured bicyclists, the agency staff were able to
conduct a "risk analysis" by comparing the characteristics and use patterns of injured riders
with those who were not injured. In effect, the exposure data were used as "control data"
against which to compare the characteristics and use patterns of injured bicyclists. The aim of
the risk analysis was to determine and quantify the rider and environmental factors associated
with higher risk levels.
The Division of Human Factors (HF) reviewed the injury and exposure survey data
bases in light of behavioral studies applicable to bicycle riding. HF also evaluated the literature
on bicycle safety education and training, with emphasis on the developmental capabilities of
children.
The CPSC does not collect information on all bicycle-related deaths. However,
because deaths constitute an important bicycle hazard pattern, the study provides a brief
description and analysis of information on bicyclist deaths obtained from the National Center
for Health Statistics and from the National Highway Traffic Safety Administration's (NHTSA)
Fatal Accident Reporting System.
To complement the analysis of bicycle use and risk patterns, the agency purchased data
from a comprehensive 1990 survey of adult bicyclists commissioned by Rodale Press, the
publishers of Bicycling magazine (Rodale Press, 1991). The Rodale Press survey was
conducted by National Family Opinion, Inc., from its national consumer mail panel, and
included interviews with over 3,200 adult bicyclists who were 18 years-of-age and older.
Although the survey did not gather information on bicycle use by children, a major focus of the
CPSC project, it did gather data on a wide range of topics relevant to an analysis of the risk
and safety behavior of adult bicyclists. In many cases, its results were directly comparable to
the results of the CPSC exposure survey. It also provided market data, such as plans for future
purchases of bicycles and equipment, which were unavailable from other sources.
Characteristics of Riders and Injury Statistics
This section summarizes some of the important descriptive results from the 1991 injury
and exposure surveys, including the characteristics and use patterns of riders, and injury
statistics.
Rider Characteristics and Use Patterns
The results of the exposure survey are detailed in Part II. The exposure survey
confirmed the popularity of bicycle riding in the U.S. There are about 67 million bicyclists
3

who ride a total of about 15 billion hours annually. Most bicycle riding is for recreational
purposes, but almost 9 percent of riders use their bicycles primarily for commuting to work or
school.
Just over half of all bicyclists (52 percent) are males. In addition, a large proportion of
bicyclists are young. About 22 percent are under the age of 10 years and 40 percent are under
age 15. Young bicyclists ride more than the average for all bicyclists. Riders under age 15
reportedly ride about 300 hours per year, about 50 percent more than the average reported for
riders age 15 and older.
Most bicyclists (64 percent) ride a substantial proportion of the time on neighborhood
streets with low traffic volume, but sizable proportions also spend a lot of their riding time on
sidewalks and playgrounds (29 percent), bike paths (17 percent), and unpaved roads (18
percent); smaller proportions ride on major thoroughfares with high traffic volume (7 percent)
and on other unpaved surfaces or trails (11 percent).
Children under age 10 ride primarily on sidewalks, playgrounds, and neighborhood
streets; riders over age 10 are more likely to be found on neighborhood streets, bike paths, or
major thoroughfares. About 12 percent of bicyclists ride at least occasionally after dark.
However, less than one-third of these nighttime riders use headlights or taillights.
There are about 96 million bicycles in existence, but only about 66 million (69 percent)
were used in the year prior to the survey. Children's models (i.e., sidewalk or BMX/high rise)
account for over one-fourth of the bicycles in use. Of the adult models (i.e., lightweight
racing/touring, mountain, and middleweight/cruisers), the lightweight racing and touring
bicycles are the most common and account for about one-third (34 percent) of the bicycles in
use. Mountain bikes were first marketed in substantial numbers in the early 1980s, and now
account for about 17 percent of the bicycles in use.2
The Rodale Press survey findings for adult bicyclists (age 18 and over) are described in
Part VII of the study. They are generally consistent with the findings of the CPSC exposure
survey. The majority of adult bicyclists (62 percent) rode most often on neighborhood streets.
In addition, over half (57 percent) had access to community bike paths, and 28 percent had
access to extra wide roads or bike lanes. Many bicyclists said that "having safer places to go
riding" (35 percent) or "being able to ride to work" (14 percent) would encourage them to ride
their bicycle more often.
About one-fifth of the Rodale Press survey respondents expected to purchase a new
bicycle within 2 years. The mean expected outlay was $334, an 82 percent increase over the
mean price paid ($183) by recent purchasers. The Rodale Press results also indicate that

2

The term mountain bike refers to the class of bicycles that includes city, all-terrain, or
mountain bicycles.
4

mountain bicycles are increasing in market share. While 14 percent of recent purchasers said
that they had bought a mountain bicycle, 44 percent of those planning a purchase expected to
buy a mountain bicycle.
Characteristics of Victims, Injuries, and Injury Location
According to the analysis of the injury survey results, which are detailed in Part III of
the study, there were an estimated 588,000 bicycle-related injuries treated in U.S. hospital
emergency rooms in 1991. About 531,000 (90 percent) involved bicycle operators; the
remainder involved primarily passengers and bystanders.
About 62 percent of the injured operators were male. Most were also children: about
37 percent of the injured operators were under age 10, and 71 percent were under age 15.
Non-operators who were injured (i.e., primarily passengers and bystanders) were younger than
injured operators; about 66 percent of the injured non-operators were under age 10.
Injured bicycle operators also tend to be younger than the general population of bicycle
riders. Table 1 compares the ages of injured operators with those of the general rider
population from the exposure survey. As can be seen, children between the ages of 5 and 14
are disproportionately involved in accidents resulting in injury. While 5-to-14 year-old
bicyclists represent about 36 percent of riders, they account for about 68 percent of all
emergency room treated injuries.
Table 1: Distributions of Riders by Age
Age

Injured
All
Operators
Riders
(years)
(Percent)
(Percent)
# 4
3.1
3.8
5-9
33.6
17.7
10-14
34.7
18.6
15-24
10.9
16.3
25-44
13.4
32.7
45-64
3.1
9.5
$ 65
1.1
1.4
Total
100.0
100.0
Source: 1991 CPSC Bicycle Injury and Exposure Surveys.
Almost one-third (30 percent) of all operator injuries involved the head or face; 27
percent of these head/face injuries involved potentially serious diagnoses, such as fractures,
internal injuries, or concussions. Young children suffered a significantly higher proportion of
head injuries than older victims; 50 percent of the injuries suffered by children under age 10
involved the head or face, compared with 19 percent for riders age 10 or older.

5

Less than 3 percent of injury victims were admitted for hospitalization. This is about
the same rate of hospitalization (about 4 percent) for all product-related injuries reported
through NEISS in 1991.
Just over half of the operator injuries (53 percent) happened on roadways (i.e., surfaces
designed for use by motorized vehicles). About three-quarters of the roadway injuries
occurred on neighborhood streets; the remainder were on major thoroughfares and unpaved
roads. Riders age 25 and older were injured on highways or major thoroughfares more
frequently than younger riders. Bicyclists under age 25 who were injured on roadways were
more likely to be injured on neighborhood streets.
Another 12 percent of the operator injuries occurred on sidewalks and playgrounds;
most involved children under age 10. About 5 percent of the incidents occurred on unpaved
roads, 5 percent occurred on trails, and less than 1 percent occurred on bike paths.
Injury Hazard Patterns and Risk Factors
A major focus of the bicycle study was the evaluation of bicycle hazard patterns and the
bicycle risk analysis. This section summarizes the results of these analyses, which are
contained in Part III of the study. It also presents some complementary results from a risk
analysis of the Rodale Press survey data base, which can be found in Part VIII.
Hazard Patterns and Contributing Factors.3
An estimated 15 percent of the injuries involved collisions with moving objects, such as
motor vehicles, other bicycles, or animals. Another 13 percent involved collisions with nonmoving objects, such as parked cars, traffic signs, or fences. Incidents involving collisions or
near-collisions (i.e., swerving to avoid collisions) with moving motor vehicles accounted for
only about 10 percent of the injuries.
About 11 percent of the incidents occurred while victims were performing stunts, such
as jumping over ramps or speed bumps, or performing "wheelies." About 88 percent of the
incidents that occurred while performing stunts involved children under age 15, and 80 percent
involved male riders.
Respondents reported a number of factors that contributed to the incidents. Uneven
riding surfaces (e.g., bumps, ruts, curbs) contributed to about 27 percent, slippery surfaces
contributed to about 15 percent, and "going too fast" contributed to about 22 percent. Other
miscellaneous reported factors included mechanical failure, rider inexperience, inattention,

3

Unless otherwise noted, all injuries refer to injuries suffered by bicycle riders (i.e., bicycle
operators rather than passengers or bystanders).
6

riding the wrong size bicycle, or riding at night without a light. The use of earphones or
carrying young children in child carriers did not play a major role in injury scenarios.
Just over one-fifth of the injuries occurred under non-daylight conditions: about 5
percent were at night and 16 percent were at dawn or dusk. However, about 35 percent of the
injuries on major thoroughfares occurred under non-daylight conditions. Less than 3 percent
of the injuries occurred in rain or snow.
Risk Analysis and Risk Factors
Table 2 presents information on injury rates for various age groups. Based on the
results of the injury and exposure surveys, there were about 8.8 bicycle-related injuries treated
in hospital emergency rooms for every 1,000 riders in 1991. Riders 5-to-14 years of age have
the highest injury rate with about 17 injuries per thousand riders. The injury rate for all riders
age 15 and older is considerably lower than the child injury rate. However, when adjusted for
hours of annual use, the injury rate for riders over age 64 is similar to the child injury rate.
Although based on a small sample of older riders, riders over the age of 64 (who ride much
less than children) have an adjusted injury rate comparable to that of riders in the 5-to-14 yearold age group.
Table 2: Bicycle Injury Rates, by Age Group
Victim
Injuries per
Injuries per
Age
Thousand
Million
(years)
Riders
Hours of Use
All Ages

8.8

37.2

# 4
7.3
25.3
5-9
16.9
57.0
10-14
16.7
55.4
15-24
5.9
29.4
25-44
3.6
18.1
45-64
2.9
22.0
$ 65
7.2
61.0
Source: 1991 CPSC Bicycle Injury and Exposure Surveys
EP staff used a logistic regression model to determine and quantify the factors
associated with the injury risk.4 They estimated a general model which included riders from all
age groups. In addition, they estimated two separate risk models, one for riders under age 15
("children"), and one for riders 15 years of age and older ("adults"), because of the significant
risk differential between these two groups.
4

This statistical technique is used to determine the independent impact of each of several
factors on the injury risk. It is useful when a number of factors simultaneously affect the injury
risk.
7

The general model found a significantly higher risk for children under age 15. Holding
all other factors constant, the risk for a child under age 15 was over 5 times the risk for an
older rider. Most of the other results for the two age-specific models were similar to those in
the general model.
In the children's model, higher risks were associated with certain riding surfaces, time
of day, and population density. Children who rode during non-daylight hours, on streets, and
who lived in areas with greater population density were more likely to be injured. The risk on
streets was about 8 times the risk on bike paths, 3.4 times the risk on unpaved surfaces, and
about 1.7 times the risk on sidewalks. Riding under non-daylight conditions (e.g., at night,
dusk, or dawn) was about 3.6 times more risky than riding during the daytime. Rider gender
had no statistically significant effect on the injury risk.
In the model for riders 15 years of age and older, risk was also affected by riding
surface. As in the children's model, the adult risk was higher on paved roadways. The risk on
neighborhood streets was about 7 times the risk on bike paths and about 9 times the risk on
unpaved surfaces. Moreover, the risk on major thoroughfares, the highest risk riding surface,
was about 2.5 times the risk on neighborhood streets. As in the children's model, risk was
higher for riders who lived in areas with greater population density. However, there was no
significant difference in risk between daylight and non-daylight hours. Nor did rider gender
independently affect the injury risk.
In the Rodale Press survey of adult riders, about 9 percent of respondents reported that
they had had accidents in which they had crashed or fallen off their bicycle within 12 months of
the survey. These accidents may or may not have resulted in a medically-attended injury. An
analysis of factors associated with this accident risk was highly consistent with the results of
the EP risk analysis of riders 15 years of age and older. One especially noteworthy finding was
that the accident risk rose for riders over age 64; the risk for riders over age 64 was
significantly higher than for riders 25-to-64 years of age. (This finding was suggested in the
EP risk analysis, but was not significant, probably because of the small sample of riders over
age 64.) In addition, the accident risk was substantially higher on off-road trails (a type of
riding surface not evaluated directly in the EP risk assessment) than on other riding surfaces.5
Human Factors Evaluation of Children's Risk
HF reviewed the bicycle injury data on children and the existing literature on safety
education and training. This analysis, which can be found in Part IV, provides some
explanation for the higher risks for children under age 15, based primarily on the cognitive
immaturity of children. According to HF, bicycle riding is a complicated activity in which a lot

5

Riding on "unpaved roads" was combined with "other unpaved surfaces and trails" in the
EP risk analysis.
8

of information is vying for the attention of children. Children often do not have the ability to
filter all the information, or to filter it correctly.
According to available literature, children 5-to-14 years of age begin to test their skills
and experience many physical and cognitive changes. They may push their bodies physically in
ways that can lead to injury. In addition, boys tend to be more risk-taking than girls, as
evidenced in many studies. These factors may help explain why 88 percent of those injured
while performing stunts were under age 15, and why 80 percent involved boys.
The egocentric behavior of children (i.e., the inability to perceive other people's
viewpoints) also helps explain their higher injury risk. It is not until around the age of 10 that
children are able to consider the consequences of their actions. For example, children under
age 10 may not consider their behavior unexpected when they suddenly turn in front of a car or
dart out of a driveway, because that appears to them as the only way to go.
Evaluation of Mechanical Hazards
EP identified 41 incidents (i.e., injury accidents reported through NEISS) which might
have involved mechanical failure or design problems. These cases represented about 13
percent of the operator injuries, and were all assigned for on-site investigations. ES evaluated
the incident investigations to determine if there were systematic mechanical hazards which
might be addressed by revisions or amendments to the existing mandatory standard. This
analysis can be found in Part V.
The most frequently reported problems involved bicycle chains breaking or falling off,
brakes failing, and various components such as handlebars and brake components coming
loose. By mechanical component group, the 41 cases involved: brakes (15 cases); chains (13
cases); handlebars (6 cases); tires (2 cases); and gear cables, seats, spokes, handgrips, and
pedals, with one case each.
Although the cause of these alleged mechanical failures could not be absolutely
determined, ES concluded that poor bicycle maintenance and/or bicycle modifications were
contributors in a minimum of 9 cases and possible contributors in an additional 11 cases.
External conditions, such as slick road surfaces, were probable contributors in 4 cases. In
addition, operator behavior and unfamiliarity with a bicycle were described as possible
contributors in 12 cases.
Only 15 of the cases (representing an estimated 4 percent of emergency room treated
injuries) reported component malfunctions without indicating other likely contributing factors.
However, information was insufficient to determine if these incidents resulted from inherent
mechanical failure not attributable to poor maintenance, ill-advised modifications, or other
factors. ES concluded that there were no significant mechanical failure patterns that warranted
amendment or revision to the mandatory bicycle standard.

9

Bicycle-Related Deaths
Information on bicycle-related deaths is available from two sources: the National
Center for Health Statistics (NCHS) and NHTSA's Fatal Accident Reporting System (FARS).
In Part III of the study, NCHS data on deaths are discussed and compared to data from the
injury survey. The NCHS identified about 890 bicyclist deaths in 1989, the most recent year
for which data from that source are available. About 90 percent of the deaths involved motor
vehicles, compared to about 10 percent of the nonfatal injuries treated in hospital emergency
rooms.
According to the NCHS data, bicycle injury victims who died tended to be older than
those who were treated for nonfatal injuries in hospital emergency rooms. As shown in Table
3, about 63 percent of those who died were age 15 or older, and about 17 percent were age 45
or older. In contrast, about 29 percent of the nonfatal injury victims were age 15 or older, and
only about 4 percent were age 45 or older. In addition, fatal accidents were more likely to
involve males. About 85 percent of the fatality victims were male, in contrast to about 62
percent of the nonfatal injury victims.
Table 3: Age and Gender of Victims, by
Percent of Deaths and Injuries

Age (years)
# 4
5-14
15-24
25-44
45-64
$ 65
Total

Deaths
(NCHS, 1989)
(Percent)
2.0
35.0
21.3
25.2
9.5
7.0
100.0

Injuries
(NEISS, 1991)
(Percent)
3.1
68.4
10.9
13.4
3.1
1.1
100.0

Gender
Female
14.7
37.7
Male
85.3
62.3
Total
100.0
100.0
Source: National Center for Health Statistics 1989,
and the 1991 CPSC Bicycle Injury Survey
Fatal injuries also tended to involve a greater proportion of head injuries than did
nonfatal injuries treated in hospital emergency rooms. While the injury survey indicated that
30 percent of emergency room treated injuries involved the head or face, Sacks et al. (1991)
estimated that about 62 percent of all bicycle-related deaths involved head injury.
In Part VI of the study, bicycle-related deaths reported through NHTSA's Fatal
Accident Reporting System (FARS) are evaluated in conjunction with data from the exposure
10

survey. The FARS data are limited to deaths resulting from crashes with motor vehicles on
public roadways (about 90 percent of deaths), but since data were available for 1991, the
FARS data were directly comparable to data from the 1991 exposure survey. It was therefore
possible to estimate comparative risk factors for various gender and age categories by
comparing the distribution of the 1991 FARS deaths with estimates of riding exposure from
the 1991 CPSC exposure survey.
This analysis revealed that the fatality risk for male bicyclists, adjusted for riding
exposure, was almost five times the risk for female bicyclists. In addition, when adjusted for
exposure, the fatality risk for 16-to-24 year-old bicyclists was about 2.1 times higher than for
bicyclists under age 16. The relative risk of fatality was even higher for riders over the age of
44, and was highest for those over age 64. Riders over age 64 were about 3.2 times more
likely to be involved in fatal accidents than 16-to-24 year-old riders, and about 6.6 times more
likely to be involved in fatal accidents than riders under age 15.
Finally, riding after dark appears to contribute to the fatality risk. An estimated 23.5
percent of the deaths occurred between the hours of 9:00 p.m. and 5:59 a.m. Although
daylight conditions vary during the year and by region, most of these deaths probably occurred
after dark. Another 22.9 percent of the deaths occurred between 6:00 p.m. and 8:59 p.m.,
some of which probably occurred after dark. In contrast, only about 12.4 percent of riders
from the exposure survey reported that they engage in nighttime riding at least some of the
time. Nighttime riding therefore appears to be an important contributing factor in bicycle
deaths.
Bicycle Helmet Findings
While recent studies show substantial safety benefits from helmet use, they also reveal
that only a small proportion of riders actually use helmets. The exposure survey provides
valuable insights into current helmet usage patterns and on the reasons why riders use or do
not use helmets. This section summarizes the helmet usage patterns of bicyclists and the
statistical analysis of factors associated with helmet use, which are detailed in Part II. It also
describes the attempt in Part III to evaluate the impact of helmet use on the likelihood of head
injury.
Descriptive Results
The exposure survey found that only 11.8 million (18 percent) of the entire population
of about 67 million bicyclists wear helmets all or most of the time. Another 6 percent,
representing about 4 million riders, reported that they wear helmets sometimes, but less than
half of the time.
The proportion of children under age 15 who wear helmets all or most of the time was
about 15 percent. HF reports (in Part IV) that the low usage rate for children may be partly
related to peer pressure. Some studies show that children are not inclined to wear helmets if
11

their social group disapproves of helmet use. However, helmet use in all age groups appears
to be increasing. Just over half of the current users (53 percent) began wearing helmets in the
last two years.6
Nearly all of the 9 million riders who always wear helmets described "safety" as an
important reason for doing so. The "insistence of family members," was also important to
about half of those who always wear helmets. Usage patterns for 6.8 million riders who wear
helmets sometimes, but not all of the time, are apparently affected by risk perceptions. Many
said that they usually wear helmets when in traffic (40 percent) and when on long rides (25
percent). Many also reported that they are less likely to wear helmets when riding only a short
distance and when not riding in traffic.
Finally, when non-helmet users were asked why they do not wear helmets, nearly half
(48 percent) reported that they had never considered wearing helmets, 21 percent said helmets
were unnecessary, 19 percent said they did not wear helmets because they seldom ride in
traffic, and 16 percent said they had not gotten around to wearing them.
Helmet Use Patterns
In an analysis of factors associated with helmet use, the exposure survey data revealed
that the likelihood of helmet use increases with the amount of riding time. It is higher for those
who ride on major thoroughfares and bike paths, and is lower for those who ride on
neighborhood streets and on sidewalks and playgrounds. The relationship between age and
helmet use is more complex, suggesting that helmet use increases with age for frequent riders
and declines with age for infrequent riders. The results also suggest that children age 10 and
under are more likely to wear helmets, relative to older riders, than can be otherwise explained
by the general relationship between age and risk. The likely explanation is that enough parents
of young children require their children to wear helmets so that helmet use patterns of children
are distinguished from those of older bicyclists. Helmet use also increases substantially with
higher household education levels.
These relationships are illustrated for individual riders in a table at page 53. For
example, consider a male who rides 300 hours per year on neighborhood streets, and who has
(or, for children, whose parents have) no more than a high school education. The expected
likelihood of helmet use decreases from 9.9 percent for a 10 year-old rider to 6.8 percent for a
20 year-old rider. However, it rises again to 10.5 percent for a 40 year-old rider. In contrast,
for a 30 year-old female who rides about 50 hours a year on neighborhood streets, the

6

The Rodale Press findings for adults, described in Part VII, were similar. In 1990, only
about 15 percent of adult bicyclists wore helmets all or some of the time. However, the results
also suggested that helmet use was likely to increase substantially. About 10 percent of riders
who did not own helmets said they planned to buy one within 2 years. If plans materialized,
helmet usage rates would have increased to about 25 percent by 1992.
12

likelihood of helmet use rises from 5.4 percent if she has a high school education, to 16.4
percent if she has a college education. It rises further to 37.1 percent if she not only has a
college education but also rides primarily on major thoroughfares.
The analysis of the Rodale Press survey data on helmet usage patterns (in Part VIII)
came to similar conclusions. Helmet use increased with riding distances, and was higher for
bicyclists who ride primarily on major thoroughfares and off-road trails. In addition, helmet
use increased with household income, a variable not included in the analysis of helmet use
patterns from the exposure survey.
Helmet Effectiveness
Since helmets are intended to reduce the likelihood of head injury, EP used injury
survey data to examine the safety effects of helmet use by estimating the conditional probability
of head injury given that a helmet was worn. As described in Part III, the results of this
analysis were inconclusive, probably because the sample of helmet users was small (only about
12 percent of the injured riders were wearing a helmet at the time of accident), and possibly
because no information was available on riders who avoided injuries or whose injuries were
less severe because they were wearing helmets.
However, EP found evidence that helmets prevented or reduced the severity of some
head injuries. Helmets were damaged in 16 of the injury cases, about one-third of the cases in
which they were worn. In 11 of these cases (69 percent), the victim did not sustain a head
injury. In addition, in all 16 cases, the victim expressed the opinion that the helmet prevented a
head injury or made it less severe.
Conclusions and Implications for Injury Reduction
The bicycle study documented the large number of bicycle-related injuries and deaths
that occur every year, and evaluated the use and hazard patterns of bicyclists in the United
States. While the costs to society of bicycle-related injuries and deaths are enormous -- on the
order of $8 billion annually -- the bicycle study does not indicate any simple or direct remedies
to the hazards of bicycle riding.
Bicycle accidents result from a complex interaction of behavioral, environmental, and
mechanical factors. Efforts to reduce injuries must therefore be based on long term strategies
which systematically address risk factors on a number of fronts at the same time. The
behavioral factors leading to injuries, for example, might be addressed by training, or by
strategies that make riders aware of safe riding practices and the consequences of unsafe riding
practices. Environmental factors might be addressed by improving road design, or by
promoting the development of bike lanes and bike paths. Similarly, mechanical factors might
be addressed by product modification. In addition, all of these factors may be addressed by the
use of safety equipment which prevents or mitigates the severity of injury when accidents
occur.
13

Although the bicycle study could not quantify the causal relationship between the
behavioral, environmental, and mechanical factors and the injury risk, the study's results
indicate that the behavioral factors constitute an important component. A large proportion of
bicycle injuries result from behaviors which are risky or reflect poor riding judgment (e.g.,
stunting or riding too fast given the riding conditions). In addition, the cognitive and physical
immaturities of children are likely contributing factors in many of their injuries. The bicycle
study also found that environmental factors, such as riding terrain and riding conditions, play
an important role in the injury risk. On the other hand, while poor bicycle maintenance was a
hazard factor, the structure of the bicycle itself appeared to play little role in the injury risk.
The remainder of this section discusses these general conclusions, and their implications
for injury reduction.
Mechanical Factors
One purpose of the bicycle project was to determine whether there are significant
mechanical failure patterns that warrant amendments or revisions to the existing mandatory
standard for bicycles. Although there was no reason at the outset of the project to believe that
revisions were necessary, possible mechanical hazard patterns have not been evaluated on a
systematic basis since the standard went into effect almost 20 years ago. In addition, changes
in the bicycle market (such as the availability of mountain bikes) may have resulted in new
mechanical hazard patterns not envisioned in the original standard.
The bicycle study, however, provides no evidence that any bicycle type (e.g.,
lightweight racing, BMX, mountain, etc.) is inherently more hazardous than any other. Hazard
patterns involving bicycle types were found to be related primarily to the age and riding
patterns of users.
In addition, the ES review of the injury data found no evidence of systematic
mechanical hazards that would warrant amendments or revisions to the existing mandatory
standard for bicycles. Although mechanical failure was identified as a possible contributing
factor in as many as 13 percent of the injury reports, ES concluded that a large proportion of
these injuries involved poor bicycle maintenance and/or bicycle modifications, as well as
external riding conditions such as wet, slippery riding surfaces. Because of the findings
concerning bicycle maintenance and modification, ES recommends that both adults and
children be made aware of the importance of maintaining a bicycle in good working condition
and of the risks of modifying a bicycle.
Environmental Factors
The risk analysis revealed a substantial risk differential between paved roadways (which
are shared with motor vehicles) and bike paths (which are generally shared with other bicycles,
joggers, walkers, and skaters). When holding other factors statistically constant, the risk of
injury on neighborhood streets was about seven to eight times the risk on bike paths, and the
14

risk on major thoroughfares was even greater than on neighborhood streets. Moreover, about
90 percent of bicyclist deaths involve crashes with motor vehicles on public roadways.
These findings suggest that the riding environment should be an important focus of
efforts to reduce bicycle injuries and deaths. Such efforts might focus on improvements in
roadway design aimed at reducing many of the serious injuries involving collisions with
automobiles every year. The development of bike paths (i.e., paths that separate bicycles from
parallel motor vehicle traffic) and bike lanes (i.e., designated lanes on roadways which are offlimits to motor vehicles) should also be considered.
Efforts to improve the bicycle riding environment are already underway at all levels of
government. As mentioned above, the DOT's 1991 Appropriations Act instructed DOT to
develop a plan to promote bicycling and walking, and to enhance the safety of these
transportation modes. The goals of the plan are to double the percentage of trips made by
bicycling and walking by the year 2000, and to simultaneously reduce by 10 percent the
number of bicyclists and pedestrians killed or injured in traffic crashes (Federal Highway
Administration, 1994). DOT hopes to do this by, among other things, promoting the use of
federal funds for the development of a bicycle-friendly infrastructure (i.e., riding surfaces,
lighting at night, and facilities), and for education and training. The Intermodal Surface
Transportation Efficiency Act (ISTEA) also promotes improvements in the riding environment.
ISTEA requires that all state and local governments incorporate programs and facilities for
bicyclists in transportation plans. ISTEA also requires states to establish and fund bicycle and
pedestrian coordinator positions for promoting and facilitating the increased use of nonmotorized modes of transportation.
The higher injury risk on roadways also suggests that motorists and bicyclists need to
be educated in bicycle safety. Motorists need to be aware of the many road hazards that
confront bicyclists, to help them avoid collisions when approaching bicyclists on the road.
Being aware of road hazards confronting bicyclists can also help them better assess high risk
areas, such as intersections, and be more attentive in areas where bicyclists may not be clearly
in view. Safety programs geared toward adult bicyclists who ride in traffic, such as the League
of American Bicyclist's hands-on training program "Effective Cycling," should also be
encouraged.
Behavioral Factors
The bicycle study found that many of the bicycle-related injuries and deaths every year
are related to what the rider does and how the rider interacts with environmental factors.
Riding practices that are risky, that reflect poor riding judgment, or that fail to account for
environmental conditions, play a major role in injury and fatality scenarios. This finding
suggests that information and education (I&E) might play a role in injury reduction.
Many groups and organizations, including the CPSC, actively promote bicycle safety
through informational efforts. The promotion of bicycle safety through public service
15

announcements, brochures, poster campaigns, and other means must continue. These
messages reach new audiences and reinforce safety behavior. However, I&E efforts,
particularly those which are short term or do not present new information to consumers, may
have limited additional impact on rider behavior. Moreover, information by itself is unlikely to
change the behavior of children.
One of the most striking findings of the study is the higher risk of injury for children.
About 71 percent of the emergency room treated injuries, and 37 percent of the deaths
involved children under age 15. In addition, when other factors are held statistically constant,
the expected injury risk for a child under age 15 is over 5 times the risk for an older rider.
A clear implication is that there is a potentially big injury reduction payoff that may be
gained by focusing on the behavior of the highest risk population, children. One remedy is to
train children in safe riding practices. Child training programs need to be developed
judiciously. From a review of the available bicycle training literature, HF finds a consensus
among child development experts that many safety concepts cannot be learned by children
before a certain maturational level, regardless of the amount of training. In large part, this is
because of children's physical and cognitive limitations in dealing with a complex and
constantly changing riding environment. Determining the time appropriate to begin bicycle
safety education is therefore essential in designing effective programs.
Existing behavioral studies find that the optimal time for intensive bicycle safety
education for children is between the third and sixth grades (i.e., riders 9 to 12 years of age).
(See references in Part IV.) This does not mean that younger children should not have any type
of training, but that a comprehensive program is most effective beginning in the third or fourth
grades, with refresher courses for older children and adults. By the sixth grade, most children
have the ability to understand and perform the taught behaviors.
The analysis of risk and hazard patterns reveals several areas that should be stressed in
training programs for children. Helmet use should be encouraged to reduce the incidence of
head injury, which was especially high for children. Roadway skills should be emphasized, as
indicated by the substantially higher risks on streets. The higher risks during non-daylight
hours indicate that night riding by children should be discouraged. Training courses should
also include some basic information on how to maintain bicycles in good working order.
It would also be useful to convey child safety information to parents who, if they were
aware of risks, might encourage safer riding habits, such as the use of helmets. Given
children's risk patterns and available human factors information on the cognitive and physical
development of children, parents might want to discourage or prohibit children under the age
of about 10 from riding on roadways (without direct parental supervision) or from riding at all
during non-daylight hours.

16

Other Implications: Safety Equipment
The importance of the behavioral and environmental factors in hazard patterns also has
implications for the use of protective safety equipment, which can prevent or mitigate injuries
when accidents occur. Encouraging children to use safety equipment, such as helmets, is
especially important because of the difficulty in teaching young children certain safety skills.
Head injuries represent the most serious and potentially life threatening injuries that can
be sustained by bicyclists. According to the injury survey results, almost one-third of hospital
emergency room treated injuries involve the head, and children under age 10 are significantly
more likely than older riders to suffer head injuries. In addition, Sacks et al. (1991) estimate
that about 62 percent of all U.S. bicycle-related deaths involve injuries to the head. Based on
these estimates, the societal costs associated with the bicycle-related injuries and deaths
involving head injury amounted to more than $3 billion in 1991.
Available evidence indicates that helmets reduce both the likelihood and severity of
head injury (Dorsch et al, 1987; Thompson et al., 1989). Results from the exposure survey,
however, indicate that only about 17.6 percent of bicyclists currently wear helmets. This is
higher than the 5 to 10 percent usage rate estimated in studies conducted only a few years ago
(see references at Part II), and suggests that attitudes towards helmet use are improving.
Nevertheless, helmet usage rates remain low. Increasing helmet use may therefore be the
single most important factor in reducing the incidence of serious bicycle injuries.
The high incidence of fatal accidents after dark also suggests night riding is an area for
future safety efforts. People who ride at night should be aware of the need to see and be seen.
This suggests that the use of bicycle headlights and reflective clothing should be encouraged.
Night riders should also make sure that their bicycles are equipped with reflectors, as required
by the CPSC bicycle standard.

17

References
Bicycle Institute of America. Bicycling Reference Book: Transportation issue. Washington,
DC: Author; 1993.
Cross, K. D., and Fisher, G. A study of bicycle/motor-vehicle accidents: Identification of
problem types and countermeasure approaches (Technical Report DOT-HS-803 315).
Washington, DC: National Highway Traffic Safety Administration; 1977.
Dorsch, Margaret M.; Woodward, Alistair J.; Somers, Ronald L. Do bicycle safety helmets
reduce severity of head injury in real crashes? Accident Analysis and Prevention 19(3): 183190; 1987.
Federal Highway Administration. National Bicycling and Walking Study. (Technical Report
FHWA-PD-94-023), Washington, DC: Author; 1994.
National Sporting Goods Association. Sports participation in 1991. Mt. Prospect, Ill.:
Author; 1992.
Rodale Press. The cycling consumer of the 90's: A comprehensive report on the U.S. adult
cycling market. Emmaus, PA: Author; 1991.
Roland, H.E., Hunter, W.W., Stewart, J.R., and Campbell, B.J. Investigation of motor
vehicle/bicycle collision parameters (Technical Report DOT-HS-804 840). Washington, DC:
National Highway Traffic Safety Administration; 1979.
Sacks, J.J.; Holmgreen, P.; Smith, S.M.; and Sosin, D.M. Bicycle-associated head injuries and
deaths in the United States from 1984 through 1988: How many are preventable. Journal of
the American Medical Association 266(21): 3016-3033; December 4, 1991.
Thompson, Robert S; Rivara, Frederick P.; Thompson, Diane C. A case control study of the
effectiveness of bicycle safety helmets. The New England Journal of Medicine 320(21): 13611367; May 25, 1989.

18

Part II.
Bicycle and Bicycle Helmet Use Patterns in the United
States: A Description and Analysis of National Survey
Data
____________________
Gregory B. Rodgers, Ph.D.
Directorate for Economic Analysis
November 1992

Introduction
Bicycle riding is an important means of transportation, as well as one of the most
popular recreational activities in the United States (National Sporting Goods Association,
1992). This popularity is accompanied by a large number of injuries and deaths every year.
Based on data from the U.S. Consumer Product Safety Commission's (CPSC) National
Electronic Injury Surveillance System (NEISS), a stratified random sample of U.S. hospital
emergency rooms (CPSC, 1988), there are more than an estimated 500,000 nonfatal bicyclerelated injuries treated in the nation's hospital emergency rooms annually. When other
medically-attended injuries are counted, such as injuries treated in physicians' offices, there may
be on the order of about one million medically-attended injuries involving bicycles every year
(J. Robb Associates, 1976). In addition, based on information from the National Safety
Council (1992), there are almost 1,000 bicycle-related deaths annually.
The societal costs of bicycle-related injuries and deaths are large. Based on the
CPSC's Injury Cost Model (Technology & Economics, 1980), the costs of the medicallyattended injuries amount to about $6 billion annually. In addition, based on an imputed cost of
$2 million per life lost, fatalities add $2 billion annually. The total estimated societal costs of
bicycle-related injuries and deaths may therefore be about $8 billion annually.
In spite of the large number of injuries and deaths, there has never been a
comprehensive national survey designed to gather information on the characteristics and use
patterns of the general population of bicyclists.1 The published literature on bicycle hazards
consists primarily of injury analyses, most of which have been carried out at the level of the

1

Rodale Press recently conducted a major survey of adult bicycle riders in the United
States (Rodale Press, 1991). However, the Rodale Press survey was limited to bicycle
riders age 18 and older who had acquired new bicycles, and accounts for only about 60
percent of all U.S. bicycle riders.
19

individual hospital or in limited geographical areas. Several recent studies have also attempted
to measure the effectiveness of helmets in reducing head injuries.
Injury studies provide valuable information about injury characteristics and scenarios.
However, in the absence of control (or "exposure") data describing the characteristics and use
patterns of the rider population, injury studies are not enough to allow us to quantify the injury
and fatality risks associated with bicycle use (Dewer, 1978; HDR Engineering, 1991).2
This report presents the results of a comprehensive, nationwide 1991 survey of U.S.
bicycle riders (the "exposure survey") conducted by the U.S. Consumer Product Safety
Commission. It provides information on: the number of riders and bicycles in use; the
demographic characteristics of rider households; rider characteristics and use patterns; helmet
use patterns; and the types of bicycles in use.
The report also presents an analysis of the factors associated with helmet use. These
factors are determined and quantified with a probit regression model, a qualitative response
model that can be used to estimate helmet use probabilities for individual bicyclists and for
various population subgroups.

Survey Methodology
Abt Associates, Inc. ("Abt"), a survey firm located in Cambridge, Massachusetts,
designed for the CPSC a telephone survey to provide a national probability sample of
households with bicycle riders in the 48 contiguous states and the District of Columbia. The
survey used the Mitofsky-Waksberg method of random-digit-dialing (Waksberg, 1978), a
two-stage sampling procedure intended to give all telephone numbers in the continental U.S.
an equal probability of selection. The survey's initial goal was to complete about 1,150
interviews with bicycle riders from around the nation. A detailed description of the sampling
procedure is provided in the appendix.

2

Consider an example. About 70 percent of nonfatal injuries treated in hospital
emergency rooms involve children under the age of 15 (Tinsworth, 1987). However, there
are no nationwide data describing the riding patterns and behaviors of these children, or the
amount of riding they engage in. This makes it difficult to determine whether the large
proportion of injuries suffered by children results from high levels of exposure (i.e.,
aggregate riding times), risky riding patterns, or limitations in motor or cognitive skills.
Exposure information, as well as injury information, is needed to evaluate these risks, to
determine the relative importance of the various hazard patterns, and, ultimately, to develop
effective intervention strategies to reduce injuries.
20

The survey questionnaire was designed by CPSC staff, in consultation with Abt and
interested user and industry groups.3 Although no major problems were found in a pretest of
the questionnaire, responses in the pretest resulted in revisions and refinements to several of
the questions.
The survey was conducted during June and July 1991 (Abt, 1991). When households
were reached, respondents were asked the number of bicycle riders (i.e, those who rode a
bicycle at least once during the year prior to the survey) in the household. If there was more
than one rider in the household, one was selected at random to be interviewed. If the selected
rider was a young child, an adult in the household was asked to respond on the child's behalf.
In total, 6,076 residential numbers were called. The disposition of telephone calls to
residential numbers is described in Table 1. (Tables begin on page 41.) A maximum of six
attempts were made to obtain an answered call for each sampled telephone number. The 1,009
telephone numbers which were busy or for which there were no answers were presumed to be
residential numbers, although some are likely to have been nonresidential.
Of the residential numbers called, 4,346 households were successfully screened to
determine whether or not a household member qualified as a respondent. The screening
resulted in 1,254 completed interviews with bicycle riders (or designated respondents for
young children). The remainder was comprised of 2,613 households that owned no bicycles,
and 479 screenings with households that owned bicycles that had not been ridden during the
previous year.
The response rate can be measured in several ways. Since 4,346 screenings were
completed (the sum of rows 1 to 3 of Table 1), the response rate was 92 percent of the 4,705
cases (rows 1 to 5) in which contact was made with the appropriate respondent and 86 percent
of the 5,067 cases (rows 1 to 6) in which contact was made with a household member but not
necessarily the respondent. Finally, when the 1,009 cases (row 7) in which the telephone rang
busy or there was no answer on all attempts are included, the minimum response rate was 71
percent. The 1,254 interviews exceeded the 1,150 target number because the minimum
response rate was higher than expected.
After the survey data were collected, the sample was weighted to make population
projections of bicycle use in the continental U.S. In order to make the projections reflect the
estimated 94 million U.S. households in 1991, each of the successfully screened sample
households received a weight of 21,629.1 (i.e., 94 million/4,346). That is, each of the
successfully screened households was assumed to represent 21,629.1 U.S. households. In
addition, since only one rider per household was interviewed, the household weight for each of
3

The groups included the Bicycle Federation of America, the Bicycle Helmet Safety
Institute, the Bicycle Manufacturers of America, the National Highway Traffic Safety
Administration, and the National Safe Kids Campaign.
21

the 1,254 sample households containing one or more bicycle riders was multiplied by the
number of bicycle riders in the household. This yields a "rider" population weight reflecting
the total number of bicycle riders in the U.S. (Kish, 1965).
Information on the age and gender of all bicycle riders in the 1,254 sample households
was also gathered in the survey. This enabled further refinement of the weighting process to
account for the apparent over or under-representation of some of the age-gender categories
interviewed. For example, while male riders under age 10 accounted for about 9.3 percent of
those interviewed, they accounted for about 11.9 percent of the total number of household
riders. The "rider" population weight for these riders was therefore adjusted by the ratio of
1.28 (i.e, 0.119/0.093) to account for the apparent under-representation of riders in this
category. The ratio adjustment factors for eight age categories and two gender categories,
ranged from 0.79 (25-34 year-old females) to 1.40 (10-14 year-old females).
The survey results are subject to some nonsampling errors (Abt, 1991). First, the
survey excluded households that do not own telephones, about six percent of all use
households. In addition, Alaska and Hawaii were excluded from the survey. However, these
states only account for about 0.6 percent of total households in the U.S.
Rider Characteristics and Use Patterns
Population Estimates and Household Demographics
Based on the survey results, there were an estimated 66.9 million bicycle riders
residing in about 27.1 million households in 1991. Thus, there were riders in an estimated 28.8
percent (27.1 million/94 million) of all U.S. households. These riders used an estimated 65.9
million bicycles during the year.4 This indicates that the vast majority of bicyclists ride a
specific bicycle not shared with other household members.
The survey also gathered information on the number of bicycles owned by households,
and whether or not the bicycles had been ridden during the previous year. In total, there were
an estimated 96.0 million bicycles in about 37.4 million bicycle-owning households. Thus,
about 40 percent of the 94 million U.S. households have one or more bikes, but about 31
percent of the 96 million bicycles in these households had not been used during the past year.
Table 2 summarizes data on riders and bicycle ownership, per household, and calls
attention to the large proportion of households with multiple riders and bicycles. Over 70
percent of households with riders have more than one rider, and about 23 percent have four or
4

This estimate is considerably less than the National Safety Council's estimate of 105
million bicycles in use in 1991 (National Safety Council, 1992, p. 65). The National Safety
Council's estimate was based on a ten-year total of domestic production plus imports less
exports.
22

more riders; similarly, about 23 percent of bicycle-owning households had four or more
bicycles. On average, there are about 2.5 riders per household with riders, and about 3.4
bicycles per bicycle-owning household.5
Table 3 compares rider households with census data on all U.S. households. There are
no major regional differences in the location of rider and U.S. households. On the other hand,
though the population density figures from the survey and the 1990 census are not directly
comparable, it appears that greater proportions of rider households are located in low density
areas. About 57 percent of rider households live in a "small city or town" or "open country or
farm," compared with 32 percent of all households which are in non-Metropolitan Statistical
Areas (MSA) and MSAs with a population of less than 0.25 million. Only about 20 percent of
rider households are from a "large city or suburbs," compared with 31 percent of households in
MSAs with a population of 2.5 million or more.
Rider households are larger than the U.S. norm, reflecting the large number of children
who ride bicycles. Two-thirds of all rider households have four or more members, compared
with only about 26 percent for all U.S. households. In contrast, single-person households
account for 24.5 percent of U.S. households, but only 3.1 percent of rider households.
Rider households also have higher education levels and incomes than the U.S. norm.
Almost 50 percent of rider households have at least one college graduate, compared to 23
percent of all U.S. households. In addition, while the median U.S. income was $30,000 in
1990, the median income was about $40,000 or more in rider households. The higher income
for rider households reflects the larger average household size and the higher education levels.
Characteristics of Riders
Characteristics of the rider population are shown in Table 4. Rider ages varied widely,
from 2 to 77 years, but 25.2 percent were under age 11 and about half (49.9 percent) were
under age 21. Only about 6 percent were over age 50. In addition, just over half of all riders
(52.3 percent) were male.
About 81 percent of bicyclists learned to ride at least four years prior to the survey, and
just over half (52.4 percent) learned to ride more than 10 years prior to the survey. As in other
recreational activities, the number of years since a bicyclist learned to ride is a measure of
riding "experience," and hence riding skills (see, e.g., Rodgers, 1990). However, this variable
is also highly correlated with rider age (r = 0.76, p < 0.01) and suggests (not surprisingly) that
most individuals learn to ride bicycles as children. Consequently, this particular measure of

5

For households which own bicycles but have no riders there are an average of 1.9
bicycles per household.
23

"experience" may be a weak predictor of bicycle riding skills, especially for adults, since there
may have been long intervals in which bicycles were not used.
Bicycle Use Patterns
Since the injury risk is affected by the ways in which bicycles are used (Dewer, 1978),
substantial amounts of information were gathered on rider use patterns. The amount of time
spent riding a bicycle, a measure of rider exposure to risk, was estimated from a series of
questions intended to determine (1) the number of months bicycles were used in the previous
year, and (2) the number of hours individuals spend riding in an average month.6
According to the results shown in Table 5, the estimated mean and median annual
riding times for bicyclists are 236 and 105 hours per year. These estimates imply an aggregate
of about 15 billion hours of bicycle riding annually in the continental U.S. However, riding
times vary substantially from individual to individual. Over 20 percent ride less than 25 hours
per year, and about 12 percent ride more than 400 hours per year.
Table 5 also provides information on riding times by age category. Annual riding times
are highest for the youngest riders, and generally decrease with age. Children under age 11
ride for an average of about 318 hours per year (about 35 percent more than the average of
236 for all riders) and are followed by 11-to-14 year-olds with an average of about 262 hours
per year (about 11 percent more than average). Because of these higher averages, younger
bicyclists account for a disproportionate amount of riding time. Riders under age 11 account
for about 33.9 percent of all riding time, and riders under age 15 account for about 49.8
percent.
Table 6 provides information on the relative amount of riding time spent in various
environments or on various riding surfaces. Such information is important in analyzing risk
patterns since different environments are likely to have varying impacts on the injury risk.
Relative frequencies were quantified by means of the following discrete categories:7

6

The Rodale Press survey estimated bicycle use in terms of distance (i.e., miles in an
average warm weather month). However, following discussions with industry and user
groups, it was concluded, since bicycle riding is primarily a recreational activity, that hours
of rider use is a better measure of exposure, especially for children. It was also believed
that bicyclists are able to estimate hours of use more accurately than miles ridden.
7

Since frequency responses were requested for multiple surface types, responses from
some riders were not internally consistent. Some riders, for example, indicated that they rode
"more than half of the time" on more than one surface type, a logical impossibility. The
responses are nevertheless quite instructive of basic riding patterns since they provide an
approximate ranking of surface types.
24

- (1) always or almost always;
- (2) more than half of the time;
- (3) less than half of the time;
- (4) never or almost never.
Table 6 presents two measures of the "frequency" associated with the various use
patterns. The top line for each category provides the percentage of riders at each frequency
response; the second line (numbers in parentheses) adjusts rider responses for estimated annual
riding times. For example, the 18.4 percent of riders who "always or almost always" ride on
sidewalks or playgrounds account for 19.3 percent of total riding time.
The predominant riding surface is neighborhood streets. Over 60 percent of bicyclists
ride primarily (i.e., spend all or most of their riding time) on neighborhood streets with low
traffic volume.8 Almost 30 percent (mostly children) ride primarily on sidewalks and
playgrounds.
Only about 6.8 percent ride primarily on major thoroughfares and highways with high
traffic volume and 16.9 percent ride primarily on bike paths.9 Finally, about 17.6 percent ride
primarily on unpaved roads and about 10.5 percent ride primarily on other types of unpaved
surfaces or trails. Adjustments for riding time do not substantially alter these estimates, but
they do suggest that riders who spend all or most of their time on major thoroughfares,
highways, or unpaved surfaces tend to ride more than the average.
Table 6 also indicates the relative time spent in several other activities or practices.
About 17.6 percent of bicyclists wear helmets all or most of the time and account for about
20.6 percent of aggregate riding times. Another 6.0 percent said they wear helmets some (i.e.,
less than half) of the time. Helmet use is discussed more in the next section of this report.
About 8.7 percent of riders (representing a projected 5.8 million bicyclists) spend all or
most of their riding time commuting to work or school and account for about 12.6 percent of
total annual riding times. Another 10.1 percent use their bicycles for commuting less than half
of the time.
A relatively small proportion of bicycle riding takes place after dark. About 12.4
percent of bicyclists indicated that they ride at least occasionally after dark, but only 3.1
percent of the bicyclists ride primarily after dark. Despite the relatively small amount of

8

This response was so pervasive that it may include some bicyclists who frequently ride
on neighborhood streets to get to other surfaces such as bike paths.
9

However, the respective proportions were higher for 21-30 year-olds; 12.8 percent of
these bicyclists report that they ride primarily on major thoroughfares and highways and 24.1
percent ride on bike paths.
25

nighttime riding, available studies suggest that nighttime accidents account for a large share of
bicycle-related injuries and deaths (Cross, 1977; Ferguson and Blampied, 1991; NHTSA,
1993; Tinsworth, 1987).10 Lights are considered important safety equipment when riding after
dark. However, of those who ride at least occasionally after dark, less than one-third use
headlights or tail lights.
About 2.8 of riders wear earphones all or most of the time; another 4.5 percent do so at
least occasionally. Finally, about 2.3 percent of respondents said that when riding they carry
infants all or most of the time. Another 1.7 percent indicated they do so at least occasionally.
However, these are generally infrequent riders and account for only about 1.5 percent of total
riding time. In addition, about 56 percent of those who carry young children reported that the
child always wears a helmet, and another 5 percent reported that the child wears a helmet most
of the time.
Characteristics of Bicycles In Use
Table 7 presents information about the bicycles used most frequently by the
respondents. (See Figure 1 for pictorial representations of the various bicycle types.)
Lightweight racing or touring bicycles are the most common, with about 34.5 percent of the
total. Only 14.1 percent of riders use BMX or high rise bicycles, but, because BMX and high
rise bicycles are used largely by children who ride more than an average amount of time, these
riders account for 24.5 percent of total bicycle use.
Mountain, city, or all-terrain bicycles, which were first marketed on a large scale in the
mid-1980s, have become increasingly popular with recreational riders in recent years.
Although they account for only about 17 percent of all bicycles in use, they account for about
25 percent of the newer bicycles purchased within a year of the survey. On the other hand,
lightweight racing and touring bicycles are becoming correspondingly less popular: while they
account for 34.5 percent of all bicycles in use, they account for only 26.6 percent of those
acquired in the year prior to the survey.11

10

According to NHTSA (1993), about 23.5 percent of the bicyclist fatalities involving
motor vehicles in 1991 occurred between the hours of 9:00 PM and 5:59 AM.
11

Market share estimates of the bicycle types purchased during the year preceding the
survey (shown in Table 7) are generally consistent with 1990 domestic sales estimates
provided by the Bicycle Manufacturers Association (BMA, 1991). BMA sales estimates
indicate that about 20 percent of 1990 domestic sales were children's sidewalk models, 26
percent were BMX and high rise models, 14 percent were lightweight racing and touring
models, and about 40 percent were in the mountain, all-terrain, and city bicycle or
middleweight/cruiser categories.
26

Table 7 also describes the types of bicycles used by various age groups. Not
surprisingly, most children under age 11 use children's sidewalk bicycles (38.2 percent), or
BMX or high rise bicycles (31.1 percent). Lightweight racing or touring bicycles are the most
commonly used bicycle for riders over age 10, accounting for over 40 percent of the total.
Mountain, city, or all-terrain bicycles appear to be most popular with 21-to-30 year-old riders,
with about 27.6 percent of the total, but they are also used by 19.1 percent of 11-to-20 yearold riders. Middleweight and cruisers are most popular with older riders. Although not shown
in the table, 34.2 percent of riders over the age of 50 reported that they used a middleweight or
cruiser.
The bicycles in use tend to be relatively new. About 28.5 percent were acquired during
the year preceding the survey, and another 54.2 percent were acquired from one to five years
before the survey was conducted. The mean and median length of time since the bicycles were
acquired were 3.6 and 2.0 years. In addition, about 19 percent were acquired used, indicating
a substantial aftermarket for bicycles.
Few (less than 3 percent) of the bicycles had been substantially modified since
acquisition. Reported modifications included changes to the handlebars (1.0 percent), wheels
(0.6 percent), and seats (1.1 percent). Tail lights and headlights which, according to Ferguson
and Blampied (1991), can substantially reduce the nighttime accident risk, were the most
widely reported bicycle safety accessories. Tail lights and headlights were respectively
reported on 20.6 percent and 14.5 percent of bicycles.
Helmet Use Patterns
Sacks et al. (1991) estimate that 62 percent of all U.S. bicycle-related deaths and 32
percent of bicycle-related injuries treated in hospital emergency rooms involve head injuries.12
Recent studies reveal substantial safety benefits from helmet use: Dorsch et al. (1987) showed
that helmets substantially reduce the severity of head injury when head injuries occur;
Thompson et al. (1989) found that helmets can reduce the likelihood of head injury by 75 to 85
percent. Helmet usage rates have nevertheless been found to be generally low. Although no
firm nationwide data were available prior to this survey, estimates generally put helmet use at
under 10 percent for all riders (Wasserman et al., 1988; Weiss, 1990). Moreover, studies of
specific localities found that less than 5 percent of school age children wore helmets
(DiGuiseppi, 1989; Howland, 1989; Weiss, 1986 and 1992). This section discusses the bicycle
survey findings regarding helmet use patterns and presents the results of an analysis to
determine the factors that go into the decision to use a bicycle helmet.

12

These estimates are based on data from the Center for Health Statistics and the CPSC's
National Electronic Injury Surveillance System.
27

Description of Survey Results
As shown in Table 8, about 27.3 percent of riders (representing a projected 18.3
million bicyclists) own or have the use of a helmet. About 77.9 percent of the helmets are
"hard" shell (i.e., a polystyrene shell covered with a hard plastic covering), 14.1 percent are
soft shell (i.e., a lightweight polystyrene shell with no plastic covering), and 5.1 percent are
"thin" shell (i.e., polystyrene with a light or thin plastic covering).13 However, the lighter
weight soft and thin shell helmets are becoming increasingly popular. Over 30 percent of
helmets purchased or received during the year prior to the survey were the soft or thin shell
types, compared to about 14 percent of those purchased or received three or more years ago.
Of the riders who have the use of helmets (27.3 percent of all riders), 64.2 percent
wear them all or most of the time and 21.8 percent wear them less than half of the time; 13.5
percent of riders who have helmets never use them. These figures indicate that about 17.6
percent of all riders (11.8 million bicyclists) wear helmets all or most of the time, 6.0 percent
(4.0 million) wear helmets less than half of the time, and about 76.0 percent (50.9 million)
never (or almost never) wear helmets.
Helmet use is highest for the 41-to-50 year-old age group, with a reported 24.6 percent
usage rate (i.e., percent reporting that they wear helmets all or most of the time), and lowest
for the 11-to-14 year-old age group with a 11.4 percent usage rate. The usage rate for
children under the age of 11, which was usually reported by parents who were responding for
their young children, was 17.0 percent.
These estimates, in comparison to estimates of helmet use in earlier studies, suggest
that helmet use for all riders has increased from under 10 percent to almost 18 percent in the
last few years. The survey finding that about 52.7 percent of helmet wearers began wearing
helmets in the last two years supports the conclusion that the change is real and recent.
Information was also gathered on the reasons why individuals use or do not use
helmets, as shown in Table 9. For the approximately 17.7 percent of riders who purchased
helmets for their own use, as opposed to receiving one as a gift, comfort and safety
considerations were very important in the purchase decision. Two additional comfort factors,
the weight of the helmet and ventilation, were also described as "very important" by over 40
percent of purchasers. Cost and appearance were apparently secondary considerations for
those who did buy helmets, but were still reported to be at least "somewhat important" by over
55 percent of purchasers.
Of the 13.4 percent of riders who always wear helmets, (see Table 8), nearly all (97.8
percent) described "safety" as an important reason for doing so. The "insistence of family

13

The type of helmet was unknown by about 2.9 percent of respondents.
28

members," was reported to be important by about 56 percent. In addition, local legal
requirements were mentioned by about 13.5 percent of these riders.
When riders who sometimes, but not always, wear helmets (i.e., the 10.2 percent of
riders who wear helmets "more than half of the time," or "less than half of the time") were
asked to describe the circumstances under which they "usually" wear a helmet, 40.0 percent
indicated "when riding in traffic" and 25.2 percent indicated "when on long rides." About 17.9
percent usually wear helmets when reminded to do so.
These riders were also asked when they do not wear a helmet. The most frequent
responses were when riding a short distance (31.6 percent), when not riding in traffic (23.8
percent), and when they forget (22.9 percent).
Finally, the estimated 76.0 percent of riders (see, Table 6) who said they never or
almost never wear helmets were asked why. About 21.6 percent said that they had never
thought about it. While 15.6 percent said they had not gotten around to wearing a helmet (and
thereby implied a positive attitude toward helmet use), a large proportion also indicated a lack
of need for helmet use: 21.0 percent said that helmets were unnecessary and 18.8 percent said
they did not wear helmets because they seldom ride in traffic. Smaller percentages said that
helmets were not comfortable (8.9 percent), not attractive (4.9 percent), and too expensive
(7.3 percent).
Statistical Analysis of Helmet Use
The Model. This section develops a probit regression model to determine and quantify
the factors associated with helmet use. Probit analysis, like multiple regression analysis, is a
statistical procedure in which variation in the dependent variable is explained by variation in the
explanatory variables. The probit specification of the regression model is used to examine the
relationship between a series of explanatory variables and a dependent variable that represents
two (or more) distinct alternatives (Pindyck and Rubinfeld, 1991).
In this analysis the dependent variable represents whether or not riders use helmets.
Survey respondents were assumed to be helmet users if they reported that they wore a helmet
all or most of the time. There could be some upward bias in the reported helmet use rates, as
has been described in some automobile seat belt use studies (Knapper et al., 1976; Hakkert et
al., 1981). However, since the extent of bias, if any, is unknown, it will be assumed that
reported usage rates provide a reasonable approximation to actual helmet use patterns.
The explanatory variables comprise various rider characteristics, use patterns, and
household demographic factors that may influence the bicyclist's decision to wear a helmet.
About 22 percent of the observations were lost because of missing information on the
independent variables. A sensitivity analysis, conducted by replacing missing values with the
mean value of the variable in question (Pindyck and Rubinfeld, 1991), indicated that the
29

models were not substantially affected by the missing data. Table 10 defines the explanatory
variables used in the analysis.
Statistical Results. Table 11 shows the results of three specifications of the regression
model. These specifications differ by the way in which the age and riding time variables are
entered into the models. Rider age is included as a series of "dummy" variables (i.e., AGE(110) to AGE(41-50)) in Models 1 and 2. These variables are intended to pick up the
relationship between the various discrete age categories and the likelihood of helmet use,
relative to riders over the age of 50. Model 3, in contrast, includes age as a continuous
variable (AGE). In addition, Model 1 expresses riding time (i.e., hours of exposure to risk) as
the natural logarithm of the estimated annual hours of use (LN(HOURS)).14 Models 2 and 3,
on the other hand, include riding time as part of an interaction term defined as the product of
the natural log of riding time and rider age (AGE@LN(HOURS)).
All of the equations are statistically significant. In addition, inclusion of the interaction
term improved somewhat the fit of Models 2 and 3, relative to Model 1, as is indicated by the
higher model chi-square and score statistics.
The regression results show several strong relationships between helmet use and the
surface types over which bicyclists ride. These relationships are measured with a series of
dummy variables representing various riding surface types, relative to unpaved and other
surfaces. Helmet use is higher for riders who spend all or most of their riding time on bike
paths (BIKEPATH) and on major thoroughfares, highways, or streets with high traffic volume
(HIGHWAY). In contrast, helmet use is lower for riders who spend all or most of their riding
time on neighborhood streets with low traffic volume (STREET). Helmet use on playgrounds
or sidewalks (SIDEWALK) is not significantly different from use on unpaved and other
surfaces, but it is significantly lower than use on bike paths and major thoroughfares.
There is also a strong positive relationship between riding time and helmet use -helmet use increases with riding time. This relationship is evidenced clearly in Model 1, where
riding time is entered as a natural logarithm (LN(HOURS)). It is also indicated in Models 2
and 3, where riding time is entered as part of the interaction term with age
(AGE@LN(HOURS)). However, in contrast to the specification of the riding time variable in
Model 1, the coefficients for the interaction terms in Models 2 and 3 suggest that helmet use
increases with riding time at an increasing rate for older riders. That is, the change in the rate
of helmet use is more sensitive to changes in riding time for older riders.
These relationships provide some evidence that riders are more likely to wear helmets
if, by virtue of riding a lot or by riding frequently on major thoroughfares with high traffic
14

Transforming the riding time variable to a natural logarithm (as opposed to using it as a
continuous linear variable) increased its explanatory power by reducing the distorting effect
of outliers on the results.
30

volume, they face potentially higher accident rates.15 The relationships are also consistent with
some recent analyses of behavioral response in inherently risky activities, such as in automobile
driving. These studies indicate that consumers increase safety efforts (i.e., by wearing seat
belts) in response to greater perceived risk (Blomquist, 1988 and 1991; McCarthy, 1986).
Helmet use is only slightly related to specific rider characteristics. Rider experience
(LN(EXPER))and gender (GENDER), for example, have no independent statistical impact on
helmet use.16 Nor do the results of Model 1 indicate any measurable relationship between age
and helmet use. However, when age is allowed to interact with riding time, as in Models 2 and
3, the results suggest that helmet use is systematically related to age, though in a somewhat
complex way.
For riders over the age of 10, the Model 3 coefficients for the AGE and
AGE@LN(HOURS) variables indicate that helmet use increases with age for bicyclists who ride
more than about 20 hours per year (about three-quarters of bicyclists), and decreases with age
for those who ride less than 20 hours per year. More generally, the results suggest that
bicyclists who ride a lot of the time are more inclined to wear helmets as rider age increases; on
the other hand, infrequent bicyclists are less likely to wear helmets as age increases.
Notice also that Model 3 includes as a shape parameter a dummy age variable set equal
to one for riders 1-to-10 years of age (AGE(1-10)). The significant positive coefficient for this
variable indicates a higher likelihood of helmet use for these riders than can otherwise be
explained by the overall relationship between age and helmet use (as expressed by the
coefficients for the AGE and AGE@LN(HOURS) variables). The obvious implication, assuming
accurate responses to the helmet use questions, is that, on balance, young children tend to
wear helmets because their parents require it.17 This may suggest that the substantial recent
publicity in favor of helmet use has influenced the behavior of parents.

15

There was also some evidence that helmet use increases if the riders experienced an
accident requiring medical attention during the three years prior to the survey. However, this
relationship was not significant at the usual 5 percent significance level (p = 0.06, two-tailed
test).
16

A small proportion of bicyclists (about 2 percent), reported a greater number of years
of riding experience than their age. However, a sensitivity analysis, conducted by eliminating
the inconsistent observations from the analysis and by imposing a plausible replacement
scheme for the inconsistent observations, indicated that these inconsistencies did not affect
the results.
17

This does not mean that a majority of parents require their children to wear helmets, but
rather that enough do so that the helmet use patterns of children can be distinguished from
those of older bicyclists.
31

The regression results also show that helmet use is influenced by household
demographic factors, such as education and geographical location. Households headed by
members who attended college use helmets more frequently than households headed by
members with less education, as indicated by the positive and increasing coefficients for SCH2
and SCH3.18
There also appears to be some regional variation in helmet use. The regional variables
are included as a series of dummy variables and indicate regional differences in helmet use
relative to use in the Pacific Coast States. Helmets are used more frequently in the Pacific
Coast and Northeast States than in the Midwest, Southern, and Mountain States.
Predicted Helmet Usage Rates
Table 12 shows the estimated probability of helmet use for various combinations of
rider and bicycle characteristics, based on the econometric results from Model 3. Each of the
estimates is obtained by evaluating the probability of helmet use for a bicyclist who has five
years of riding experience and who resides in a large or medium sized city in a northeastern
state. Values for the other characteristics are specified in the table.
Part A shows the effect of age on the likelihood of helmet use, by selected hours of
annual riding time and riding terrain, for a male bicyclist. The first column shows helmet use
estimates for a male bicyclist who rides 10 hours per year on quiet residential streets, and
whose household members have no more than a high school education. The expected
likelihood of helmet use declines slightly with age from 3.9 percent for a 20 year-old rider to
3.4 percent for a 50 year-old rider. This illustrates the finding that helmet use declines with
age for relatively infrequent riders. The second column, in contrast, shows that if riding time is
300 hours per year, helmet use increases with age for riders over the age of 10 -- the expected
probability of helmet use increases from 6.8 percent for a 20 year-old rider to 13.0 percent for
a 50 year-old rider. The third column shows that the expected probability of helmet use
increases substantially for the bicyclist (of any age) if he rides 300 hours per year on a bike
path rather than on quiet residential streets. Finally, in all the columns of part A, the expected
probability of helmet use is higher for a 10 year-old rider than it is for a 20 year-old rider,
reflecting in part the impact of the dummy age variable (AGE(1-10)) on the risk estimates.
Part B shows the effect of hours of annual riding time on the likelihood of helmet use,
by selected combinations of education and riding terrain, for a 30 year-old female bicyclist.
The first column provides helmet use estimates for the bicyclist if she rides primarily on quiet
residential streets and if no household member has more than a high school education. If, for
example, she rides 200 hours per year the probability of helmet use is 7.5 percent. If, however,
18

Although income is not included in the model, helmet use also increases with household
income. Income was excluded from the model because it was highly correlated with the
schooling variables, and because data on household income were frequently missing from the
database.
32

she comes from a household with a college graduate, the probability of helmet use increases to
21.8 percent. Finally, if she rides primarily on highways or major thoroughfares with high
traffic volume, the probability of helmet use further increases to 45.6 percent.
Table 13 provides another view of the sensitivity of helmet use to discrete changes in
the independent variables by reporting the average predicted probability of helmet use for
various population subgroups. These estimates, which are also based on the econometric
results of Model 3, do not statistically hold other variables constant, but they do provide
consistent group estimates of the proportion of individuals who will choose to wear a helmet
(Train, 1986).
Predicted helmet usage rates are lowest for the 11-to-20 year-old age group, and,
except for the over 50 year-old age bracket, generally increase with age.19 Predicted helmet
usage rates are also higher for male bicyclists, who generally have greater annual riding times
than females.
Helmet use is higher for more frequent riders. Bicyclists who ride 100 hours per year
or more are twice as likely to wear helmets as bicyclists who ride less than 25 hours per year.
Helmet use also increases with household education and income. While helmet use appears to
increase gradually with income, households with college graduates are about three times as
likely to wear helmets as households with members having no higher than a high school
education.
Individuals from large or medium size cities are more likely to wear helmets, as are
individuals who ride all or most of the time on highways and major thoroughfares with high
traffic volume, or bike paths. In fact, bicyclists who ride primarily on highways and bike paths
are roughly twice as likely to wear helmets as other bicyclists. In contrast, helmet use is lower
for bicyclists who ride primarily on quiet residential streets, and somewhat lower for bicyclists
who ride primarily on sidewalks or playgrounds.
Finally, there are notable differences in the predicted helmet use in the various regions
of the country. Helmet use is highest in the Pacific Coast States (26.7 percent) and lowest in
the Mountain States (12.7 percent).
Summary and Discussion
This report described and evaluated the results of the 1991 CPSC bicycle exposure
survey. The survey was based on a "random-digit-dialing" sampling methodology. Its primary
goal was to collect statistically sound information on the characteristics of the general

19

Average usage rates are lower for bicyclists over the age of 50 primarily because their
annual riding times are substantially lower than they are for other age groups.
33

population of riders, their use patterns, and the types of bicycles they use, data which are
necessary to quantify risk and hazard patterns.
Survey results confirm the popularity of bicycle riding in
the U.S. There are an estimated 66.9 million bicyclists who reportedly spend about 15 billion
hours riding bicycles in a year. Although most bicycle riding is for recreational purposes,
bicycles are also widely used as a form of transportation. About 8.7 percent of riders, who
account for almost 13 percent of total riding times, use their bicycles primarily for commuting
to school or work.
Bicycles are used by riders of all ages, but young riders tend to predominate. About
one-fourth of bicyclists are under age 11, and about half are under age 21. Moreover, young
bicyclists ride more than the average for all bicyclists; when riding times are taken into
account, bicyclists under age 11 account for about one-third of all riding time, and those under
age 21 account for about 61 percent of all riding time.
Riding patterns and behaviors are closely tied to accident risk and must therefore be
considered when evaluating risk patterns. Most bicyclists said that they ride on quiet
neighborhood streets with low traffic volume, but sizable proportions also ride on sidewalks
and playgrounds, bike paths, and unpaved surfaces. On the other hand, a relatively small
proportion rides primarily on busy streets or major thoroughfares with high traffic volume, a
surface type which is associated with a higher likelihood of collisions with automobiles.
Similarly, relatively small proportions of bicyclists engage in potentially unsafe practices, such
as riding after dark, carrying infants, and wearing earphones.
While there are about 96.0 million bicycles available for use, about 65.9 million had
been used in the year prior to the survey. Lightweight racing or touring bicycles are the most
commonly used models by all riders over the age of 10. However, these models, which were
highly popular through the early-1980s, have been losing their relative sales share to the
mountain, city, or all-terrain models which have become increasingly popular in recent years.
Not surprisingly, age appears to be an important factor in the choice of bicycle types.
Children tend to ride sidewalk, BMX, or high rise bicycles, and the mountain/city/all-terrain
bicycles appear to be most popular with young adults in the 21-to-30 year-old age group.
Middleweight and cruisers are most popular with older riders.
The survey also obtained a substantial amount of information on helmet use patterns,
some of which was used to model the helmet use decision. The results of the probit regression
analysis indicate that helmet use is systematically related to riding patterns, household
demographic characteristics, and some personal rider characteristics. Riding time is a major
determinant of helmet use; riders who spend more time riding are more likely to wear helmets.
Other major determinants include the primary riding surface and demographic factors. Helmet
use tends to be higher for those who ride primarily on highways and bike paths, and lower for
34

those who ride primarily on neighborhood streets with low traffic volume and on sidewalks and
playgrounds. Helmet use also increases with household education levels.
The relationship between age and helmet use is complex, suggesting that helmet use
increases with age for frequent riders and declines with age for infrequent riders. There is also
evidence that children under age 11 are more likely to wear helmets, relative to older riders,
than can be explained by the general relationship between age and risk. This suggests,
assuming responses to the helmet use questions were accurate and unbiased, that parents are
requiring their young children to wear helmets more frequently than they had in the past.
The survey finding that about 17.6 percent of bicyclists wear helmets all or most of the
time (two- to three-times the usage rate found in studies conducted only a few years ago)
suggests that attitudes towards helmet use have been improving. The reasons for the change in
helmet use patterns are probably related to the increasing publicity given to the benefits of
helmet usage in the popular and scientific press in recent years. Improvements in helmet
construction (i.e., the development of soft- and thin-shelled helmets), that reduce the
discomfort of helmet usage by increasing helmet ventilation and by making helmets lighter and
more attractive, may also have played a role in increasing helmet use.20
In addition, the growing trend toward local helmet use laws, which have also been
widely publicized in the media, may also play a role (see, e.g., Beyers, 1992). According to the
National Safe Kids Campaign (1992), five states and seven local jurisdictions have enacted
some form of helmet requirements. All but one of the state requirements apply to child
passengers, and the local requirements apply generally to children who are operators or
passengers.21 For the most part, these requirements do not appear to be rigidly enforced, and
may therefore be viewed as strong informational warnings with minor penalties under some
circumstances. Nevertheless, a recent study of the effects of one locality's helmet requirements
for child operators suggests that they may have significantly increased helmet usage rates (Cote
et al., 1992).22

20

The long run increasing wealth of society (or segments of society) may also be an
underlying factor, since increased wealth is likely to increase the private demand for safety
(Viscusi, 1983).
21

The New Jersey state law applies to operators and passengers under the age of 14;
requirements in Chico, California and Rockland, New York, apply to all operators.
22

Cote et al. (1992) report, in an observational study, that child helmet usage rates
increased from about 4 percent to 47 percent in Howard County, Maryland, following the
institution of helmet requirements of children under the age of 16 in July 1990. This
compared with an increase from 8 to 19 percent over the same time period in Montgomery
County, Maryland, a county which had sponsored a community education program at about
the same time. Montgomery Country later enacted helmet requirements similar to those in
35

In spite of improvements in the overall helmet use rate, it remains low. Less than onefifth of all bicycle riders regularly wear helmets. The reported effectiveness of helmets in
reducing head injuries (see, e.g., Thompson et al., 1989 and Dorsch et al., 1987) suggests that
innovative informational and educational efforts designed to increase helmet use by riders of all
ages should be encouraged. Moreover, the survey results provide some reason to believe that
such efforts may be at least somewhat effective. Almost 40 percent of survey respondents who
do not own or wear helmets reported that they had never thought about doing so or that they
had simply not gotten around to buying a helmet. Many of these riders may respond to
information and education efforts that explain honestly the advantages of helmet use.

Appendix: Sampling Methodology
The Mitofsky-Waksberg method of random-digit-dialing (Waksberg, 1978) is a
two-stage sampling procedure intended to give all telephone numbers in the continental U.S.
an equal probability of selection. In the first stage, all of the nation's active area code/central
office telephone code combinations (called prefix codes) are stratified by the nine Census
Divisions. Prefix codes are randomly selected from each Census Division, and four-digit
random numbers are appended to the selections to generate complete telephone numbers. The
complete numbers are then dialed to determine which are residential, as opposed to
commercial.
The first stage residential numbers are used to generate a second-stage sample. The
first eight digits of these residential numbers are referred to as prefix areas in the sampling
literature, and each prefix area defines a cluster of 100 contiguous telephone numbers.23 At the
second stage, complete (10 digit) telephone numbers are randomly sampled from each prefix
area until a fixed number of residential numbers, referred to as the cluster size, have been
sampled from all sample prefix areas.
In the bicycle survey, 10 telephone numbers were initially sampled from each prefix
area. Nonresidential numbers were replaced until a total of 10 residential numbers were found
from the prefix area. This one-to-one replacement yields a self-weighting sample of residential
telephone numbers since the same number of residences are sampled from each prefix area.
Given the expected incidence of bicycle ownership, 682 residential telephone numbers
were generated in the first stage of the Mitofsky-Waksberg random-digit-dialing procedure to
provide the prefix areas. These prefix areas were randomly divided into 11 replicates, or

Howard County.
23

For example, the first-stage prefix area given by the number 301-504-09XX defines a
cluster of 100 telephone numbers ranging from 301-504-0900 to 301-504-0999.
36

subsamples, of 62 prefix areas residential numbers each. Each replicate can be viewed as
providing a miniature national sample of residential numbers. The second stage was
administered on a replicate-by-replicate basis to come as close as possible to completing the
desired 1,150 interviews.
Given the actual incidence of bicycle use found in the survey, only 10 replicates were
activated to generate the desired number of interviews. Thus, the first stage sample contained
620 prefix areas (i.e. 62 prefix areas per replicate times 10 replicates). In addition, since the
cluster size was set at 10 residences per prefix area, the second stage sample consisted of about
6,200 residential telephone numbers (i.e. 620 prefix areas times 10 residential numbers per
prefix area).
As described above, a total of 6,076 residential numbers were called. This is less than
the expected number of 6,200 because in a small number of prefix areas less than 10 residential
numbers were found when all 100 residential numbers were called. In addition, this total
excludes 4,343 calls to nonworking numbers and non-residential working numbers, which were
replaced as part of the sampling procedure.

37

References
Abt Associates, Inc. Final report: Bicycle exposure survey. Contract CPSC-C-89-2031, No.
008, Cambridge, MA: Author; November 1991.
Beyers, Dan. Helmet laws gain a track record. The Washington Post, p. B1; December 13,
1992.
Bicycle Manufacturers Association. The 1990 bicycle market in review (mimeo). Washington,
DC: Author; 1991.
Blomquist, Glenn. The regulation of motor vehicle and traffic safety. Boston: Kluwer; 1988.
Blomquist, Glenn C. Motorist use of safety equipment: Expected benefits or risk
incompetence? Journal of Risk and Uncertainty 4:135-152; 1991.
Cote, T.R.; Sacks, J.J.; Lambert-Huber, B.A.; Dannenberg, A.L.; Kresnow, M.; Lipsitz, C.M.;
Schmidt, E.R. Bicycle helmet use among Maryland children: Effect of legislation and
education. Pediatrics 89(6): 1216-1220; June 1992.
Cross, Kenneth D. Bicycle-safety education: Facts and issues.
Barbara, CA: Anacapa Sciences, Inc.; August 1978.

Technical Report. Santa

Dewer, Robert E. Bicycle riding practices: Implications for safety campaigns. Journal of
Safety Research 10(1): 35-42; Spring 1978.
DiGuiseppi, Carolyn G. Bicycle helmet use by children: Evaluation of a community-wide
campaign. Journal of the American Medical Association 262(16): 2256-2261; October 27,
1989.
Dorsch, Margaret M.; Woodward, Alistair J.; Somers, Ronald L. Do bicycle safety helmets
reduce severity of head injury in real crashes? Accident Analysis and Prevention 19(3): 183190; 1987.
Ferguson, Brett; Blampied, Neville M. Unenlightened: An unsuccessful attempt to promote
the use of cycle lights at night. Accident Analysis and Prevention 23(6): 561-571; 1991.
Hakkert, A.S.; Zaidel, D.M.; Sarelle, E. Patterns of safety belt usage following introduction of
a safety belt warning law. Accident Analysis and Prevention 13: 65-82; 1981.
HDR Engineering, Inc. National bicycling and walking study: Interim report. Report No.
FHWA-PD-92-003. Washington, DC: Federal Highway Administration; November 1991.

38

Howland, Jonathan et al. Barriers to bicycle helmet use among children. American Journal of
Diseases of Children 143: 741-744; June 1989.
J. Robb Associates. Impute measures of severity for NEISS data. Contract CPSC-P-76-681,
Houstan, TX: Author; 1976.
Kish, Leslie. Survey sampling. New York: John Wiley & Sons; 1965, pp. 398-401.
Knapper, C.K.; Cropley, A.J.; Moore, R.J. Attitudinal factors in the non-use of seat belts.
Accident Analysis and Prevention 8: 241-246; 1976.
McCarthy, Patrick S. Seat belt usage rates: A test of Peltzman's hypothesis. Accident
Analysis and Prevention 18(5): 425-438; 1986.
National Sporting Goods Association. Sports participation in 1991. Mt. Prospect, Ill.:
Author; 1992.
National Highway Traffic Safety Administration. Fatal accident reporting system 1991.
Washington, DC: Author; 1993.
National Safe Kids Campaign. Mandatory bike helmet legislation. Washington, DC: Author;
October 1992.
National Safety Council. Accident facts, 1992 edition. Itasca, IL: Author; 1992.
Pindyck, R.S.; Rubinfeld, D.L. Econometric models and economic forecasts. New York:
McGraw-Hill; 1991.
Rodale Press. The cycling consumer of the 90's: A comprehensive report on the U.S. adult
cycling market. Emmaus, PA: Author; 1991.
Rodgers, Gregory B. Evaluating product related hazards at the Consumer Product Safety
Commission: The case of all-terrain vehicles. Evaluation Review 14(1): 3-21; February 1990.
Sacks, Jeffery J.; Holmgreen, Patricia; Smith, Suzanne M.; Sosin, Daniel M. Bicycleassociated head injuries and deaths in the United States from 1984 through 1988: How many
are preventable. Journal of the American Medical Association 266(21): 3016-3033;
December 4, 1991.
Technology & Economics, Inc. The Consumer Product Safety Commission Injury Cost
Model. Contract CPSC-C-78-0091, Cambridge, MA: Author; July 1980.

39

Thompson, Robert S; Rivara, Frederick P.; Thompson, Diane C. A case control study of the
effectiveness of bicycle safety helmets. The New England Journal of Medicine 320(21): 13611367; May 25, 1989.
Tinsworth, Deborah K. Results of NEISS surveillance level study on bicycles. Technical
Report. Washington, DC: U.S. Consumer Product Safety Commission; March 16, 1987.
Train, K. Qualitative choice analysis: Theory, econometrics, and an application to
automobile demand. Cambridge, MA: The MIT Press; 1986.
U.S. Consumer Product Safety Commission. The NEISS sample: Design and implementation.
Technical Report. Washington, DC: Author; 1988.
Viscusi, W. Kip. Risk by choice. Cambridge, MA: Harvard University Press; 1983.
Waksberg, J. Sampling methods for random digit dialing. Journal of the American Statistical
Association 73(361): 40-46; 1978.
Wasserman, Richard C.; Waller, J.A.; Monty, Melanie J.; Emery, Alice B.; Robinson, Danielle
R. Bicyclists, helmets and head injuries: A rider-based study of helmet use and effectiveness.
American Journal of Public Health 78(9): 1120-1121, September 1988.
Weiss, Barry D. Bicycle helmet use by children. Pediatrics 77: 667-679; 1986.
Weiss, Barry D. Bicycle helmets: Effective but underused. Journal of the American Medical
Association 266(21): 3032-3033; December 4, 1991.
Weiss, Barry D. Trends in bicycle helmet use by children: 1985 to 1990. Pediatrics, 89(1):
78-80; January 1992.

40

Table 1.

Disposition of Telephone Calls

Disposition of Call

Number

(1)

Completed interview with rider

1,254

20.6

(2)

Screened out: household owns no bicycles

2,613

43.0

(3)

Screened out: household owns a bicycle,
but no household member rode in past year

479

7.9

(4)

Refused to answer any question

323

5.3

(5)

Broke off interview before completion

36

0.6

(6)

Contact with household made, but
interview could not be conducted*

362

6.0

Busy or no answer on all attempts

1,009

16.6

Total number of attempted calls

6,076

100.0

(7)

%

* Interview could not be conducted because of language barrier
or because the designated respondent was not available during
interviewing period.
Source: Abt Associates

41

Table 2.

Riders and Bicycles Per Household

Riders Per Household*
Riders
(millions)
1
2
3
4
5
$ 6
unknown

Percent

Total

Bicycles Per Household**

Households
(millions)

28.0
30.5
18.0
13.9
6.5
2.7
.4

7.6
8.3
4.9
3.8
1.7
.7
.1

100.0

27.1

Bicycles
1
2
3
4
5
$ 6
unknown
Total

* For households with riders.
** For households with bicycles.

42

Percent

Households

27.7
29.8
19.3
11.5
6.1
5.0
.6

10.4
11.1
7.2
4.3
2.3
1.9
.2

100.0

37.4

Table 3.

Household Demographics
Survey Results
(%)

1990 Census
(%)

Geographic Region
New England
Middle Atlantic
East North Central
West North Central
South Atlantic
East South Central
West South Central
Mountain
Pacific

5.5
12.6
19.7
9.8
15.8
5.1
11.8
6.2
13.5

5.4
15.1
17.0
7.3
17.9
6.2
10.5
5.5
15.1

Population Density
Large City or Suburbs
Medium City or Suburbs
Small City or Town
Open Country or Farm

20.5
22.5
35.1
21.9

NA
NA
NA
NA

MSA,
$ 2.5 million pop.
NA
MSA, 1.0-2.5 million pop.
NA
MSA, .25-1.0 million pop.
NA
MSA,
< .25 million pop.
NA
non-MSA
NA
(MSA=Metropolitan Statistical Area)

31.3
18.0
19.0
9.2
22.5

Household Size
One Person
Two Persons
Three Persons
Four Persons
Five Persons
Six or More Persons

3.1
13.3
16.0
31.1
23.5
13.0

24.5
32.3
17.3
15.5
6.7
3.7

Highest Education Attainment
High School or Less
Trade or Vocational Sch.
Some College
College Graduate
Attended Graduate School

25.5
3.3
22.0
33.8
15.4

58.4
NA
18.4
12.8
10.4

Total Household Income
Less than $15,000
$15,000-$29,999
$30,000-$44,999
$45,000-$59,999
$60,000 or more

8.5
22.5
30.2
20.7
18.1

24.4
25.7
20.3
12.7
16.9

Note: The unknown values from the survey (population density,
0.4 percent; income, 14.6 percent; household size, 0.5 percent;

43

and, education, 0.7 percent) were distributed evenly among the
other categories.

44

Table 4.

Profile of Riders

Characteristics

Riders
(%)

Projected Riders
(Millions)

Age (years)
10 or less
11-14
15-20
21-30
31-40
41-50
51 or more
Unknown
Total

25.2
14.3
10.4
15.6
18.3
8.6
6.1
1.5
100.0

16.9
9.5
7.0
10.4
12.2
5.8
4.1
1.0
66.9

Gender
Female
Male
Unknown
Total

47.0
52.3
0.7
100.0

31.4
35.0
0.5
66.9

Years Since Learned to Ride
3 or less
4-6
6-9
10 or more
Unknown
Total

18.9
15.9
12.2
52.4
0.6
100.0

12.6
10.6
8.2
35.1
0.4
66.9

45

Table 5.

Bicycle Use, Hours Per Year

Annual Riding Time
(hours/year)
1-24
25-49
50-99
100-199
200-399
400-599
600 or more
Unknown

% of Riders
20.2
8.7
11.8
15.2
15.2
3.9
8.1
16.9

Total

100.0

Mean Riding Time
Median Riding Time

Age Group
# 10 years
11-14 years
15-20 years
21-30 years
31-40 years
41-50 years
> 50 years
Age Unknown
Total

236.2 hours/year
105.0 hours/year

Mean Annual
Riding Time
(hours/year)
317.6
262.4
250.4
216.3
164.4
177.0
103.4
195.5
236.2

46

Estimated
Riding Time
(%)
33.9
15.9
11.0
14.3
12.7
6.4
2.7
3.1
100.0

Table 6.

Rider Practices and Use Patterns

Practice
Always
%

Proportion of Time:
More than Less than
Half
Half
Never
%
%
%

Unknown
%

Time Spent Riding on:
Sidewalk/Playground

18.4
(19.3)

10.8
(13.0)

17.7
(17.9)

53.1
(49.8)

0
(0)

Streets with Low
Traffic Volume

42.5
(38.8)

21.6
(19.7)

13.6
(14.5)

22.2
(26.9)

0.1
(0.1)

Major Thoroughfares
or Highways

2.9
(3.8)

3.9
(4.5)

12.9
(14.0)

80.1
(77.3)

0.2
(0.4)

Bike Paths

7.1
(6.1)

9.8
(11.6)

18.2
(19.4)

64.7
(62.8)

0.2
(0.1)

9.9
(12.3)

7.7
(11.0)

17.6
(17.6)

64.6
(58.8)

0.2
(0.3)

5.7
(8.2)

4.8
(7.7)

14.9
(21.4)

74.3
(62.6)

0.3
(0.1)

Commuting

5.0
(8.0)

3.7
(4.6)

10.1
(10.5)

80.8
(75.9)

0.4
(1.0)

Riding After Dark

1.0
(1.6)

2.1
(3.0)

9.3
(11.2)

87.1
(84.0)

0.5
(0.2)

13.4
(12.7)

4.2
(7.9)

6.0
(4.9)

76.0
(74.2)

0.4
(0.3)

1.5
(.4)
1.0
(.7)

0.8
(.1)
1.8
(1.5)

1.7
(1.0)
4.5
(6.7)

96.0
(98.5)
92.5
(90.8)

0
(0)
0.2
(0.3)

Unpaved Roads
Other Unpaved
Surfaces or Trails
Time spent:

Wearing Helmet
Carrying Infant in
Carrier or Trailer
Wearing Earphones

* Percentages without parenthesis represent percentages of
bicyclists. Percentages in parenthesis are adjusted to account for
estimated annual riding times.
47

Table 7.

Profile of Bicycles in Use

Bicycle Type Used Most Frequently
By Respondent

Estimated
Riding Time
%
14.3
24.5
9.0
16.6
26.2
9.4
100.0

%
13.5
14.1
14.1
17.3
34.5
6.5
100.0

Children's Sidewalk
BMX/High Rise
Middleweight/Cruiser
Mountain/City/All-Terrain
Lightweight Racing/Touring
Unknown
Total

Bicycle Type Used, by Length of Time Since Acquired
Years Since Acquired
# 1
%
18.7
16.5
8.0
24.9
26.6
5.3
100.0

Type
Children's Sidewalk
BMX/High Rise
Middleweight/Cruiser
Mountain/City/All-Terrain
Lightweight Racing/Touring
Unknown
Total

1 to <3
%
15.2
18.0
10.2
19.5
30.1
7.0
100.0

$ 3
%
6.1
5.7
25.8
6.8
48.7
6.9
100.0

Total
%
13.5
14.1
14.1
17.3
34.5
6.5
100.0

Bicycle Type Used, by Age of Rider
Rider Age, in years
Type
Children's Sidewalk
BMX/High Rise
Middleweight/Cruiser
Mountain/City/All-Terrain
Lightweight Racing/Touring

# 10
%
38.2
31.1
5.0
6.5
11.3

48

11-20
%
5.8
16.1
9.9
19.1
40.7

21-30
%
5.1
4.8
12.8
27.6
44.9

$ 30
%
4.8
4.0
24.4
19.3
42.5

All
%
13.5
14.1
14.1
17.3
34.5

Unknown
Total

7.9
100.0

49

8.4
100.0

4.8
100.0

5.0
100.0

6.5
100.0

Table 7 (continued)
Length of Time Since Acquired
One year or less
2-3
4-5
6-7
8 or more
Unknown
Total

%
28.5
42.1
12.1
6.2
10.9
0.2
100.0

Mean Number of Years Since Acquired
Median Number of Years Since Acquired

3.6
2.0
%
80.6
19.2
0.2
100.0

How Acquired
New
Used
Unknown
Total

%
38.8
30.5
26.3
3.9

General Condition of Bicycle
Like New
Better Than Average
About Average
Poor (i.e., abused, scarred,
rusted, etc.)
Unknown
Total

0.5
100.0
%
97.0
2.9
(1.0)
(0.6)
(1.1)
(0.2)
0.1
100.0

Modifications Made to Bicycle
None
Modification
Handlebars
Wheels
Seat
Others
Unknown
Total
Yes
(%)
14.5
20.6
12.5
4.5
8.5
8.8

Bicycle Accessories
Headlight
Tail Lamp
Bell or Horn
Child Carrier
Front Basket
Rear Basket or Carrier
50

No
(%)
85.2
78.7
87.4
95.5
91.4
90.9

Unknown
(%)
0.3
0.7
0.1
0
0.1
0.3

51

Table 8.

Helmet Use Information
Riders
(%)

Projected
Riders
(millions)

Rider Owns or Has Use of Helmet
Yes
No
Unknown
Total

27.3
72.4
0.3
100.0

18.3
48.4
0.2
66.9

Length of Time Owned or Had Use of Helmet*
Less than 1 year
1 to < 2 years
2 or more years
Unknown
Total

30.1
22.6
45.8
1.5
100.0

5.5
4.1
8.4
0.3
18.3

Total
%
77.9
14.1
5.1
2.9
100.0

14.3
2.6
0.9
0.5
18.3

13.4
4.2
6.0
76.0
0.4
100.0

9.0
2.8
4.0
50.9
0.2
66.9

17.0
11.4
13.7
18.5
19.7
24.6
23.1
15.4
17.6

2.9
1.1
1.0
1.9
2.4
1.4
0.9
0.2
11.8

Type of Helmet, by Length
of Time Owned (or Had Use of)*
Type
Hard Shell
Soft Shell
Thin Shell
Unknown
Total

< 1
%
64.9
21.5
8.6
5.0
100.0

Years
1 to <2
$ 3
%
%
80.9
85.0
15.9
8.4
0
5.4
3.2
1.2
100.0
100.0

Proportion of Time Spent Wearing Helmet
(for All Riders)
Always/Almost Always
More Than Half of Time
Less Than Half of Time
Never or Almost Never
Unknown
Total
Proportion of Riders Who Wear Helmets
All or Most of the Time
(for All Riders)
# 10 years
11-14 years
15-20 years
21-30 years
31-40 years
41-50 years
> 50 Years
Age Unknown
Total
52

* For the 27.3% of riders (projected at 18.3 million) who own or
have the use of helmets.

53

Table 9.

Helmet Use Patterns

For the 17.7 percent of Riders (projected at 11.8 million) Who
Purchased a Helmet (as opposed to receiving as a gift)
Importance of Factors
Some
in Purchase Decision:
Very
What
Not
Unknown
(%)
(%)
(%)
(%)
Cost
13.5
47.3
38.8
0.4
Appearance
10.6
45.0
42.6
1.8
Comfort
74.0
21.3
4.4
0.3
Weight of Helmet
46.7
31.7
19.8
1.8
Ventilation
41.9
35.0
22.2
0.9
Safety Certification
77.2
13.6
7.6
1.6
For the 13.4 percent of Riders (projected at 9.0 million) Who
Always Wear Helmets
Reasons for Wearing Helmet:
Yes
No
Unknown
(%)
(%)
(%)
Safety Reasons
97.8
2.2
0
Family Members (i.e.,
parent, spouse) insist
55.8
44.2
0
Local Legal Requirement
13.5
78.0
8.5
For the 10.2 percent of Riders (projected at 6.8 million) Who
Sometimes Wear a Helmet
Circumstances Under Which
Usually Wear Helmet:
%
When Riding in Traffic
40.0
When on Long Rides
25.2
When Remember To
12.4
When Riding with Family Members
11.6
When Reminded
17.9
Reasons Why Riders Do Not
Always Wear A Helmet:
Rider Forgets
Helmet Uncomfortable
When Riding A Short Distance
When Not Riding in Traffic
When Riding at Low Speeds
When on Bike Paths

%
22.9
14.4
31.6
23.8
6.6
6.9

For the 76.0 percent of Riders (projected at 50.9 million) Who Do
Not Own or Do Not Use a Helmet
Reasons Why Not Use Helmet
%
Do Not Ride Often
8.6
Never Thought About
21.6
Haven't Gotten Around To It
15.6
Seldom Ride in Traffic
18.8
Helmet Not Comfortable
8.9
Helmet Not Attractive
4.9
Helmets Are Too Expensive
7.3
54

Helmets are Unnecessary
Did Not Buy One for Child

21.0
17.2

55

Table 10.

Independent Variable Definitions

Rider Characteristics
-LN(EXPER)

-GENDER

The natural logarithm of years since learned to
ride a bicycle,
The natural logarithm of the estimated number of
riding hours per year,
1 if the rider is male, 0 if the rider is female,

-AGE

Rider age,

-AGE(X-Y)

1 if aged X to Y, 0 otherwise

-LN(HOURS)

Riding Surfaces
-SIDEWALK

1 if sidewalks or playgrounds are ridden on all or
most of the time, 0 otherwise,

-STREETS

1 if neighborhood streets with low traffic volume
are ridden on all or most of the time, 0
otherwise,
1 if major thoroughfares, highways, or streets
with high traffic volume are ridden on all or most
of the time, 0 otherwise,
1 if bike paths that are separate from roadways
are ridden on all or most of the time, 0
otherwise,

-HIGHWAY
-BIKEPATH

Demographic and other factors
-SCH1

1 if no household member has more than a high
school education, 0 otherwise,

-SCH2

1 if at least one household member attended
college but no household member graduated, 0
otherwise,
1 if any household member was a college graduate,
0 otherwise,

-SCH3
-CITY

1 if the household resides in a large or medium
size city (or suburbs of), 0 otherwise,

-NORTHEAST

1 if the rider household is in the New England or
Middle Atlantic States, 0 otherwise,

-MIDWEST

1 if the rider household is from the East North
Central or West North Central States, 0 otherwise,

-SOUTH

1 if the rider household is from the South
Atlantic, East South Central, or West South
Central States, 0 otherwise,
1 if the rider household if from the Mountain
States, 0 otherwise,

-MOUNTAIN

56

-PACIFIC

1 if the rider household is from the Pacific Coast
States, 0 otherwise.

57

Table 11. Regression Results -Factors Associated with Helmet Use
MODEL 1
VARIABLE
INTERCEPT
LN(EXPER)
LN(HOURS)
GENDER
AGE(1-10)
AGE(11-20)
AGE(21-30)
AGE(31-40)
AGE(41-50)
AGE
AGE@LN(HOURS)
SIDEWALK
STREET
HIGHWAY
BIKEPATH
CITY
SCH2
SCH3
NORTHEAST
MIDWEST
SOUTH
MOUNTAIN

COEFF.

SE

-1.745
0.072
0.136**
0.030
-0.113
-0.452
-0.269
-0.190
-0.206
---0.172
-0.214*
0.481**
0.629**
0.066
0.540**
0.776**
-0.253
-0.390**
-0.337*
-0.583*

0.415
0.080
0.033
0.099
0.284
0.250
0.243
0.234
0.251
--0.118
0.108
0.173
0.124
0.103
0.152
0.141
0.168
0.152
0.151
0.244

MODEL 2
COEFF.

MODEL 3
SE

-2.308
0.476
0.065
0.079
--0.028
0.099
0.981** 0.378
0.467
0.326
0.397
0.293
0.278
0.265
0.068
0.265
--0.005** 0.001
-0.156
0.117
-0.232* 0.108
0.473** 0.173
0.630** 0.124
0.062
0.103
0.557** 0.153
0.779** 0.141
-0.259
0.168
-0.399** 0.152
-0.336* 0.151
-0.605* 0.244

COEFF.

-1.680
0.272
0.091
0.081
---0.020
0.099
0.412** 0.158
---------0.015* 0.007
0.005** 0.001
-0.160
0.117
-0.229* 0.107
0.460** 0.173
0.626** 0.124
0.061
0.103
0.546** 0.152
0.774** 0.140
-0.259
0.167
-0.400** 0.152
-0.334* 0.151
-0.600* 0.243

** significant at p # 0.01, two-tailed test
* significant at p # 0.05, two-tailed test
N (helm=1)
N (helm=0)
DF
Model Chi-Sq
Score

177
795
19
116.591
115.269

177
795
19
119.862
119.867

58

SE

177
795
16
121.404
121.349

Table 12. Expected Probability of Helmet Use,
for Selected Rider-Bicycle Characteristics

Part A

Age
10
20
30
40
50

Male,
High School,
10 hours/year,
Street
(%)
7.7
3.9
3.8
3.6
3.4

Male,
High School,
300 hours/year,
Street
(%)
9.9
6.8
8.4
10.5
13.0

Male,
High School,
300 hours/year,
Bike path
(%)
30.2
22.2
26.6
31.5
36.9

Part B

Hours
Per Year
25
50
100
200
400
600

Female,
High School,
Street,
30 years
(%)
4.6
5.4
6.3
7.5
8.8
9.6

Female,
College,
Street,
30 years
(%)
14.1
16.4
18.9
21.8
24.9
26.9

59

Female,
College,
Highway,
30 years
(%)
33.1
37.1
41.3
45.6
49.9
52.5

Table 13.

Average Predicted Probability of Helmet Use,
for Selected Population Subgroups

Variable

Group

Average
Probability (%)

Age

# 10
11-20
21-30
31-40
41-50
> 50

16.5
14.5
19.3
18.8
27.6
24.4

219
246
179
170
78
80

Hours/Year

< 25
25-99
100-400
> 400

12.2
18.0
21.0
21.4

253
250
339
130

Gender

Male
Female

19.6
16.6

546
426

School

H. Sch. or less
Some College
College Grad.

7.8
18.4
23.7

259
241
472

Income
(in $1000)

< $15
$15-$29.9
$30-$44.9
$45-$59.9
$60 or more

15.2
15.3
17.7
20.1
22.4

72
202
256
166
139

Region

Northeast
Midwest
South
Mountain
Pacific

19.6
15.7
17.2
12.7
26.7

168
285
327
52
140

Live in City

No
Yes

16.3
20.7

537
435

No
Yes

19.0
16.4

686
286

Street

No
Yes

21.5
16.3

335
637

Highway

No
Yes

16.7
37.3

884
88

Ride on:
Sidewalk

60

N

Bike Path

No
Yes

15.1
34.8

61

797
175

Part III.
Bicycle-Related Injuries: Injury, Hazard, and Risk Patterns
____________________
Deborah Kale Tinsworth, Curtis Polen, and Suzanne Cassidy
Division of Hazard Analysis, Directorate for Epidemiology
January 1993

Introduction
The Directorate for Epidemiology estimates that about one-half million bicycle-related
injuries are treated annually in U.S. hospital emergency rooms. An additional 1,000 fatalities
occur each year, according to the National Safety Council (1). About two-thirds of the injuries
and about one-third of the deaths involve children under the age of 15 years. Head trauma has
been reported to be associated with the majority of these deaths and a substantial portion of the
injuries (2, 3, 4, 5).
The U.S. Consumer Product Safety Commission's (CPSC) early commitment to reducing
this annual toll was evidenced by the promulgation of a mandatory safety standard for bicycles in
1976 (6). This standard included requirements for mechanical and performance aspects of
bicycles, as well as requirements for instructions on bicycle assembly, operation, and maintenance.
In more recent years, CPSC has been involved in the development and revision of voluntary safety
standards [American Society for Testing and Materials (ASTM), American National Standards
Institute (ANSI)] for bicycle helmets (7, 8). A recently established Memorandum of
Understanding between CPSC and the National Highway Traffic Safety Administration (NHTSA)
promotes cooperative efforts between these agencies in the area of bicycle safety (9).
This report provides the results of a 1991 Directorate for Epidemiology special study of
the circumstances contributing to bicycle-related injuries treated in U.S. hospital emergency
rooms. It incorporates information from a 1991 Directorate for Economics exposure survey of
the current U.S. population of bicycle users and their patterns of bicycle and helmet use (10). A
brief overview of data on bicycle-related deaths is also included. Together, these data were used
to quantify and evaluate risk factors associated with bicycle use.
This information was developed in support of Commission efforts to address bicyclerelated hazards, and may be used as a resource by other organizations and individuals who have
an interest in bicycle safety.

55

Data and Methodology
The first stage of this analysis involved the descriptive presentation of data collected about
the circumstances involved in bicycle-related injuries treated in U.S. hospital emergency rooms.
The second stage involved the development of statistical models to identify and evaluate
risk factors associated with bicycle use. These models were developed to assess factors
associated with the general risk of injury to children and to adults. They utilized parallel sets of
data collected from injured as well as non-injured U.S. bicyclists so that risk comparisons could be
made.
While the primary purpose of the study was to evaluate risk factors associated with
bicycle-related injuries, national mortality data was presented to highlight some of the
circumstances involved in the most serious incidents involving bicycles. The following sections
describe the data sources and methodologies used for this study.
Injury Data
The bicycle-related incidents used for this study were identified through the National
Electronic Injury Surveillance System (NEISS). This system consists of a nationally
representative sample of over 90 U.S. hospital emergency rooms that report consumer productrelated injuries to CPSC on a daily basis. Information routinely collected through NEISS includes
the type of product associated with the injury; the victim's age, sex, diagnosis, disposition, and
body part injured; and a short narrative description of how the injury occurred. When further
information is needed, selected incidents may be followed by a more detailed investigation.
From January through December 1991, injuries reported to have involved bicycles were
sampled for telephone investigation to obtain additional information about the circumstances
involved in the incident and the victim's general patterns of bicycle use. Incidents in which a
mechanical failure of the bicycle was reported were reassigned for on-site investigation for further
examination of the bicycle.
CPSC investigators were instructed to discuss the incident circumstances and usage
patterns with the injured victim whenever possible. However, for cases in which the victim was
under the age of 15 years, the investigators were asked to contact the parent, guardian, or other
adult familiar with the incident.
In the telephone investigations, open-ended questions were used to obtain a narrative
description of the incident scenario and the nature and extent of the injury received. Other
questions were primarily closed-ended and were used to collect such information as: where the
victim was riding at the time of the incident; time of day; daylight conditions; use of lights and
reflectors; type, ownership, and condition of the bicycle; use of and damage to a helmet; and
other factors that may have contributed to the incident.

56

Data intended specifically for use in developing models to assess factors contributing to
the risk of bicycle-related injuries in general were collected primarily through closed-ended
questions. These questions were usually identical to those asked of the non-injured respondents
in the "exposure" survey described in the following section. They related to the frequency,
location, and time of bicycle use; the frequency and circumstances of bicycle helmet use; and
victim demographics such as area of residence, income, and education.
In all, 597 cases were selected (every 23rd bicycle case reported). Of these, 123 could not
be verified as being within the scope of the study because the victim could not be contacted or
was unwilling or unable to provide information about the incident. Of the remaining 474 cases,
11 (about 2 percent) were determined to be out-of-scope (e.g., involved a tricycle, motorcycle,
etc.). This analysis contains information from the 463 injury reports verified to have involved
bicycles. However, the majority of this analysis is based on the 420 cases that involved injury to
the bicycle operator, rather than to passengers, bystanders, and others.
Thus, information from the injury cases identified through NEISS was used 1) to provide
a general description of the circumstances involved in bicycle-related incidents, 2) to develop
statistical models that identified risk factors associated with bicycle use in general, when combined
with comparable information from the "exposure" survey.
Exposure Data
The "exposure" data noted earlier was obtained from a national probability survey of
about 1,250 households with bicyclists. It was conducted by Abt Associates during June and July
1991, for CPSC's Directorate for Economics. One rider was randomly selected in multiple rider
households; an adult was asked to respond for any randomly selected child. The sample was
weighted to reflect the number of riders in each household, and further weighted to represent the
94 million U.S. households in 1991.
Preliminary highlights from this survey indicated that there were about 67 million bicycle
riders in U.S. households in 1991, about one-half under 21 years of age. More than one-half of
nearly 12 million helmet wearers began wearing helmets in the last two years, and about 18
percent wear helmets all or more than half of the time (10).
Risk Comparisons
The national probability sample of verified in-scope bicyclist injury cases identified
through NEISS was combined with non-injury cases from the exposure survey. Respondents
were asked to provide information about the characteristics of persons, bicycles, and
environments involved in either bicycle-related injuries or bicycle use. From the information
received, the factors associated with injuries to children under 15 years of age, and with injuries to
adults age 15 and older, were examined.

57

Logistic regression techniques were used to develop statistical models (logit models) to
answer each question. Logit models are used to examine the relationship between a set of factors,
such as characteristics of persons, bicycles, or environments, and a dichotomous outcome, such as
injury or non-injury. This type of analysis is typically used to examine the contribution of each
factor while holding other factors constant (11).
Death Data
Information on bicycle-related fatalities was obtained primarily from two sources, the
National Center for Health Statistics (NCHS), and the National Highway Traffic Safety
Administration (NHTSA).
NCHS collects information on all deaths that occur in the United States each year. Data
on deaths involving bicycles were obtained from NCHS mortality data tapes for 1989, the latest
year available. Using international classifications published by the World Health Organization,
bicycle-related deaths were selected from External Cause of Death Codes E800 through E807,
with fourth digit .3; E810 through E819, with fourth digit .6; E826.1; and E826.9 (12).
NHTSA maintains the Fatal Accident Reporting System (FARS), which includes
information on fatal traffic crashes in the United States, including bicyclist fatalities involving
motor vehicles. Published FARS data from 1991 were used for this report (13, 14).
Analysis of Operator Injury Data
The 1991 estimate of bicycle-related injuries treated in U.S. hospital emergency rooms
was 588,000.1 From the special study, about 90 percent (about 531,000) of those injured were
the bicycle operators, about 4 percent were passengers, and about 3 percent were bystanders. An
additional three percent were injured in such other ways as when repairing or tripping over a
bicycle, etc.
Those injured as non-operators (i.e., passengers, bystanders, etc.) were significantly
younger (p <.01) than those injured while operating the bicycle. About 66 percent of the nonoperators were under the age of 10 years whereas 37 percent of the operators were under age 10
(See Appendix Table A1.) Of note was that none of the passengers treated in hospital emergency
rooms was reported to have been in a child carrier or wearing a helmet.
The remainder of this section includes information from 420 incidents in the sample in
which the bicycle operator was injured. Described are the riders' age, sex, rate of injury, body
part injured, diagnosis and disposition; hazard patterns; location of incident; time of day and
daylight conditions; helmet use; and bicycle type.

1

This estimate was adjusted using the proportion of special study cases verified to have involved a bicycle.

58

Victim Age, Sex, Rate of Injury
As shown in Table 1, about 37 percent of those injured were under the age of 10 years,
and about 71 percent were under the age of 15. About 62 percent of the victims were male.
Data from the CPSC Directorate for Economics' exposure survey indicated that there
were approximately 66.9 million bicycle riders in the United States in 1991. (The survey defined
a rider as a U.S. resident who rode a bicycle during the year prior to the survey.) Using this
information in conjunction with data from the injury study, age-specific injury rates were
calculated. Bicyclists were injured at an overall rate of 8.8 per thousand riders. Those in the 5-14
age group exhibited the highest rate, about 17 emergency room-treated injuries per thousand
individuals in that age group who rode bicycles. Those age 45-64 had the lowest rate, about 2.9
injuries per thousand riders in that group.
Age-specific injury rates were also calculated using annual riding times obtained from the
exposure survey. The estimated average riding time for U.S. bicyclists was about 237 hours per
year. Annual riding times were highest for younger riders, and generally decreased with age.
Overall, bicyclists were estimated to have been injured at a rate of 37.2 injuries per million hours
of use. Those age 65 and older demonstrated the highest rate of injury, about 61 injuries per
million hours of use.2 Bicyclists in the 5-14 age range also exhibited a high rate of injury, over 55
injuries per million hours of use. Bicyclists in the 25-44 age group were estimated to have had the
lowest rate of injury, about 18 injuries per million hours of use.
Thus, bicyclists in the 5-14 age range demonstrated higher rates of injury than most of the
other age groups, whether evaluated in terms of numbers of riders or in terms of annual riding
time. Riders age 25-64 demonstrated the lowest rates of injury. While the

2

Injury rates for bicyclists age 65 and older should be interpreted with caution due to small sample size for this age
group (n = 6).

59

Table 1
Bicycle Injuries: Age of Victim by Percent of Total
Injuries, Injuries per Thousand Riders,
nd Injuries per Million Hours of Use
644444444444444444444444444444444444444444444444444444444447
5 Victim
Percent of
Annual Injuries
Annual Injuries
5
Age
Total
per Thousand
per Million
5
Injuries
Riders
Hours of Use
K)))))))))0)))))))))))0)))))))))))))))0))))))))))))))))))))M
5 Total *
100.0* *
8.8
*
37.2
5
*
*
*
5
# 4 *
3.1
*
7.3
*
25.3
5
5-9 *
33.6
*
16.9
*
57.0
5 10-14 *
34.7
*
16.7
*
55.4
5 15-24 *
10.9
*
5.9
*
29.4
5 25-44 *
13.4
*
3.6
*
18.1
5 45-64 *
3.1
*
2.9
*
22.0
5
$ 65 *
1.1
*
7.2
*
61.0
5
*
*
*
9444444444N44444444444N444444444444444N444444444444444444448

5
5
5
5
5
5
5
5
5
5
5
5
5

* Column detail may not add to total due to rounding.
Source:

National Electronic Injury Surveillance System (NEISS): Special Study,
January-December 1991; National Survey of Bicycle and Helmet Use Patterns
in the United States, 1991 U.S. Consumer Product Safety Commission

sample size of injured bicyclists age 65 and older was small, the data suggested that bicyclists
of this age had a relatively low rate of injury per thousand riders in that age group. However,
they had the highest rate of injury per million hours of riding time.
Body Part, Diagnosis, Disposition
As shown in Table 2, the arm/hand and head/face areas each accounted for about onethird of the injuries, followed by the leg/foot area, accounting for about one-fifth of the
injuries.
Younger victims exhibited a significantly higher proportion of head injuries than older
victims (p < .01).3 About one-half of the injuries to those under the age of 10 years were to
the head/face area, as compared to about one-fifth of the injuries to victims over the age of 10.
Table 2
Bicycle Injuries: Body Part by Age of Victim
6444444444444L44444444444444444444444444444447
5
*
Age of Victim
5
*
5 Body Part * Total
<10 Years $10 Years
K))))))))))))3))))))))0))))))))))0)))))))))))M
5
Total * 100% *
100%
*
100%
K))))))))))))3))))))))3))))))))))3)))))))))))M
5 Arm/Hand *
32% *
19%
*
39%
5 Head/Face *
30% *
50%
*
19%
5 Leg/Foot *
22% *
21%
*
23%
5
Other *
16% *
10%
*
19%
9444444444444N44444444N4444444444N444444444448

Source:

5
5
5
5
5
5
5
5

National Electronic Injury Surveillance System
(NEISS): Special Study, January-December 1991
U.S. Consumer Product Safety Commission

Table 3 indicates that about 69 percent of the head/face injuries were lacerations,
contusions, and abrasions, relatively minor diagnoses. However, about 27 percent were
reported to be potentially serious injuries such as fractures, internal injuries, and concussions.
Injuries to the arm/hand area tended to be lacerations, contusions or abrasions (42 percent),
and fractures (40 percent). Injuries to the leg/foot area were primarily lacerations, contusions,
and abrasions (72 percent). Fractures were reported for about 13 percent of the leg/foot
injuries. Less than three percent of the victims were admitted for further hospitalization,
similar to all consumer product-related injuries reported through NEISS in 1991 (about four
percent).

3

In this section, all tests of significance were chi-square tests of independence. The p-value, or the Type I error
rate, indicates the probability of rejecting a true null hypothesis. For this study, the maximum probability for a
Type I error was set at 0.05.

61

Table 3
Bicycle Injuries:

Diagnosis by Body Part

644444444444444444444444444444444444444444444444444444444444444444444447
5
Body Part
5
5
Diagnosis
Total
Head/Face Arm/Hand Leg/Foot Other
K))))))))))))))))))))))0))))))))0))))))))))0)))))))))0)))))))))0)))))))M
5
Total * 100%* *
100%
*
100% *
100% * 100%
K))))))))))))))))))))))3))))))))3))))))))))3)))))))))3)))))))))3)))))))M
5
*
*
*
*
*
5 Lac./Contus./Abras. *
61% *
69%
*
42% *
72% * 65%
5
Fractures *
20% *
6%
*
40% *
13% * 17%
5
Strains/Sprains *
8% *
1%
*
15% *
9% *
7%
5
Internal Inj. *
5% *
14%
*
*
*
3%
5
Concussions *
2% *
6%
*
*
*
5
Other *
5% *
3%
*
4% *
6% *
8%
5
*
*
*
*
*
5
5
*
*
*
*
*
94444444444444444444444N44444444N4444444444N444444444N444444444N44444448

* Column detail may not add to total due to rounding.
Source:

National Electronic Injury Surveillance System (NEISS):
Special Study, January-December, 1991
U.S. Consumer Product Safety Commission

5
5
5
5
5
5
5
5
5
5
5
5

Hazard Patterns
Investigators recorded up to three factors that may have contributed to the incident,
using information collected through structured questions. Information collected through openended questions was used to verify these factors and to provide additional details about the
circumstances involved. Table 4 provides the proportion of cases associated with each of the
factors reported. Appendix Tables A2 and A3 provide, for each contributing factor, the
distribution of victim ages and where the incident occurred.
Table 4
Bicycle Injuries: Hazard Patterns
64444444444444444444444444444444447
5
5
Hazard Pattern
%
K)))))))))))))))))))))))))))0)))))M
5
Uneven Surface * 27%
5
Going Too Fast * 22%
5
Slippery Surface * 15%
5
Collis./Moving Object * 15%
5
Mechanical Failure * 13%
5
Collis./Non-Mov. Obj. * 13%
5
Performing Stunts * 11%
5
Obj. Caught in Spokes * 6%
5
Other * 29%
K)))))))))))))))))))))))))))2)))))M
5
Note: Column sums to more
5
than 100% due to multiple
5
causal factors contributing
5
to some incidents.
94444444444444444444444444444444448

Source:

5
5
5
5
5
5
5
5
5
5
5
5
5
5
5

National Electronic Injury Surveillance
System: Special Study, January-December, 1991
U.S. Consumer Product Safety Commission

Following is a brief summary of information on the incidents involving each hazard
pattern.
Uneven Surfaces. About 27 percent of the incidents were reported to have involved
uneven riding surfaces such as bumps, ruts, curbs, grates, holes, etc.
Riding Too Fast. About 22 percent of the incidents were reported to have involved the
victim "riding too fast." This was frequently reported in conjunction with other contributing
factors such as uneven or slippery surfaces, or performing stunts. "Lack of experience" was
also often mentioned in conjunction with this factor.

63

Slippery Surfaces. Slippery riding surfaces, such as those having loose stones/gravel,
sand, dirt, mud, grass, leaves, puddles, oil, ice, or snow were said to have contributed to about
15 percent of the incidents. About two-thirds of these incidents occurred on non-street
locations such as unpaved surfaces, trails, sidewalks, playgrounds, and bike paths.
Collisions with Moving Objects. Collisions with moving objects such as operating
motor vehicles, animals, pedestrians, and other bicyclists were reported for about 15 percent of
the incidents. Incidents associated with this hazard pattern most frequently occurred on
neighborhood streets (54 percent) or major thoroughfares (24 percent).
Mechanical Failure. Victims attributed about 13 percent of the incidents to a
mechanical or performance problem with the bicycle. CPSC's Directorate for Engineering
Sciences evaluated these cases and determined that the most frequently reported problems
were bicycle chains breaking or falling off, brakes failing, and various components such as
handlebars or brakes coming loose (15). While the causes of these reported failures could not
be absolutely determined, it appeared that poor bicycle maintenance and bicycle modification
were likely contributors to some of these incidents. In addition, it appeared possible that
factors such as a slick riding surface or unfamiliarity with the bicycle (e.g., brakes) could have
contributed to the operators' perception that a failure occurred.
Hit Non-Moving Object. About 13 percent of the incidents involved collisions with
non-moving objects, such as parked vehicles, traffic signs, posts, walls, fences, bushes, large
rocks, chains, and toys. More than two-thirds of these incidents occurred on non-street
locations such as sidewalks, playgrounds, trails, and bike paths.
Performing Stunts. About 11 percent of the incidents involved victims performing such
stunts as jumping over mounds of dirt, ramps, speed bumps, etc., and performing "wheelies."
Most of these victims (88 percent) were under the age of 15 years. Three out of four of these
incidents occurred on non-street locations such as sidewalks, playgrounds, trails, and unpaved
surfaces.
Items caught in spokes. About six percent of all incidents involved items caught in the
bicycle spokes, such as a foot/shoe, book bag, purse, pant leg, book, board, etc.
Other. Other factors reported as contributing to incidents were victims' inexperience,
inattention, and unfamiliarity with braking systems (either hand or foot brakes). Victims were
also reported to have been riding at night without a light; to have been riding a bicycle either
too big or too small for them; and to have fallen while avoiding motor vehicles, other
bicyclists, pedestrians, and animals.
Incidents involving motor vehicles involved collisions with both operating and parked
vehicles, and cases in which the victim swerved to avoid collision with a moving vehicle.

64

Those that involved collision or near-collision with an operating motor vehicle (not parked)
accounted for only about 10 percent of all incidents.
In the development of the study plan, there was interest expressed, from both within
and outside CPSC, in the contribution of radios or other devices with earphones to bicycle
incidents. However, there was little indication that these devices played a major role in the
incidents reported. Very few of the victims were said to have been wearing earphones at the
time of the incident.
Location of Incident
Overall, streets with relatively low traffic volume (e.g., neighborhood streets) were
associated with 41 percent of the injuries, more than any other location (Table 5). This was
followed by sidewalks or playgrounds, where about 12 percent of the injuries were reported to
have occurred. Major thoroughfares were associated with about 7 percent of the injuries;
unpaved roads and trails, each with about 5 percent; and bike paths, with less than 1 percent.
"Other" locations included driveways, yards, parking lots, alleys, etc.
When injuries were grouped to compare bicyclists age 25 and older to those of younger
age, and major thoroughfares to other locations, riders age 25 and older were injured on major
thoroughfares more frequently than younger riders
(p <.01). About 25 percent of those 25 and older were injured on major thoroughfares, as
compared to about 3 percent for younger victims.
Table 5
Bicycle Injuries: Location of Incident
by Age of Victim
64444444444444444444444444444444444444444444444444444447
5
Victim Age
5
5
5
5
Location
Total
<10 10-14 15-24 $25 5
K))))))))))))))))))))))0))))))0)))))0)))))0)))))0))))))M
5
Total * 100%** 100%* 100%* 100%* 100% 5
K))))))))))))))))))))))3))))))3)))))3)))))3)))))3))))))M
5 Neighborhood street * 41% * 33%* 50%* 50%* 38% 5
5 Sidewalk/playground * 12% * 19%* 10%*
9%*
8% 5
5
Major thoroughfare *
7% *
- *
6%*
7%* 25% 5
5
Unpaved road *
5% *
6%*
6%*
2%*
1% 5
5
Trail *
5% *
4%*
6%*
7%*
- 5
5
Bike path * <1% *
- *
1%*
- *
3% 5
5
Other * 28% * 38%* 20%* 25%* 25% 5
94444444444444444444444N444444N44444N44444N44444N4444448

* Column detail may not add to total due to rounding.
Source:

National Electronic Injury Surveillance System
(NEISS): Special Study, January-December,1991
65

U.S. Consumer Product Safety Commission

66

Time of Day, Daylight Conditions
Overall, about two-thirds of the incidents occurred between 6:00 am and 6:00 pm,
primarily in the afternoon. About one-third occurred between 6:00 pm and midnight, and less
than one in one hundred occurred between midnight and 6:00 am.
To accommodate seasonal variations in the length of day, a question was also asked
about the daylight conditions at the time of the incident (Table 6). This revealed that about 79
percent of the incidents occurred during daylight, about 16 percent during dawn or dusk, and
about 5 percent at night.
Incidents were combined to compare daylight with non-daylight conditions and major
thoroughfares to other locations. Non-daylight incidents (i.e., occurring at dawn, dusk or
night) were significantly more common on major thoroughfares than in non-thoroughfare
locations (p < .01). About 35 percent of the incidents that occurred on major thoroughfares
occurred during non-daylight conditions, as compared to about 19 percent of the incidents in
other locations.
While the small number of cases precluded drawing specific conclusions about the use
of bicycle lights during incidents that occurred in non-daylight conditions, less than eight
percent of all bicycles involved in incidents were reported to have been equipped with lights.
Of those with lights, victims were not always using the lights during incidents that occurred at
dawn or dusk. About nine out of ten victims were reported to have had reflectors on their
bicycles. Reflectors are currently required by the CPSC mandatory standard for bicycles.
Table 6
Bicycle Injuries: Location of Incident by Daylight Category
6444444444444444444444444444444444444444444444444444444447
5
Daylight Category
5
5 Location
Total
Dawn Daylight Dusk Night
K)))))))))))))))))))))0))))))0))))))0))))))0)))))0)))))))M
5
Total * 100% * <1% * 79% * 15% *
5%
K)))))))))))))))))))))3))))))3))))))3))))))3)))))3)))))))M
5 Sidewalk/playground * 100% *
- * 75% * 21% *
5%
5 Neighborhood street * 100% * <1% * 81% * 15% *
4%
5 Major thoroughfare * 100% *
5% * 65% * 26% *
4%
5
Bike path * 100% * 20% * 80% * - *
5
Unpaved road * 100% *
- * 87% * - * 13%
5
Trail * 100% *
- * 98% * 2% *
5
Other * 100% * <1% * 78% * 16% *
6%
5
*
*
*
*
*
9444444444444444444444N444444N444444N444444N44444N44444448

Source:

National Electronic Injury Surveillance System
(NEISS): Special Study, January-December 1991
U.S. Consumer Product Safety Commission
67

5
5
5
5
5
5
5
5
5
5
5
5

Helmets
About 12 percent of the victims were wearing a helmet. However, victims under 15
years of age were significantly less likely than older victims to have been wearing a helmet (p
<.01). See Table 7. About five percent of the victims under age 15 were reported to have
been wearing a helmet, as compared to about 30 percent of the victims age 15 and older.
Table 7
Bicycle Injuries: Age of Victim
by Helmet Use at Time of Incident
64444444444444444444444444444444444444447
5
Helmet Worn
5
5 Age Group
Total
No
Yes
K))))))))))))))0)))))))))0)))))))0))))))M
5
All Ages
*
100% * 88% * 12%
:44444444444444P444444444P4444444P444444<
5 Total <15
*
100% * 95% * 5%
5
<10
*
100% * 93% * 7%
5
10-14
*
100% * 98% * 2%
K))))))))))))))3)))))))))3)))))))3))))))M
5 Total $15
*
100% * 70% * 30%
5
15-24
*
100% * 84% * 16%
5
$25
*
100% * 61% * 39%
5
*
*
*
944444444444444N444444444N4444444N4444448

Source:

5
5
5
5
5
5
5
5
5
5
5

National Electronic Injury Surveillance System
(NEISS): Special Study, January-December, 1991,
U.S. Consumer Product Safety Commission

About one-half of the helmets were reported to have been hard shell types, about onethird were soft-shell, and the remainder were primarily thin-shell.
In about one-third of the cases in which the victim was wearing a helmet, it was
reported that the helmet was damaged in some way during the incident. Damage included
scrapes, scratches, dents, and cracks.
Of the 16 helmets in the sample that were damaged, 11 (69 percent) were worn by
victims who did not sustain head injuries. However, in all of the cases in which damage was
observed, the victim expressed the opinion that the helmet prevented a head injury or made the
head injury less severe. For helmet-wearers in general (regardless of helmet damage), about
two-thirds expressed the opinion that the helmet prevented a head injury or made the head
injury less severe.

68

Bicycles
BMX, freestyle, sidewalk, and high-rise bicycles were reported to have been involved
in about 43 percent of the incidents. Mountain, city, or all-terrain bicycles were said to have
been involved in about 25 percent of the incidents; lightweight racing or touring bicycles in
about 19 percent; and middleweight or cruisers in about 6 percent (Table 8).
An estimated 56 percent of the victims under 15 years of age were injured while using
types of bicycles typically intended for use by children, such as BMX/freestyle, sidewalk, or
high-rise bicycles. Only about 10 percent of those 15 and older were injured using these types
of bicycles. Instead, these victims were most frequently injured while using mountain, city, or
all-terrain bicycles, and lightweight racing or touring bicycles.
About 86 percent of the bicycles involved were owned by the victim's household, rather
than borrowed or rented. Of the bicycles owned by the victim, about 80 percent were
purchased new, rather than used. About three-fourths of the bicycles owned by the victim
were described as being like new, or in better than average condition at the time of the
incident. For bicycles purchased new, about one-half had been purchased less than one year
prior to the incident, and almost all (97 percent) had been purchased five years or less prior to
the incident.
The data suggested that borrowed or rented bicycles may not have been in as good
condition as those owned by the victim's household. About one-half of the borrowed or rented
bicycles were reported to have been in average or poor condition rather than like new or in
better than average condition.
About 94 percent of the respondents indicated that the structure or design of the
bicycle had not been changed or modified (other than minor repairs) prior to the incident.
Where modifications were made, they were primarily to the wheels, seat, and handlebars.
Other
About 80 percent of the injuries occurred during the six month period from April
through September. The months of June and July were associated with the greatest proportion
of injuries, each accounting for about 15 percent of the total injuries reported for the year.
January had the fewest injuries reported, accounting for about one percent of the total injuries
for the year.
Less than three percent of the incidents were reported to have occurred during
conditions of rain or snow, suggesting that precipitation was not a major contributor to
bicycle-related injuries treated in U.S. hospital emergency rooms.

69

Bicycle Injuries:

Table 8
Type of Bicycle by Age of Victim

64444444444444444444444444444444444444444444444444444444447
5
Age of Victim
5
Bicycle Type
Total
<15 Years
$15 Years
K))))))))))))))))))))))))0)))))))))0))))))))))0)))))))))))M
5
Total * 100%* * 100%
* 100%
K))))))))))))))))))))))))3)))))))))3))))))))))3)))))))))))M
5 BMX/Freestyle,
*
43%
*
56%
*
10%
5
Sidewalk, High-rise *
*
*
5 Mountain, City,
*
25%
*
21%
*
33%
5
All-Terrain
*
*
*
5 Lightweight
*
19%
*
14%
*
34%
5
Racing/Touring
*
*
*
5 Middleweight,
*
6%
*
4%
*
13%
5
Cruiser
*
*
*
5 Other
*
7%
*
6%
*
10%
9444444444444444444444444N444444444N4444444444N444444444448

5
5
5
5
5
5
5
5
5
5
5
5

* Column detail may not add to total due to rounding.
Source:

National Electronic Injury Surveillance System
(NEISS): Special Study, January-December 1991
U. S. Consumer Product Safety Commission

Risk Assessment
Logit models were used to examine factors that may contribute to the risk of injury for
1) children and 2) adults. Variables that were included in the models were age, gender,
number of hours ridden per month4, location where the bicycle was most frequently ridden,
size of metropolitan area in which the bicyclist resided, and whether a person rode
predominantly during the day or when there was less light. Some variables were not retained
in the models because they were not significantly related to the risk of injury and there was not
a strong theoretical basis for their inclusion. Other variables were excluded because of small
sample sizes.

4

Missing values for hours of use per month reduced the sample size of the models. No variable in the data set
seemed to predict the average hours of use per month for an observation, so the mean value was imputed for the
missing values in the injury and exposure samples. The highest correlation between any variable and log hours of
use was 0.12, indicating that other parameters would change little when missing values were imputed. Imputation
increased the sample size by nine percent for the children's model and by 15 percent for the adults' model. This
also decreased the standard errors for the variables in the models, allowing several variables which had been
borderline to emerge as significant.

70

The relative risk between levels of these variables was evaluated by estimating odds
ratios. These were calculated from the coefficients obtained from the logistic analysis. Odds
ratios indicate the change in risk between two categories of a variable being compared.
The variables were coded so that meaningful comparisons could be made. For
example, the variable that identified the location upon which a person predominantly rode was
coded so that the effect of riding on each location (i.e., riding on major thoroughfares,
sidewalks, bike paths, or unpaved surfaces) could be directly compared to riding on streets.
The variable that identified the size of the metropolitan area in which a person resided was
coded so that the effects of living in a suburb, a smaller city, and a rural environment were
compared to living in a large city. Age of bicyclist was coded in continuous form to measure
the effect of maturing one year. Average hours of use per month was coded in logarithmic
form to reduce the effect of extreme responses. For easier interpretation, the effect of more
riding time was estimated by doubling the number of hours of riding per month. Gender and
daylight were coded to compare the effect of being male to female and the effect of riding in
daylight to a category that included riding in dawn, dusk, or night.
Separate models were developed for children and adults in order to provide a clear
assessment of the risk factors associated with each age group. Of note was that while victim
age was not significantly related to the risk of injury in either the children's or adults' models,
children were found to be at significantly greater risk than adults when data from both age
groups were combined.5 The definitions of the independent variables and results of the
regression analysis are presented in Appendix Tables A4 through A6.
Model I: Factors Associated with Children's Bicycle-Related Injuries
Table 9 presents the variables that were included in the model for children, along with
the comparison groups and the changes in relative risk between comparison groups.
Increased risk of injury was associated with certain riding locations, time of day, and
place of residence. Children who rode in non-daylight hours, on streets,6 and who lived in
large cities were more likely to be injured.

5

In a combined model (Appendix Table A5), victims' ages were grouped for easier comparison. Those under
10 years of age were 6.28 times as likely to be injured in a bicycle-related incident as those age 45 and older.
Those age 10 through 14 years were about 6.83 times as likely to be injured as those age 45 and older. In addition,
as a group, those riders under age 15 were almost six (5.66) times as likely to be injured as those 15 years of age
and older.
6

Data for major thoroughfares and streets were combined in the children's model because of the relatively small
number of cases reported for major thoroughfares.

71

Table 9
Bicycle Injuries:

Relative Risks for Characteristics of Children

644444444444444444444444L444444444444444444444444L44444444444444444447
5 Characteristics of
*
Comparison Group
* Risk Relative to
5 Bicycle Riding
*
* Comparison Group;
5 and Bicyclist
*
* Odds Ratio
:44444444444444444444444P444444444444444444444444P4444444444444444444<
5
*
*
5 Rider aging one year * Rider age
* Non-significant
5 Males
* Females
* Non-significant
5 Double hrs. of use/mo.* Hours of use per month * Non-significant
5 Living in suburbs
* Living in large cities * Non-significant
5 Living in small towns * Living in large cities * 0.56 (1.79)*
5 Living in rural areas * Living in large cities * 0.47 (2.12)
5 Riding on sidewalks
* Riding on streets
* 0.60 (1.65)
5 Riding on bike paths * Riding on streets
* 0.12 (8.02)
5 Riding on unpaved
* Riding on streets
* 0.29 (3.44)
5
surfaces
*
*
5 Riding in daylight
* Riding at dawn, dusk, * 0.27 (3.64)
5
*
or night
*
5
*
*
944444444444444444444444N444444444444444444444444N44444444444444444448

5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5

* Numbers in parentheses are inverted odds ratios, and may be used for
easier interpretation when the direction of the risk comparison is
reversed; e.g., living in large cities was 1.79 times riskier than
in small towns (1 ÷ 0.56 . 1.79).
Source:

National Electronic Injury Surveillance System (NEISS):
Special Study, January-December 1991; National Survey of
Bicycle and Helmet Use Patterns in the United States, 1991

U.S. Consumer Product Safety Commission

73

For each measured effect on relative risk, all other effects that were modeled were held
constant. A child in a large city would be 1.79 times more likely to incur a bicycle-related
injury than a child in a small town, if they have the same riding habits.7 Similarly, a child in a
large city would be 2.12 times more likely to sustain an injury than a child in a rural
environment. Riding predominantly on streets was estimated to be 1.65 times more risky than
riding on sidewalks, 8.02 times more risky than riding on bike paths,8 and 3.44 times more
risky than riding on unpaved surfaces. The risk of riding predominantly at night was estimated
to be 3.64 times greater than riding in daylight.
The variables age, gender, hours of use per month, and living in suburbs were not
found to be significant predictors of bicycle injury to children.
Model II: Factors Associated with Adults' Bicycle-Related Injuries
Table 10 presents the variables that were included in the model for adults, along with
the comparison groups and the changes in relative risk between comparison groups. Bicyclists
age 15 and older differed from younger riders in 1) the effects which were significant for the
comparison groups and 2) the magnitude of change in relative risk for the significant effects.
Adults who rode more frequently, who rode on streets and major thoroughfares, and
who lived in large cities, were more likely to be injured.
An evaluation of relative risk revealed that for bicyclists age 15 and older, doubling the
number of riding hours per month increased the relative risk of bicycle injury by an estimated
1.40 times. Living in large cities was about 2.08 times as risky as living in small towns. As
with children, riding on streets was more risky than riding on bike paths and unpaved surfaces.
Riding on streets was estimated to be 6.93 times more risky than

7

Within the discussion of these models, some of the differences in relative risk were expressed using inverted
odds ratios. This reversed the risk comparison for greater ease of interpretation. For example, it was more
meaningful to say that a child living in a large city was about 1.79 times more likely to be injured than a child in a
small town, rather than to say that a child living in a small town would be about 0.56 times as likely to be injured
as a child in a large city (i.e., 1 ÷ 0.56 . 1.79). Calculated values may differ from those in the text due to rounded
odds ratios presented in Tables 9 and 10.

8

The number of children who rode predominantly on bike paths in the exposure sample was 26 (7.3 percent).
In the injury group, there were only three observations (2 percent). This may be due to bike paths being safer
places for children to ride a bicycle or it may be due to sample variation yielding an injury estimate that was very
small.

74

Table 10
Bicycle Injuries:

Relative Risks for Characteristics of Adults

644444444444444444444444L444444444444444444444444L4444444444444444447
5 Characteristics of
*
Comparison Group
* Risk Relative to 5
5 Bicycle Riding
*
* Comparison Group;5
5 and Bicyclist
*
* Odds Ratio
5
:44444444444444444444444P444444444444444444444444P444444444444444444<
5
*
*
5
5 Rider aging one year * Rider age
* Non-significant 5
5 Males
* Females
* Non-significant 5
5 Double hrs. of use/mo.* Hours of use per month * 1.40
5
5 Living in suburbs
* Living in large cities * Non-significant 5
5 Living in small towns * Living in large cities * 0.48 (2.08)*
5
5 Living in rural areas * Living in large cities * Non-significant 5
5 Riding on sidewalks
* Riding on streets
* Non-significant 5
5 Riding on major
* Riding on streets
* 2.45
5
5
thoroughfares
*
*
5
5 Riding on bike paths * Riding on streets
* 0.14 (6.93)
5
5 Riding on unpaved
* Riding on streets
* 0.11 (8.84)
5
5
surfaces
*
*
5
5 Riding in daylight
* Riding at dawn, dusk
* Non-significant 5
5
*
or night
*
5
5
*
*
5
944444444444444444444444N444444444444444444444444N4444444444444444448

*Numbers in parentheses are inverted odds ratios, and may be used for
easier interpretation when the direction of the risk comparison is
reversed; e.g., living in large cities was 2.08 times riskier than in
small towns (1 ÷ 0.48 . 2.08).
Source:

National Electronic Injury Surveillance System (NEISS):
Special Study, January-December 1991; National Survey of

Bicycle and Helmet Use Patterns in the United States, 1991
U.S. Consumer Product Safety Commission

76

riding on bike paths and 8.84 times more risky than riding on unpaved surfaces. Riding on
major thoroughfares was found to be 2.45 times more dangerous than riding on streets.9
The variables age, gender, living in suburbs, living in rural areas, and riding
predominantly on sidewalks were not found to be significant predictors of bicycle injury to
adults.
Other
Currently, there is widespread interest in the effectiveness of helmets in preventing head
injuries. Therefore, logit models were developed to identify factors associated with head
injuries, using data from the injury cases identified through NEISS. These models, however,
were affected by small sample size and significant conclusions could not be drawn about the
effects of helmets.
Bicycle-Related Deaths
The focus of this study was to evaluate bicycle-related injuries treated in U.S. hospital
emergency rooms. Almost all were non-fatal, as were the great majority of casualties
associated with bicycles. The majority did not involve motor vehicles. However, some
information about bicycle-related fatalities, which are primarily traffic-related, has been
included. This section provides a brief overview of these deaths, and highlights some of the
differences in the patterns of injury associated with fatal versus non-fatal incidents. Hazard
patterns specifically associated with fatal bicycle incidents involving motor vehicles also are
described.
According to the National Safety Council, an average of about 1,000 bicycle-related
deaths occur annually in the United States (1). Additional information about these fatalities
was obtained from the National Center for Health Statistics (NCHS).10 In 1989, the latest year
for which data were available, there were about 890 deaths reported in which a person was
fatally injured while riding a road transport vehicle operated solely by pedals, such as a bicycle.

9

As with the children's sample, the sample sizes for some categories of injuries occurring to adults was small.
This may be because the risk of bicycle-related injury for adults was less than that for children. Only 79 injury
observations had the information required for the analysis of adult risk factors. There were small sample sizes in
the injury sample for living in rural areas, and for being injured on sidewalks, bike paths and unpaved surfaces.
There were only three injury observations for bike paths (4 percent of the injury sample) and only two injury
observations for unpaved surfaces (3 percent of the injury sample). However, the small sample sizes for injuries in
these locations would make the estimated odds ratios sensitive to small changes.

10

In recent years, CPSC has not collected death certificate data for bicycle-related incidents involving motor
vehicles.

77

About 90 percent of these deaths involved motor vehicles, in contrast to about 10 percent of
the injuries treated in U.S. hospital emergency rooms.

78

As shown in Table 11, those who died from bicycle-related incidents tended to be older
than those treated in hospital emergency rooms for primarily non-fatal injuries. About 63
percent of those who died were age 15 or older, and about 17 percent were 45 or older. This
compares to about 29 percent and 4 percent, respectively, of the victims reported through
NEISS.
In addition, fatal incidents appeared more likely to have involved males. About 85
percent of the fatalities were male, in contrast to about 62 percent of those treated in hospital
emergency rooms.
Table 11
Bicycle Injuries: Age and Gender of Victims
by Percent of Deaths and Injuries
64444444444444444444444444444444444444444447
5
5
Victim Age
Deaths
Injuries
5
and Gender
(NCHS)
(NEISS)
K))))))))))))))))0))))))))))0))))))))))))))M
5
Total * 100.0 % * 100.0 %
K))))))))))))))))3))))))))))3))))))))))))))M
5
# 4 *
2.0
*
3.1
5
5-14 *
35.0
*
68.4
5
15-24 *
21.3
*
10.9
5
25-44 *
25.2
*
13.4
5
45-64 *
9.5
*
3.1
5
$ 65 *
7.0
*
1.1
5
*
*
5
Male *
85.3
*
62.3
5
Female *
14.7
*
37.7
5
*
*
94444444444444444N4444444444N444444444444448

Source:

5
5
5
5
5
5
5
5
5
5
5
5
5
5

National Center for Health Statistics (NCHS),
1989; National Electronic Injury Surveillance
System (NEISS): Special Study, January-December
1991, U.S. Consumer Product Safety Commission

Fatal incidents involved a greater proportion of head injuries than those resulting in
non-fatal injuries treated in hospital emergency rooms. A recent study (2) indicated that about
62 percent of all bicycle-related deaths involved head injury, as compared to about 30 percent
of the non-fatal injuries reported through NEISS. The majority (about 87 percent) of the
bicycle-related head injury deaths involved motor vehicles.
The National Highway Traffic Safety Administration's (NHTSA) Fatal Incident
Reporting System (FARS) contains additional information on pedalcyclist fatalities involving

79

motor vehicles (13, 14). In 1991, 841 pedalcyclists died from motor vehicle traffic crashes,
about two percent of all motor vehicle traffic crash fatalities for that year.
For about two-thirds of the pedalcyclist incidents included in FARS, police reported at
least one factor that may have contributed to the crash. Failure to yield right-of-way was the
most frequently reported factor, contributing to about 22 percent of the fatal incidents.
Improper crossing of roadway or intersection was the second most frequently reported factor,
contributing to about 13 percent of the incidents. Other factors included failure to obey traffic
signs, signals, or officer (9 percent); inattention (8 percent); operating vehicle in erratic,
reckless, careless, or negligent manner (6 percent); failure to keep in proper lane or running off
road (5 percent); and operating without required equipment (5 percent). About three-fourths
of these deaths occurred at non-intersection locations.
Almost one-half (46 percent) of the fatal incidents involving motor vehicles occurred
between 6:00 pm and 5:59 am, suggesting that night riding may be a contributing factor in a
portion of bicycle-related deaths.
For more than one-third of the motor vehicle crashes that resulted in a cyclist fatality,
alcohol involvement was reported for either the motor vehicle driver or the cyclist. For
pedalcyclists specifically, about one-fourth had blood alcohol concentration levels of 0.01
grams per deciliter (g/dl) or greater, and one-sixth were intoxicated (blood alcohol
concentration levels 0.10 g/dl or greater).
Discussion
Issues included for additional discussion were those related to head injuries and helmet
use, the need for modifications to the mandatory standard for bicycles, and risks associated
with night riding.
Head Injuries/Helmet Use
In recent years, there has been a growing public awareness of head injuries and helmet
safety. Data collected for this study corroborated this concern.
Almost one-third of all bicycle-related injuries treated in U.S. hospital emergency
rooms in 1991 involved the head/face area. About one-fourth of these injuries involved
potentially serious injuries such as fractures, internal injuries, and concussions. A recent study
indicated that almost two-thirds of all bicycle-related deaths involved head injury (2).
Children appeared to be at particular risk of head injury. About one-half of the injuries
to children under the age of ten involved the head and face, as compared to about one-fifth of
the injuries to older children. Children were also less likely to have been wearing a helmet at
the time of a bicycle-related incident than adults.
80

Examination of the 16 cases in the sample for which helmet damage was reported
revealed that in 11 cases (69 percent), the helmets were worn by victims who did not sustain
head injury. Using the assumption that the helmet damage resulted from head impact, a "best
guess" estimate of the effectiveness of helmets in preventing head injury would therefore be
about 69 percent. It is clear, however, that this estimate was based on a very limited number
of cases.
Statistical modeling techniques used to evaluate factors that contributed to the risk of
bicycle-related head injury were not conclusive about the effects of helmet use.
Neither the examination of damaged helmets nor the statistical modeling techniques
should be viewed as definitive methods of measuring helmet effectiveness. Neither included
cases in which bicyclists were involved in helmet impact incidents and did not seek emergency
room treatment (because injuries were prevented or were minor). While these cases were
beyond the scope of this study, the inclusion of such cases probably would have increased the
potential for accurately estimating the benefits of helmet use. It is also possible that if
sufficient data had been available to discriminate adequately between levels of head injury
severity, benefits could have been estimated.
Nevertheless, other research has shown that helmets substantially reduce the risk of
head injuries to bicyclists (16). This is also consistent with the opinion expressed by about
two-thirds of those injured while wearing helmets, which was that the helmet either prevented
a head injury or made the head injury less severe.
From all available information, it is clear that helmet use should be encouraged.
Modifications to the Mandatory Bicycle Standard
The mandatory safety standard for bicycles includes requirements for mechanical and
performance aspects of bicycles, as well as requirements for instructions on bicycle assembly,
operation, and maintenance.
For about 13 percent of the estimated injuries, the victim attributed the injury to a
mechanical or performance failure of the bicycle. Reported problems included chains breaking
or falling off, brakes failing, and various components such as handlebars or brakes coming
loose.
CPSC's Directorate for Engineering Sciences evaluated these cases and determined that
while the causes of these reported failures could not be absolutely determined from the
investigations, poor bicycle maintenance and bicycle modification were likely contributors to a
number of incidents. In addition, it appeared possible that environmental factors or
unfamiliarity with the bicycle (e.g., brakes) could have contributed to the perception that a
failure occurred. Based on information from this study, modifications to the standard were not
81

recommended. However, it would be appropriate to emphasize the importance of periodic
bicycle maintenance in future information and education activities.
Night Riding
Riding during non-daylight conditions (i.e., dawn, dusk, and night) was a significant
factor in the risk of injury for children. Non-daylight incidents were more common on major
thoroughfares than in other locations. NHTSA data on pedalcyclist deaths involving motor
vehicles suggest that night riding also may be a contributing factor in fatal incidents (14).
While it seems intuitively apparent that riding during dawn, dusk, or night would be
riskier than at other times, it is possible that some people perceive reflectors as adequate
protection at times when they may not be sufficient. It was reported that most bicycles were
equipped with reflectors. However, it was beyond the scope of this study to determine the
adequacy of the mandatory standard's reflector requirements.
While specific conclusions could not be drawn about the use of bicycle lights during
non-daylight incidents, less than eight percent of all bicycles involved in injuries were reported
to have been equipped with lights (regardless of daylight conditions). Of those with lights, a
few were involved in incidents that occurred at dawn or dusk while the light was not being
used.
Night riding may be an area deserving future information and education efforts (e.g. the
need for bicycle lights, reflective clothing, etc.).
Other
As might be expected, there was a higher risk of injury on streets and major
thoroughfares than in such areas as bike paths. Adults were at particular risk on major
thoroughfares. National data on bicycle-related deaths indicated that the great majority of fatal
incidents involved motor vehicles, and that adults were most often involved. From a safety
standpoint, it is reasonable to encourage bicycle use in low-risk locations appropriate for the
age of the rider.
Conclusions
These findings suggested the need for wider use of safety helmets, particularly by
children. In addition, riding in non-daylight conditions significantly increased the risk of injury
for children. Night riding may be an area for future information and education efforts. While
some mechanical problems associated with bicycle assembly, operation, and maintenance were
observed, no modifications to the existing mandatory safety standard for bicycles are
recommended based on this study.

82

References
1.

National Safety Council. Accident Facts, 1992 Edition. Itasca, Illinois: National
Safety Council, 1992.

2.

Sachs, Jeffrey J., MPH; Holmgreen, Patricia, M.S.;Smith, Suzanne M., M.D.; Sosin,
Daniel M., M.D.Bicycle-Associated Head Injuries and Deaths in the United States
From 1984 Through 1988." Journal of the American Medical Association 266
(December 1991):3016-3018.

3.

McKenna, Peter J., M.D.; Welsh, David J., M.D.; and Martin, Lester W., M.D.
"Pediatric Bicycle Trauma." The Journal of Trauma 31 (March 1991): 392-394.

4.

Selbst, Steven M., M.D.; Alexander, David, M.D.; and Ruddy, Richard, M.D. "BicycleRelated Injuries." American Journal of Diseases of Children 141 (February 1987):
140-144.

5.

Fife, Daniel, M.D.; Davis, Joseph, M.D.; Tate, Lawrence, M.D.; Wells, Joann K., B.S.;
Mohan, Dinesh Ph.D.; and Williams, Allan, Ph.D. "Fatal Injuries to Bicyclists: The
Experience of Dade County, Florida." The Journal of Trauma 23 (August 1983): 745755.

6.

16 CFR Section 1500.18(a)(12) Part 1512--Requirements for Bicycles.

7.

American Society for Testing and Materials (ASTM). ASTM F 1447 Specification for
Protective Headgear Used in Bicycling (proposed). Philadelphia, Pennsylvania: ASTM,
1992.

8.

American National Standards Institute, Inc. (ANSI). ANSI Z90.4-1984 American
National Standard for Protective Headgear - for Bicyclists. New York, New York:
ANSI, 1992.

9.

U.S. Consumer Product Safety Commission, National Highway Traffic Safety
Administration. Memorandum of Understanding: The U.S. Consumer Product Safety
Commission and the National Highway Traffic Safety Commission Join to Promote
Bicycle Safety. May 1989.

10.

Rodgers, Gregory B. Bicycle and Bicycle Helmet Use Patterns in the United States: A
Description and Analysis of National Survey Data. Washington, D.C.: U.S. Consumer
Product Safety Commission, Directorate for Economic Analysis, November, 1992.

11.

Hosmer, David W. and Lemeshow, Stanley Applied Logistic Regression. New York:
John Wiley and Sons, 1989.
83

12.

World Health Organization. Manual of the International Statistical Classification of
Diseases, Injuries, and Causes of Death, Ninth Revision, 1975. Geneva, Switzerland:
World Health Organization, 1977.

13.

National Highway Traffic Safety Administration (NHTSA). 1991 Pedalcyclist Fatal
Crash Facts. Washington, D.C.: National Center for Statistics and Analysis, 1993.

14.

National Highway Traffic Safety Administration (NHTSA). Fatal Accident Reporting
System 1991. Washington, D.C.: National Center for Statistics and Analysis, 1993.

15.

Heh, Scott R. Role of Mechanical Design and Performance in Selected Bicycling
Incidents. Washington, D.C.: U.S. Consumer Product Safety Commission, Directorate
for Engineering Sciences, September, 1992.
Thompson, Robert S., M.D.; Rivara, Frederic P., M.D., MPH; and Thompson, Diane
C., M.S. "A Case Control Study of the Effectiveness of Bicycle Safety Helmets." The
New England Journal of Medicine 320 (May 1989): 1361-1367.

16.

84

Appendix: Tables
Table A1
Bicycle Injuries: Age of Victim
by Mode of Involvement
644444444444444444444444444444444444444444444444444447
5
5
5
Mode of Involvement
5
5
5
5 Age of
Opera- Passen- Bystan5
5 Victim
Total
tors
gers
ders
Other 5
K))))))))0)))))))))0))))))))0)))))))0)))))))0))))))))M
5
* 587,970 *531,450 *22,240 *16,830 *17,450 5
5 Total * (100%) *(100%) *(100%) * (100%)* (100%) 5
K))))))))3)))))))))3))))))))3)))))))3)))))))3))))))))M
5
0-4 *
8%
*
3%
* 55% * 66% *
25% 5
5
5-9 *
32%
* 34%
* 19% * 13% *
17% 5
5 10-14 *
33%
* 35%
* 23% * 18% *
19% 5
5 15-24 *
11%
* 11%
*
2% *
*
24% 5
5 25-64 *
15%
* 17%
*
*
3% *
15% 5
5 $ 65 *
1%
*
1%
*
*
*
5
5
*
*
*
*
*
5
944444444N444444444N44444444N4444444N4444444N444444448

Source:

National Electronic Injury Surveillance System
(NEISS); Special Study, January-December 1991
U.S. Consumer Product Safety Commission

85

Table A2
Bicycle Injuries:

Hazard Pattern by Age of Victim

6444444444444444444444444444444444444444444444444444447
5
Age Group
5
5
Hazard Pattern
Total < 10 10-14 15-24
K))))))))))))))))))))))0))))))0)))))0)))))0)))))0)))))5
5
Total
* 100%** 37% * 35% * 11% *
5
*
*
*
*
*
5
Uneven Surface * 100% * 38% * 42% * 7% *
5
Going Too Fast * 100% * 47% * 26% * 10% *
5
Slippery Surface * 100% * 35% * 46% * 8% *
5 Collis./Mov. Object * 100% * 18% * 44% * 12% *
5
Mechanical Failure * 100% * 19% * 35% * 27% *
5 Collis./Non-Mov. Obj.* 100% * 42% * 39% * 14% *
5
Performing Stunts * 100% * 49% * 39% * 11% *
5 Obj. Caught in Spokes* 100% * 29% * 43% * 10% *
5
Other * 100% * 42% * 29% * 9% *
5
*
*
*
*
*
94444444444444444444444N444444N44444N44444N44444N444448

5
5
$25 5

18% 5
14%
16%
11%
27%
19%
6%
2%
19%
20%

5
5
5
5
5
5
5
5
5
5
5

* Row detail may not add to total due to rounding.
Source:

National Electronic Injury Surveillance System
(NEISS): Special Study, January-December, 1991
U.S. Consumer Product Safety Commission

86

Table A3
Bicycle Injuries:

Hazard Pattern by Location of Incident

644444444444444444444444444444444444444444444444444444444444444444444444444444444444447
5
5
Location
5
5
Neighbr. Sidwk/.
Major
Unpav.
Bike
5
Hazard Pattern Total Street
Plgrnd.
Thor.
Road
Trail Path
Other
K))))))))))))))))))))))0))))))0)))))))0))))))))0))))))))0)))))))0))))))0))))))0)))))))M
5
Total * 100%** 41% * 12%
*
7%
*
5% * 5% * < 1% * 28%
5
*
*
*
*
*
*
*
*
5
Uneven Surface * 100% * 37% *
7%
*
6%
*
8% * 14% * < 1% * 27%
5
Going Too Fast * 100% * 30% * 17%
*
3%
*
7% *
3% * < 1% * 40%
5
Slippery Surface * 100% * 34% *
4%
* -* 13% * 10% *
1% * 38%
5
Collis./Moving Obj.* 100% * 54% *
8%
* 24%
*
3% * -- *
1% *
9%
5
Mechanical Failure * 100% * 50% * 22%
*
6%
* -* -- *
1% * 21%
5 Collis./Non-Mov. Obj.* 100% * 25% * 30%
*
6%
* -*
9% *
2% * 29%
5
Performing Stunts * 100% * 17% * 14%
*
2%
* 13% * 14% * -- * 40%
5
Caught in Spokes * 100% * 71% *
4%
* 10%
* -* -- * -- * 15%
5
Other * 100% * 41% * 17%
*
4%
*
3% *
5% *
1% * 29%
5
*
*
*
*
*
*
*
*
94444444444444444444444N444444N4444444N44444444N44444444N4444444N444444N444444N44444448

* Row detail may not add to total due rounding.
Source:

National Electronic Injury Surveillance System (NEISS): Special Study,
January-December 1991
U.S. Consumer Product Safety Commission

5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5

Table A4
Definitions of Independent Variables in Risk Models
64444444444444444444444444444444444444444444444444444444444444447
5 Age
= Bicyclist's age
5
5 Age < 10
= 1 if bicyclist was under age 10; 0 otherwise
5
5 Age 10-14
= 1 if bicyclist was age 10-14; 0 otherwise
5
5 Age 15-24
= 1 if bicyclist was age 15-24; 0 otherwise
5
5 Age 25-44
= 1 if bicyclist was age 25-44; 0 otherwise
5
5 Gender
= 1 if male; 0 if female
5
5 Ln(hrs/mnth) = Natural log of hours ridden per month
5
5 Sidewalk
= 1 if bicyclist was injured on a sidewalk
5
(injury sample), or if bicyclist rode over 50
5
percent of the time on sidewalks (exposure
5
sample); 0 otherwise
5
5 Bike path
= 1 if bicyclist was injured on a bike path
5
(injury sample), or if bicyclist rode over 50
5
percent of the time on bike paths (exposure
5
sample); 0 otherwise
5
5 Highway
= 1 if bicyclist was injured on a major
5
thoroughfare (injury sample), or if bicyclist
5
rode over 50 percent of the time on major
5
thoroughfares (exposure sample); 0 otherwise
5
5 Unpaved
= 1 if bicyclist was injured on an unpaved
5
surface (injury sample), or if bicyclist rode
5
over 50 percent of the time on unpaved
5
surfaces (exposure sample); 0 otherwise
5
5 Suburb
= 1 if bicyclist lived in a suburb; 0 otherwise
5
5 Small town
= 1 if bicyclist lived in a small town; 0
5
otherwise
5
5 Rural
= 1 if bicyclist lived in a rural area; 0
5
otherwise
5
5 Daylight
= 1 if bicyclist was injured in daylight (injury
5
sample), or if bicyclist rode over 50 percent
5
of the time in daylight (exposure sample);
5
0 otherwise
94444444444444444444444444444444444444444444444444444444444444448

Source:

Bicycle Special Study, January-December 1991
U.S. Consumer Product Safety Commission

5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5

Table A5
Bicycle Injuries: Risk Factors, All Victims
(Results of Regression Analysis)
644444444444444444444444444444444444444444447
5 Variable
Coeff.
Std.
Stdzd.
5
Error
Coeff.
K)))))))))))))))))))))))))))))))))))))))))))M
5
5 Intercept
-1.9912
0.4118
.
5 Age < 10
1.8375* 0.3559
0.3977
5 Age 10-14
1.9218* 0.3448
0.4248
5 Age 15-24
0.3780
0.3800
0.0790
5 Age 25-44
0.0667
0.3589
0.0171
5 Gender
0.1576
0.1587
0.0431
5 Ln(hrs/mnth)
0.2463* 0.0645
0.1743
5 Sidewalk
-0.3283
0.2203 -0.0599
5 Bike path
-2.1715* 0.4764 -0.3936
5 Highway
0.6616* 0.2889
0.0884
5 Unpaved
-1.3914* 0.2626 -0.2998
5 Suburb
-0.2510
0.2131 -0.0594
5 Small town
-0.6279* 0.2028 -0.1650
5 Rural
-0.7786* 0.2679 -0.1650
5 Daylight
-0.6304* 0.2136 -0.1216
5
K)))))))))))))))))))))))))))))))))))))))))))M
5 N (injury)
259
5 N (no injury) 1129
K)))))))))))))))))))))))))))))))))))))))))))M
5 * Significant at p < .05
944444444444444444444444444444444444444444448

Source:

5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5

National Electronic Injury Surveillance System
(NEISS): Special Study, January-December 1991;
National Survey of Bicycle and Helmet Use Patterns
in the United States, 1991
U.S. Consumer Product Safety Commission

89

Table A6
Bicycle Injuries:

Risk Factors Associated with Children and Adults
(Results of Regression Analysis)

644444444444444444444444444444444444444444444444444444444444444444444444447
5
5
MODEL I. CHILDREN
MODEL II. ADULTS
5
:4444444444444444444444444444444444444444444L44444444444444444444444444444<
5 Variable
Coeff.
Std.
Stdzd. *
Coeff.
Std.
Stdzd.
5
Error
Coeff. *
Error
Coeff.
K)))))))))))))))))))))))))))))))))))))))))))3)))))))))))))))))))))))))))))M
5
*
5 Intercept
0.7024
0.5604
.
* -2.2771
0.5698
.
5 Age
0.0085
0.0330
0.0149 * -0.0139
0.0101 -0.1072
5 Gender
0.2462
0.2011
0.0665 *
0.1768
0.2625
0.0486
5 Ln(hrs/mnth)
0.1146
0.0799
0.0775 *
0.4851* 0.1098
0.3222
5 Sidewalk
-0.5036* 0.2549 -0.1169 *
0.4278
0.4322
0.0551
5 Bike path
-2.0823* 0.6423 -0.2581 * -1.9352* 0.6121 -0.3997
5 Highway
---*
0.8959* 0.3245
0.1403
5 Unpaved
-1.2361* 0.2941 -0.2803 * -2.1790* 0.7425 -0.4513
5 Suburb
-0.0201
0.2814 -0.0047 * -0.5582
0.3336 -0.1340
5 Small town
-0.5820* 0.2639 -0.1549 * -0.7328* 0.3196 -0.1908
5 Rural
-0.7522* 0.3362 -0.1650 * -0.8958
0.4646 -0.1850
5 Daylight
-1.2917* 0.3155 -0.2131 * -0.1547
0.3158 -0.0324
5
*
K)))))))))))))))))))))))))))))))))))))))))))3)))))))))))))))))))))))))))))M
5 N (injury)
186
*
N (injury)
77
5 N (no injury) 358
*
N (no injury) 771
K)))))))))))))))))))))))))))))))))))))))))))2)))))))))))))))))))))))))))))M
5 * Significant at p < .05
944444444444444444444444444444444444444444444444444444444444444444444444448

5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5
5

Source:

National Electronic Injury Surveillance System (NEISS): Special
Study, January-December 1991; National Survey of Bicycle and Helmet
Use Patterns in the United States, 1991
U.S. Consumer Product Safety Commission

91

Part IV.
Human Factors Assessment of Bicycle Incidents and
Training
_____________________
Celestine M. Trainor
Division of Human Factors, Directorate for Epidemiology
January 1993

The Division of Human Factors reviewed bicycle injury data from the Division of
Hazard Analysis and exposure survey data from the Directorate for Economic Analysis.
Human Factors also evaluated literature on bicycle safety education and training. The
literature focused on incident data and implications for training, with emphasis on the
capabilities of children.
This paper focuses on incidents involving children under 15 years of age. The physical
and cognitive development of children in this age group is discussed. Training
recommendations are considered and issues related to helmet usage for all ages are addressed.
Physical and Cognitive Development of Children
The Consumer Product Safety Commission (CPSC) estimates that each year there are
approximately a half million bicycle-related, emergency room-treated injuries, based on the
CPSC's National Electronic Injury Surveillance System (NEISS). A 1991 special Hazard
Analysis survey of bicycle incidents indicates that approximately 71 percent of the incidents
involved children under 15 years of age, with almost two-thirds (62 percent) of them boys
(Tinsworth et al., 1993). The physical and cognitive development of children may help to
explain why these incidents are occurring.
Children between 5 and 14 years of age experience many physical and cognitive
changes. It is during this age period that children are fond of testing their skills, such as riding
their bikes without holding the handlebars (Schickedanz et al., 1982). They try to see just how
far they can push their bodies physically, and may get their answers through injuries. Boys are
more daring and reckless; girls are more cautious (Gesell et al., 1977). This may explain why
80 percent of the injuries reportedly resulting from stunts involved males.
Also, children tend to display egocentric behavior (i.e., inability to perceive other
people's positions or viewpoints). Child development experts indicate that it is not until
around the age of 10 years that a child is able to consider his or her point of view and another
person's point of view simultaneously (Schickedanz et al., 1982). Before this time, children
assume that everyone sees the world as they do and, therefore, everyone knows what they are
87

thinking and planning. For example, children do not consider their behavior unexpected when
they suddenly turn in front of a car or dart out of a driveway, because that appears to them as
the only way to go.
Children generally learn to ride on flat, straight surfaces, so when they start to ride on a
variety of surfaces, they need to readjust how they ride their bikes. Riding on uneven surfaces,
which was reportedly involved in 27 percent of all incidents, requires children to concentrate
on several factors at the same time. Children need to know where they are going, what is
directly in front of them that could interfere with the motion of the bike, how to avoid
obstacles, how to avoid hitting other riders when riding with a group, and how to maintain
their balance. Maintaining balance while riding on uneven surfaces requires more control and,
in some cases, more physical strength than riding on a straight, flat surface.
Riders need to be constantly on the alert to avoid hazardous situations and need to be
physically able to avoid them. For example, children need to be able to pull up on the
handlebars to avoid ruts and roots on the surface if there is no other way to go around them.
Children also need to be aware of the likelihood that the tires may suddenly slide or turn
because of loose gravel or other hazards on the path. According to the incident data, slippery
riding surfaces, such as loose stones, gravel, sand, dirt, mud, grass, leaves, puddles, oil, ice or
snow were reported to have contributed to about 15 percent of the incidents.
Another factor frequently associated with child-rider injuries is speed. Riding too fast
was reported in 22 percent of the incidents. Children reportedly go too fast and lose control.
Typically, they simply think about how fast they can go. They do not think about the riding
surface, or the possibilities of something getting in the way, or even the chances of losing
control because of the high speed. Just as with uneven surfaces, riding fast requires children to
mentally process a lot of information at one time and then consider the alternatives.
Unfortunately, this cognitive ability is not something that can be taught to children at an early
age, but rather is a process learned over time as children reach higher maturational levels.
Since young children do not have the ability to mentally process all factors associated with a
behavior, they cannot adequately assess the consequences.
This cognitive immaturity may also explain why the number of bicycle injuries for this
age group has typically stayed the same over the years. Simple basic training on bicycle skills
may be necessary. However, training dealing with cognitive skills that are beyond young
children's mental capacity is not effective.

88

Training
Education programs should not simply tell cyclists what to do, but should teach why.
Bicyclists need to understand their role in safe bike riding and the consequences for failing to
follow safe practices. Frequently incidents occur because cyclists only consider the risk factor
for themselves and fail to consider others. Education programs should address the issue of risk
to other cyclists and motorists when cyclists choose to disobey traffic rules.
Several research studies report that urban bicycle and motor-vehicle incidents occur
most frequently at an intersection (Cross, 1978; Mathieson, 1986; Dewar, 1978). Typically,
bicyclists fail to obey traffic regulations by disregarding signals or signs, failing to yield the
right-of-way, and entering intersections from the wrong side of the street. This failure to obey
the regulations was seldom the result of the bicyclist's misunderstanding of the law - even for
the younger-aged cyclists. Rather, it was the cyclist's misjudgment of the risk associated with
the action (Cross, 1978). Some studies have indicated that a majority of serious, non-fatal
bicycle injuries are caused by cyclist error. When child cyclists are involved, they are legally at
fault close to 90 percent of the time (Mathieson, 1986; Dewar, 1978).
The City of Santa Barbara in California conducted a study to determine if a minimum
age for bicyclists should be required (City of Santa Barbara, 1975). In the study, the
researchers examined and compared stages of cognitive development of children with bicycle
and traffic safety requirements. The results show that children are unable to perform like
adults because they lack not only the knowledge and learning, but also the cognitive capacity
to perform functions as adults perform them (City of Santa Barbara, 1975). Too much
information is bidding for the children's attention, and they do not have the ability to filter it all,
much less filter it correctly. The study recommended that children between 7 and 13 years of
age only be allowed to ride bicycles on the street when accompanied by an adult. By 13 years,
most children have reached a maturity level that allows them to understand and comply with
traffic rules. While the recommendation seems reasonable developmentally, it is unrealistic.
Children between 7 and 13 years often ride their bikes to school and on neighborhood streets.
Requiring adult supervision would essentially ban bike riding for this age group. Since this is
not practical, other means of protecting children are necessary.
Child development experts also agree that certain concepts cannot be learned before a
certain maturational level, regardless of the amount of training. Therefore, determining the
time appropriate to begin bicycle education is essential. Generally, research findings suggest
that the optimal time for intensive bicycle-safety education campaigns is between the third and
sixth grades (9 to 12 years old) (Cross, 1978; City of Santa Barbara, 1975). This is not to say
that younger children should not have any type of training, but that a comprehensive program
would be most effective beginning in the third or fourth grades with refresher courses for older
children and adults. Most children by the sixth grade have the ability to understand and
perform the taught behaviors.

89

Another area of consideration is the use of protective equipment. In addition to
teaching simple basic bicycle skills and rules, teaching children and adults to wear protective
equipment such as bike helmets may help reduce the severity of their injuries.
Bicycle Helmets
While more and more information is becoming available to the public about the benefits
of bicycle helmets for all riders, the percentage of bicycle riders who use helmets all or most of
the time is still relatively low (Rodgers, 1992).
About 12 percent of all victims injured on bicycles were wearing helmets, and only
about two percent of those in the 10-14 age group were doing so (Tinsworth, 1992). Part of
this may be attributed to peer pressure and parental influence. While peer influence and
parental influence are not mutually exclusive, peer influence tends to have more value in such
matters as tastes in music and entertainment, and fashions in clothing. Parental influence has
more weight in moral and social values and understanding the adult world (Conger, 1973).
Therefore, the use of a bicycle helmet by children in this age group may be influenced more by
their friends than their parents, but reinforced when parents have the same attitude.
In two separate studies, children in grades 4 through 6 completed self-administered
questionnaires regarding their beliefs about helmet use. According to one study, the students
least motivated to wear helmets believe the social group to which they belong disapproves of
using a helmet (Otis et al., 1992). In a study conducted by the Eau Claire, Wisconsin Police
Department (1992), 33 percent of the respondents said they would wear a helmet if their
friends did and 46 percent said they might wear one if their friends did, compared to the 17
percent who said they would not wear one even if their friends did.
Other reasons students stated for not wearing helmets had to do with parental opinion
and misconception of the benefits of a helmet. Of the students who do not own a helmet, 18
percent stated their parents did not think helmets were necessary (Eau Claire, 1992). Of all the
students, 27 percent (14 percent of helmet owners; 30 percent of non-helmet owners) said they
do not need a bicycle helmet for the kind of bicycle riding they do. Specifically, helmets are
viewed as not necessary when children are only riding occasionally, or near their house (Eau
Claire, 1992).
Both the CPSC exposure survey and the special study asked respondents of all ages the
circumstances in which they wear and do not wear helmets. The most frequently stated
reasons for not wearing helmets were: riding only a short distance and not riding in traffic.
Coincidentally, riding a long distance and riding in traffic were the positive reasons stated for
wearing helmets by individuals who do wear them.
Based on these answers it appears that consumers are aware of the benefits of wearing
helmets when riding in traffic, but appear to be less likely to wear them when they believe they
90

are in less dangerous situations. This appears to be the same conclusion bicyclists give for
disobeying traffic rules.
Conclusions
Nearly three-quarters of bicycle-related injuries treated in hospital emergency rooms
involve children under 15 years of age. Almost twice as many boys as girls are treated. The
reasons for these incidents vary, but the main contributing factor may be children's cognitive
immaturity. Children under 15 years of age go through stages of cognitive development which
cannot be taught, but are learned over time. A major implication is that children are unable to
fully process in their minds situations and consequences, thus, they have accidents. While
training in basic bicycle riding skills and laws is important, trying to teach children concepts
that are cognitively too advanced is ineffective. Because children cannot be taught certain
skills that could help them avoid accidents, using protective equipment to help reduce the
severity of injury is an alternative.
Studies show that consumers recognize the benefits of wearing bicycle helmets when
riding in traffic and on long trips. However, they assess the risk level as low when they ride
only occasionally or for just short trips in their neighborhood. They are less likely to wear a
helmet in these situations. Peer influence also appears to be a factor in determining if a child
wears a helmet. Education programs and media campaigns should demonstrate that head
injuries occur on neighborhood streets and on short trips just as they do on long trips.
Education programs should not simply tell cyclists what to do, but should teach why. The
bicyclist needs to understand his/her role in safe bike riding and the consequences for failing to
follow safe practices.

91

References
Boughton, C. J., & Broadbent, A. (1986). "Bicycle Safety Current Knowledge." Appeared in
Bike Safe 86. National Bicycle Safety Conference, Newcastle, Australia.
Burden, D. (1986). "Select Bicycle Crash Studies." Appeared in Bike Safe 86. National
Bicycle Safety Conference, Newcastle, Australia.
Conger, J. (1973). Adolescence and Youth, New York: Harper & Row, Publishers, p. 331.
Conn, D. (1982). "Avoiding Those Common Crashes." Bicycling News Canada, 4(3), pp. 1213, 15.
Cross, K. D. (1978). "Bicycle-Safety Education. Facts and Issues." Anacapa Sciences, Inc.,
Santa Barbara, CA.
Dewar, R. E. (1978). "Bicycle Riding Practices: Implications for Safety Campaigns." Journal
of Safety Research, 10 (1), pp. 35-42
Eau Claire Police Department, WI (1992). "Bike Helmets: A Study of Their Use by Children
of the Eau Claire Area" Eau Claire Police Department, WI.
Gesell, A., Ilg, F., & Ames, L. (1977). The Child from Five to Ten. New York: Harper &
Row, Publishers (revised edition), p. 255.
Mathieson, J. G. (1986). "Gaps in Current Knowledge of Bicycle Safety and Effects on
Countermeasures." Appeared in Bike Safe 86. National Bicycle Safety Conference,
Newcastle, Australia.
Maring, W. & Van Schagen, I. (1990). "Age Dependence of Attitudes and Knowledge in
Cyclist." Accident Analysis and Prevention 22 (2): 127-136.
Padgett, S. (1975). "The Evaluation of the North Carolina K-9
Traffic Safety Curriculum: Methodology, Findings, and Recommendations." North Carolina
Univ., Chapel Hill. Highway Safety Research Center.
Otis, J., Lesage, D., Godin, G., Brown, B., Farley, C., & Lambert, J. (1992). "Predicting and
Reinforcing Children's Intentions to Wear Protective Helmet While Bicycling." Public Health
Report, 107 (3), pp. 283-289.
Rodgers, G., (1992). "Bicycle and Bicycle Helmet Use Patterns in the United States: A
Description and Analysis of National Survey Data." Washington, D.C.: U.S. CPSC
Directorate For Economic Analysis.
92

Schickedanz, J., Schickedanz, D., & Forsyth, P. (1982). Toward Understanding Children.
Boston: Little, Brown and Company, pp. 429, 469.
City of Santa Barbara, CA (1975). "Children Bicyclists: Should Minimum Age be Required?"
City of Santa Barbara Dept. of Public Works, CA Division of Transportation.
Tinsworth, D. & Polen, C, Cassidy, S. (1993). "Bicycle-Related Injuries: Injury, Hazard, and
Risk Patterns." Washington, D.C.: U.S. CPSC, Directorate for Epidemiology.
University of N. Carolina Highway Safety Research Center (1990). "Basics of Bicycling" as
cited in "Project Evaluates In-School Bicycle Safety Program" North Carolina Univ., Chapel
Hill.

93

Part V.
Role of Mechanical Design and Performance in Selected
Bicycling Incidents
____________________
Scott. R. Heh
Division of Mechanical Engineering, Directorate for Engineering Sciences
September 1992

Introduction
The Directorate for Epidemiology performed a special study of bicycling incidents
reported in calendar year 1991, through the National Electronic Injury Surveillance System
(NEISS). Using systematic random selection from nearly 14,000 reported bicycle-related
injuries, 463 investigations were completed, of which 420 involved injuries to the operator.
Epidemiology identified alleged mechanical failure as a possible contributing factor in
approximately one in ten (41/420) of the operator-injury cases. The Directorate for
Engineering Sciences analyzed these 41 incidents to confirm the alleged mechanical problem
and to identify any significant patterns of bicycle mechanical failure that could be addressed by
some means to reduce the frequency of accidents caused by mechanical failure.
The following discussion examines these bicycling incidents, grouped by the particular
component that reportedly failed and induced or contributed to the incident in some manner.
The forty-one incidents were divided into the following component groups: chains (13 cases),
brakes (15 cases), handlebars (6 cases), gear cables (1 case), seats (1 case), tires (2 cases),
spokes (1 case), handgrips (1 case), and pedals (1 case). A definitive cause for component
failure in these cases could not be established from the data. However, for each component
group the report discusses factors that may have contributed to the component failure. These
factors include bicycle maintenance and bicycle modification. External conditions are also
discussed as possible contributors in reported component malfunctions. In addition, the
analysis will present other identified factors which may have contributed to incidents in a
manner typically unrelated to the alleged component failure. These factors include operator
behavior and bicycle unfamiliarity. Table 1 gives an overview of the components involved and
any contributing factors associated with each incident.

95

Bicycle Component Groups and Injuries
Chains

96

Thirteen of the forty-one incidents involved bicycle chains. Six of the thirteen cases
reported the chain fell off the sprocket while the victim was riding. Two of the thirteen cases
reported chain breaks. The rider in these incidents typically lost control of the bicycle when
the chain released, causing the victim to fall. In three cases in which the chain either fell off or
broke, the victim was thrown from the bike when the released chain jammed the rear wheel.
Three of the thirteen chain incidents involved the rider attempting to make hand
adjustments to a loose or misadjusted bicycle chain while they were riding the bike. These
incidents resulted in serious finger injuries when fingers became pinched between the drive
sprocket and chain.
In two of the thirteen chain incidents, the victims reported losing control of the bikes
after experiencing performance-related problems. In one incident, the chain reportedly jerked
while shifting gears. In the second incident, the victim reported that the chain was skipping.
Contributing Factors. Maintenance was identified as a significant contributor in three
of the six chain incidents in which the chain fell off the sprocket. For example, in one of these
incidents, a sixteen-year-old boy lost control of the bike when the chain popped off. The
investigation stated that the sprocket was loose and the condition of the bicycle was poor.
Maintenance was also a likely contributor to the incident in which the victim lost
control and fell because the chain was "skipping". The bicycle in this case was described as
very old and in poor condition.
Operator behavior was a contributor to the injuries in the three incidents in which the
rider's fingers were pinched between the sprocket and chain. It is ill-advised for a rider to
make hand adjustments to a bicycle chain while riding the bike.
In one of the three finger pinching cases, a bicycle modification was a possible
contributor to the injury, although it had no active role in the initial chain malfunction. The
chain guard had been removed from the bike, which may have facilitated finger access to the
chain/sprocket area, leading to the pinching injury.
A bicycle chain modification may have been a direct contributor in one incident in
which the victim purposely shortened his chain. This modification may have induced a chain to
slip off the gears and into the rear wheel, stopping the bike suddenly and throwing the rider.
Other modifications were identified as a possible contributor in two additional chainrelated incidents. In the first case, the handbrakes had been removed from a bicycle.
Handbrakes may have prevented this accident in which the rider lost control of her bike and
drove into a roadside ditch when the chain came off the sprocket. In the second case, the
chain fell off the sprocket of a bike in which the chain guard had been removed. If a chain

97

guard had been present, it may have prevented the released chain from becoming caught in the
rear wheel, causing the bike to suddenly stop and throw the rider.
Five of the thirteen chain investigations reported chain malfunctions without identifying
the cause or other factors contributing to the incident.
Brakes
A total of fifteen incidents involved alleged brake problems. Twelve of these fifteen
reported that the brakes failed to adequately stop the bike. Nine of these twelve alleged brake
failures were reported to involve bicycles equipped with handbrakes. Two of twelve incidents
involved coaster brakes and one incident did not report whether it was a hand or coaster type
brake.
Loose brake components were reported in the three remaining brake cases. In these
three cases, front brake components (e.g. brake pad, caliper) came loose and jammed the front
wheel, causing the bikes to stop suddenly. The victims of these incidents were subsequently
thrown over the handlebars.
Contributing Factors. Inadequate maintenance was a possible contributor to one of the
coaster brake and one of the handbrake incidents. Further, poor maintenance was likely a
significant contributor to two other handbrake cases and one other coaster brake case.
Examples include two cases in which the victims had previous knowledge that the brakes were
not working properly. In one case the victim claimed that the handbrakes "went out", leading
him to ride into a cement post. The victim stated that he was having problems with his brakes
prior to the accident. In the second case, the victim fell off the bike after she realized she had
no brakes and she was about to collide with a stopped car. The victim stated that her brakes
had broken the previous day.
Rider behavior may have been a factor in one of the handbrake cases and two of the
coaster brake cases. An example of this is a case in which the victim had been drinking at the
time of the accident and described the cause of the accident as a combination of alcohol,
inexperience, and poor brakes.
A possible reason for reports of brake failure in several cases was the rider's
unfamiliarity with the bicycle and braking system. In three of the nine handbrake cases, it was
reported that the rider was not familiar with the bike or the bike's features and that this could
have contributed to the accident. For example, in one incident a forty-two-year-old woman
claimed that the handbrakes stuck and failed to bring the bike to a complete stop. However,
the investigation further reported that the victim was riding a borrowed bike and was not used
to handbrakes.

98

It is also possible that a slick riding surface contributed to two coaster brake incidents
and one handbrake incident. An example of this is a coaster brake incident that involved an
eight-year-old boy who was riding downhill on a wet, slippery surface. The boy could not stop
the bike, causing him to run into a piece of playground equipment. Although he claimed that
he was not able to stop the bike because the "chain guard or something" got stuck in the
brakes, the wet, slippery surface may have been a contributing factor.
Six of the fifteen brake related cases did not identify a cause for mechanical failure or
other factors contributing to the incident. This included all three of the incidents involving
front brake components coming loose and jamming the front wheel.
Handlebars
There were six cases in which loose handlebars were reported to be the cause of
accidents. Typically, riders were injured by falling from the bike after handlebars slipped
within the gooseneck clamp or at the gooseneck/fork tube connection. This occurred either
while riding or while attempting to mount the bicycle.
Contributing Factors. Poor maintenance was a possible contributing factor in four of
the six handlebar incidents. Two of the four incidents were attributed to loose bolts that fasten
the handlebars to the bike. In one of the four incidents, maintenance was viewed as a
secondary contributor. This case involved a worn handgrip that exposed sharp edges of the
metal handlebar. The victim was injured after the handlebars slipped and he lost control of the
bike. The exposed metal edge from the handlebar lacerated his leg.
Modification may have played a role in one of the handlebar incidents in which the bike
was constructed from various parts of a number of bikes. Combining parts of many bikes
could have led to an improper fit between the bicycle frame and handlebars.
Significant impact loading appears to have contributed to one handlebar incident. In
this case a twenty-year-old male was jumping over railroad ties. This subjected the bicycle to
abuse which may have led to the handlebars slipping within their "brace" and causing the rider
to lose control.
Gear Cable
In another loose component case, a gear cable and retention clip came loose and
tangled with a bicycle pedal. This stopped the bike suddenly, causing the victim to be thrown
over the handlebars. No other contributing factors were identified in this case.

99

Seat

100

One incident involved a five-year-old boy who was injured when he lost control of the
bike when the seat moved. Maintenance was likely a significant factor in this case since the
seat was described as loose prior to the incident and the bike was described as very old and in
poor condition.
Tires
Two incidents involved bicycle tires. In one case, it was reported that a rear tire, low
on air, may have been a contributing factor in an incident in which the rider fell off the bike
while turning. The second incident involved a rider who was thrown from the bike when it
stopped suddenly. The bike stopped when an inner tube caught on the front spokes after a tire
blow out. The tire was described to be in such poor condition that threads were visible on the
tire sides. Maintenance was viewed as a contributor in both of these incidents.
Operator behavior may have contributed to the above incident in which the rear tire
was low on air. The investigation reported that the victim was not feeling well which may
have affected her balance when she made a sharp turn and lost control.
Spokes
One incident reported that a front wheel spoke broke and wound around the fork. This
caused the victim to be thrown when the bike stopped suddenly. No other contributing factors
were apparent in this incident.
Handgrips
This incident involved another worn handgrip that allowed sharp metal edges on the
handlebars to be exposed. Operator behavior and bicycle maintenance were determined to be
factors in this incident in which the rider had stopped the bike and leaned down to pick up
something off the ground while he was straddling the bike. The victim lost control and fell.
The exposed metal edges on the handlebar landed on and fractured his finger.
Pedals
One incident that did not involve an active component failure was one in which a rider
fell after her foot slipped off a pedal. Wet conditions appear to have been a possible
contributor to the foot slippage. Also, maintenance may have played a secondary role since the
victim stated that the gear shifter was "wearing out". This demanded more concentration to
shift gears which the victim claims contributed to her foot slipping.

101

Borrowed Bikes

102

Unfamiliarity with a bicycle may not be limited to alleged brake failures as discussed
earlier. It was observed that eleven of the forty-one incidents involved riders who were using
borrowed bikes but did not report unfamiliarity with the bike as a contributing factor. A
person using a borrowed bike may be less aware of or less likely to inspect for mechanical
problems prior to riding the bike. In addition, a person riding a borrowed bike is much more
likely to be unfamiliar with its controls and features. These factors could contribute to
bicycling accidents.
Conclusions
A special study of bicycling accidents reported in calendar year 1991 through the
NEISS system identified a total of 41 bicycling accidents in which a mechanical problem with
the bicycle may have contributed to the incident. The most frequently reported problems were
bicycle chains breaking or falling off, brakes failing, and various components such as
handlebars and brake components coming loose.
While the cause of these alleged mechanical failures could not be absolutely determined
from the investigations, this analysis found that poor bicycle maintenance and/or bicycle
modification were contributors in a minimum of nine cases and possible contributors in an
additional eleven cases. External conditions, such as a slick road surface, were possible
contributors in four incidents. In addition, factors such as operator behavior and unfamiliarity
with a bicycle were possible contributors in twelve incidents, although typically not directly
related to component malfunctions.
Fifteen of the forty-one mechanically-related cases reported component malfunctions
without identifying the cause of the malfunction or any other factors contributing to the
incident. Investigative information was insufficient to determine if these incidents resulted
from inherent mechanical failure, that is, not attributable to poor maintenance, ill-advised
modifications, or other factors.
The results of this analysis do not show any significant mechanical failure patterns that
warrant an amendment to the existing mandatory bicycle standard. However, it is certainly
desirable to include some of the findings of this report in any future information and education
efforts relating to bicycle safety. Both adults and children should be made aware of: the
importance of maintaining a bike in good condition, the possible risks of modifying a bicycle,
and the possible risks presented by external conditions such as a wet, slippery riding surface.
The results of this analysis also suggest that operator behavior and unfamiliarity with a
bicycle may be contributors to many bicycle incidents. It would be beneficial to further explore
these factors as they relate to all bicycling incidents, since they do not apply solely to those that
report a mechanical malfunction.

103

Part VI.
Bicyclist Deaths and Fatality Risk Patterns
____________________
Gregory B. Rodgers, Ph.D.
Directorate for Economic Analysis
May 1993

In addition to nonfatal bicycle-related injuries evaluated in the CPSC injury survey,
there are large numbers of deaths involving bicyclists every year. Data from the National
Safety Council (1992) indicate that there have been an average of about 1,000 bicycle-related
deaths annually since 1975. The purpose of this report is to provide some information on
bicycle-related fatalities and on fatality risk patterns.
Data and Methods
The CPSC does not gather information on all bicycle-related deaths.1 However, data
are available from the National Highway Traffic Safety Administration's (NHTSA) Fatal
Accident Reporting System (FARS). Although the FARS data are limited to bicyclist deaths
resulting from crashes with motor vehicles on public roadways, FARS captures the great
majority of bicycle deaths. Based on information provided in Sachs et al. (1991) and NHTSA
(1993), the FARS system probably captures about 90 percent of all bicycle deaths.2
According to FARS (NHTSA, 1993), there were 841 bicycle-related deaths involving
crashes with motor vehicles in 1991. Table 1 presents the death data, by age and gender of the
victim. Just over one-third of the deaths (36.9 percent) involve children age 15 and under;
21.3 percent of the victims are over age 44, and 6.6 percent of the victims are age 65 and
older. In addition, almost 86 percent of victims are male.3

1

The CPSC's main source of information on deaths is from death certificates, which are
purchased by E-code from the states. However, only a subset of the E-codes pertaining to
bicyclist deaths is collected.
2

The National Center for Health Statistics (NCHS) also provides information on bicyclerelated deaths, including non-traffic deaths. However, the most recently available NCHS data
are from 1989. Based on the 1989 data about 10 percent of the bicycle-related deaths did not
involve collisions with motor vehicles (NCHS, 1992).
3

The age and gender distribution of the 1989 NCHS data on bicycle-related deaths were
quite similar to those of the FARS data. However, the victims of non-traffic related deaths
(about 10 percent of the total) appeared to be somewhat older than the victims of traffic103

The FARS data do not provide sufficient information for a detailed risk analysis of
bicycle deaths. However, it is possible to estimate relative risk factors for the various gender
and age categories. Table 2 shows the results of applying Bayes' Rule to the distribution of the
1991 FARS deaths and estimates of riding exposure from the 1991 CPSC exposure survey.4

Results
Column 1 of Table 2 shows the percentage distribution of the deaths, by gender and
age. For example, 85.6 percent of the victims were males and 14.2 percent were under age 10.
Column 2 shows the percentage distribution of riding exposure by gender and age; it is
calculated by determining the aggregate annual riding times for riders in each category, and
dividing this figure by the estimated aggregate riding time for all riders. For example, males
accounted for about 55.0 percent of all bicycle riding, and riders under age 10 accounted for
about 27.1 percent of riding time.
The risk for each of the age and gender categories, relative to the average risk for all
bicyclists, can then be calculated by dividing the percentage of deaths by the percentage of
riding exposure. Consider an example. As shown in the table, 85.6 percent of the fatal
accidents involved males, but males accounted for only 55.0 percent of aggregate riding times.
The risk of having a fatal accident, given that a rider is male, can therefore be estimated to be
about 1.56 (0.856/0.550) times the average risk. (This figure may be referred to as an index of
risk for males, because it describes their risk relative to the average risk.) Using the same
approach, the risk of a fatal accident for female bicyclists is about 0.32 (0.144/0.450) times the
average risk.
This method also yields comparative risk factors for the various age and gender
categories, relative to one another. Dividing the risk index for males by the risk index for
females, for example, indicates that males are about 4.88 (1.56/0.32) times as likely to be
involved in fatal accidents as females. That is, the "relative risk" for males is about five times
that of females.
The results from Table 2 also reveal some interesting findings with respect to the
relationship between age and the fatality risk. Although just over one-third of the deaths
involve children age 15 and under, this group of riders has a lower than average fatality risk

related deaths. For example, while 43.4 percent of the traffic-related deaths involved children
age 15 and under, only 18.1 percent of the non-traffic deaths did. Similarly, while 14.2 percent
of the traffic-related deaths involved riders over age 44, 37.4 percent of the non-traffic related
deaths did.
4

For further discussion of the application of Bayes' Rule, see Newman (1987) or Rodgers
(1989).
104

because they account for a large proportion of total riding times. In contrast, the exposureadjusted fatality risk for the 16-to-24 year-old age group is higher than average. When these
two groups are compared, the risk for 16-to-24 year-old riders is about 2.06 (1.38/0.67) times
higher than the risk for riders age 15 and under. Moreover, risk appears to be even higher for
older adult bicyclists, particularly those age 65 and older. When adjusted for riding exposure,
riders age 65 and older are about 3.19 (4.40/1.38) times more likely to be involved in fatal
accidents than 16-to-24 year-old riders, and about 6.57 (4.40/0.67) times more likely to be
involved in fatal accidents than riders age 15 and under.5
Additional Descriptive Information
NHTSA (1992, 1993) also provides some additional description of the bicycle-related
fatalities. In about 65 percent of the cases in 1991, police reported one or more errors in
bicyclists' behavior. The factor most often noted was "failure to yield right-of-way" (21.8
percent), followed by "improper crossing of the roadway or intersection" (12.6 percent), and
"failure to obey traffic signs and traffic control devices" (8.6 percent). In addition, one-fourth
of the cyclists killed had blood alcohol concentration (BAC) levels of 0.01 grams per deciliter
(g/dl) or greater, and one-sixth were intoxicated (i.e., BAC levels of 0.10 g/dl or greater).
Less than half of motor vehicle drivers involved in bicycle deaths were cited by police
for driving errors or other factors related to driver behavior. The factors most often noted
were "driving too fast for conditions or exceeding the speed limit" (21 percent), "drivers were
inattentive" (16 percent), and "vision obscured" (14 percent).
Almost three-quarters (73.0 percent) of the bicycle-related deaths occurred at nonintersection locations of roadways. Only 26.9 percent occurred at intersections. In addition,
23.5 percent of the deaths occurred between the hours of 9:00 PM and 5:59 AM. Although
daylight conditions vary somewhat during the year and by geographical location, most of these
probably occurred after dark.6 In comparison, 12.4 percent of riders from the exposure survey
reported that they ride at least occasionally after dark.7 Since most of these nighttime riders
(i.e., those from the exposure survey) ride only a small proportion of the time after dark,
nighttime riding appears to be an important contributing factor to bicycle deaths.

5

This analysis was also conducted using the NCHS data for 1989. The results were
essentially the same, and led to the same conclusions in all cases.
6

Another 22.8 percent of the deaths occurred between 6:00 and 8:59 PM, some of which
probably occurred after dark.
7

About 3.1 percent reported that they ride more than half of the time after dark and 9.3
percent said they ride less than half of the time after dark (Rodgers, 1992).
105

Summary and Discussion
The findings of this report suggest that age and gender are risk factors in bicyclerelated fatalities. When adjusted for riding exposure, risk appears to be higher than average for
16-to-24 year-old riders and for riders over age 44. Risk is highest for riders age 65 and older,
possibly because of a deterioration in their reaction time, an important characteristic in
avoiding accidents involving motor vehicles. Older persons involved in accidents may also
tend to suffer adverse outcomes because of medical complications and other factors associated
with post-injury homeostasis. Maring and van Schagen (1990), in a study of bicycle accidents
in the Netherlands, also found an increased accident risk for bicyclists over the age of 60.
They suggested that the higher risk might be related to changing cognitive and perceptual
processes that tend to reduce the flexibility of older riders in responding to unforeseen
situations.
The exposure adjusted risk for male riders is almost 5 times the risk for females, even
when risk is adjusted for riding exposure. This is consistent with automobile and other fatal
accident rates (see, e.g., National Safety Council, 1992; Rodgers, 1990), and suggests that
males are more likely to take risks than are females.
In addition, nighttime riding appears to be a contributing factor. While only a small
proportion of bicycle riding takes place after dark, approximately one-fourth of all fatal
accidents occur at night.
Information from police reports that were reported to NHTSA also suggests that many
fatal accidents are related to bicyclist roadway errors or other factors related to the riding
behavior. One or more bicyclist errors were reported by police in almost two-thirds of the
bicyclist deaths in 1991. Failure to yield right of way was reported for one out of every five
bicyclists killed. In addition, about one-sixth of the bicyclists were intoxicated at the time of
accident.

106

References
Maring, Wim, and Ingrid van Schagen. Age Dependence of Attitudes and Knowledge in
Cyclists. Accident Analysis and Prevention 22(2), April 1990, 127-136.
National Center for Health Statistics. Mortality Detail, 1989 Data. Washington, DC: Author,
September 1992.
National Highway Traffic Safety Administration. 1991 Pedalcyclist Fatal Crash Facts.
Washington, DC: Author, 1992.
National Highway Traffic Safety Administration. Fatal Accident Reporting System 1991.
Washington, DC: Author, 1993.
National Safety Council. Accident facts, 1992 Edition. Itasca, IL: Author; 1992.
Newman, Rae. Comparison of Bayes' Theorem and "Risk Index" Used in ATV Reports.
Memorandum to Dr. Robert D. Verhalen, AED, Directorate for Epidemiology, U.S. Consumer
Product Safety Commission, November 23, 1988.
Rodgers, Gregory B. Factors Contributing to Child Drownings and Near Drownings in
Residential Swimming Pools. Human Factors 31(2), April 1989, 123-132.
Rodgers, Gregory B. Evaluating Product-Related Hazards at the Consumer Product Safety
Commission: The Case of All-Terrain Vehicles. Evaluation Review, 14(1), February 1990, 321.
Rodgers, Gregory B. Bicycle and Bicycle Helmet Use Patterns in the United States. Technical
Report. Washington, DC: U.S. Consumer Product Safety Commission, 1992.
Sacks, Jeffery J., Patricia Holmgreen, Suzanne M. Smith, Daniel M. Sosin. Bicycle-Associated
Head Injuries and Deaths in the United States from 1984 through 1988. Journal of the
American Medical Association 266(21), December 4, 1991.

107

Table 1.

1991 Bicyclist Deaths, by Age and Gender

Age

Males

Deaths
Females

(years)

cases

cases

< 10
10-15
16-24
25-34
35-44
45-54
55-64
$ 65
Unknown
Total

94
153
102
117
82
69
45
50
8
720

24
36
18
15
15
4
4
5
0
121

Total
cases (percent)*
118
189
120
132
97
73
49
55
8
841

(14.2%)
(22.7%)
(14.4%)
(15.8%)
(11.6%)
(8.8%)
(5.9%)
(6.6%)
--(100.0%)

Source: National Highway Traffic Safety Administration
a The percent column applies to the 833 cases in which the age
of the victim is known.

Table 2.

Risk Characteristics

Characteristics

(1)
Deaths
(%)

(2)
Exposure
(%)

Gender
Male
Female

100.0
85.6
14.4

100.0
55.0
45.0

1.00
1.56
.32

Age
< 10
10-15
16-24
25-34
35-44
45-54
55-64
$ 65

100.0
14.2
22.7 @36.9
14.4
15.8
11.6
8.8
5.9
6.6

100.0
27.1
27.6 @54.7
10.4
15.5
12.6
3.9
1.5
1.5

1.00
.52
.82 @.67
1.38
1.02
.92
2.26
3.93
4.40

108

(3)
Risk Index
(1)/(2)

Source: National Highway Traffic Safety Administration and the
1991 CPSC Exposure Survey

109

Part VII.
Characteristics of Adult Bicyclists in the United States:
Selected Results from a National Survey
____________________
Mary F. Donaldson
Directorate for Economic Analysis
April 1993

In 1990, Rodale Press, the publisher of Bicycling magazine, sponsored a major survey
of adult bicycle owners. The survey gathered comprehensive information about the U.S. adult
bicycling population, their use patterns, and the bicycle equipment market. The CPSC
purchased this survey data to complement the CPSC bicycle injury and exposure surveys
conducted in 1991. Although the Rodale Press survey did not collect information on children,
a major focus of the CPSC exposure survey, it provided substantial marketing information
unavailable from the CPSC survey.
This paper summarizes the Rodale Press survey methodology and highlights
information of particular interest to CPSC1.
Methodology
Commissioned by Rodale Press, National Family Opinion, Inc. (NFO) conducted the
survey using its consumer mail panel. While not a probability sample, the panel is balanced to
match U.S. statistics on five demographic variables: geographic region, population density,
household income, household size, and age of panel member.
The survey, conducted in 1990, obtained information on the bicycle riding habits and
purchasing behavior of adult bicycle owners who purchased their bicycles new. To accomplish
this, NFO screened 150,000 panel households for bicycle ownership. Of the 103,774
households responding to the screening questionnaire, 40,381 were selected for further
analysis. These households had at least one adult, aged 18 or older, who owned a bicycle for
personal use. From this sample of 40,381 households (which included 61,587 adult bicyclists),
NFO further screened out those who did not purchase their last bicycle new and those who did
not answer three screening questions on rider characteristics: the number of visits to a bicycle
shop in the past year; the number of miles ridden in an average warm weather month; and, the

1

Rodale Press, The Cycling Consumer of the 90's, A Comprehensive Report on the
U.S. Adult Cycling Market, Emmaus, PA: Author; 1991.
109

price paid for the last bicycle purchased. Of the 61,587 adult bicyclists mentioned above,
37,863 (61 percent) qualified for the survey.
NFO then applied a clustering technique, based on answers to the three screening
questions, to separate bicyclists into groups ranging from those who hardly ever ride to those
who could be considered enthusiasts. From this analysis, four distinct groups emerged. These
groups were categorized by their level of participation and were described as "enthusiast",
"moving-up, "casual", and "infrequent" riders. Enthusiasts and moving-ups are the more avid
of the four clusters. As shown in Table 1, these two groups ride more, buy more expensive
bicycles, and visit shops more often than casual and infrequent riders (Rodale Press, 1991).
(Tables start on page 116.)
From each of the four identified clusters, NFO selected at least 1,000 households to
receive an indepth questionnaire. Of 4,209 households selected, 3,248 responded, including
624 enthusiast, 895 moving-up, 875 casual, and 854 infrequent riders. NFO assigned weights
to the responses to make them representative of the U.S. population of adult bicycle riders
who purchased their last bicycle new. (Rodale Press, 1991).
Overview of Survey Results
The results of the survey are representative (for 1990) of bicycle riders 18 years of age
and older, who bought their most recent bicycle new. NFO estimates that there are
approximately 32.8 million bicyclists who fit this description. By level of participation, NFO
projects that there are 0.9 million enthusiasts representing 2.7 percent of the total; 2.4 million
moving-up riders representing 7.3 percent of the total; 6.8 million casual riders representing
20.8 percent of the total; and 22.7 million infrequent riders representing 69.2 percent of the
total. Figure A graphically presents the breakdown by cluster of the total population reflected
by the survey. (Figures start on page 132.) The discussion and attached tables present
information on demographic characteristics of riders and their households, bicycle use patterns,
helmet use and safety behavior, bicycles and equipment owned, future purchase plans, and
equipment problems.
Rider and Household Demographics
The survey obtained detailed demographic information about the respondents and their
households. Respondents provided information on age, gender, marital status, education,
employment, and income. These data are shown in Tables 2 through 9.
The mean age of adult riders is 37.1 years and there is little variation between cluster
groups. Greater disparity is found in a breakdown of cluster groups by gender. Although
about half (49.4 percent) of riders are female, three-fourths of the enthusiasts and two-thirds of
the moving-up riders are male.

110

The survey indicates that adult bicyclists tend to be more educated than the overall
population, a finding that is consistent with the results of the CPSC exposure survey. About
40 percent of all adult bicyclists are college graduates, compared with about 23 percent of the
population as a whole. Enthusiast and moving-up riders have even higher education levels,
with over 50 percent holding college degrees.
The largest proportion of employed riders, 39 percent, work in managerial and
professional fields. The second largest category, with 27 percent, includes riders in the
technical, sales, and administrative support fields. Personal employment income is higher
among the enthusiast and moving-up categories, as would be expected, given their higher
education levels. Of those who said they were not employed, retired individuals and
homemakers predominate in the infrequent and casual categories, while students and
homemakers predominate in the enthusiast and moving-up categories.
Tables 10 through 15 summarize the demographic characteristics of rider households.
Some of the information is comparable to information gathered by the 1990 U.S. Census.
There are no major regional differences in the location of rider and U.S. households.
However, there appears to be a higher concentration of enthusiast and moving-up riders in the
Mountain and Pacific regions and a higher concentration of casual and infrequent riders in the
North Central and South Atlantic regions. Adult bicyclists also live in larger households than
the U.S. population overall. Nearly 80 percent live in houses and 75 percent own or are
buying their homes. More than half (56 percent) live in metropolitan areas with populations
greater than 500,000.
Use Patterns
A substantial amount of information from the survey provided details on use patterns
and riding habits. May through September are the peak riding months. At least 74 percent of
bicyclists ride their bicycles during these months. See Figure B. However, the more avid the
bicyclist, the higher the year round level of participation. During warm weather months,
almost 50 percent of bicyclists ride at least once a week. The median number of miles ridden
during a warm weather month is 10. Average monthly mileage in warm weather is 34. By
participation level, the average monthly warm weather mileage ranges from 205 for enthusiasts
to 24 for the infrequent riders. See Tables 16 and 17.
An overwhelming majority (80 percent) of bicyclists ride on neighborhood streets at
least some of the time. They also indicate that they sometimes ride on quiet streets or quiet
highways (37 percent), bike paths/rail trails (29 percent), rural roads (26 percent), busy streets
or busy highways (21 percent), bike routes (18 percent), off-road trails (9 percent), and other
places (11 percent). The higher the level of participation, the greater the number of places
ridden. However, the relative ranking of places ridden does not vary much between clusters.
See Table 18.

111

The largest percentage of bicyclists (62 percent) ride most often on neighborhood
streets. Other places most often ridden are rural roads (10 percent), bike paths/rail trails (8
percent), quiet streets and quiet highways (7 percent), busy streets or busy highways (3
percent), bike routes (3 percent), off-road trails (1 percent), and other places (2 percent).
Among clusters, the enthusiast and moving-up groups spend a relatively smaller proportion of
their riding times on neighborhood streets and more of their time on other streets or highways,
rural roads, and bike paths. See Table 19.
When questioned about the reasons for riding their bicycles, most riders (77 percent)
indicate that they ride for fitness and exercise and almost 50 percent ride as a family activity.
Commuting is listed as a purpose by 10 percent of all riders. However, over 20 percent of
enthusiasts and moving-up riders indicate that commuting is a reason for riding their bicycles.
See Table 20.
Bicyclists say they would ride more if there were 'someone to ride with' (46 percent),
'safer places to ride' (35 percent), 'more comfortable seats' (34 percent), and 'more scenic
places to ride' (29 percent). Even so, 57 percent have access to community bike paths and 28
percent have access to extra wide roads or bike lanes within their communities. Others would
ride more if they were in 'better physical condition' (27 percent), could 'ride to work' (14
percent), had 'gears easier to shift' (11 percent), had 'access to organized riding events' (4
percent) or took 'a training course in bicycle riding' (1 percent). See Tables 21 through 23.
Helmet Use and Safety Behavior
Overall, 16 percent of bicyclists said they owned helmets. By cluster, helmet
ownership is highest among enthusiasts (69 percent) and lowest among the infrequent riders (9
percent). However, not all helmet owners wear their helmets. Of those who own helmets, 60
percent always wear them, 22 percent sometimes wear them, 8 percent rarely wear them, and 8
percent never wear them. See Tables 24 and 25.
Overall, 10 percent of adult bicyclists wear helmets all of the time, 5 percent wear
them some of the time, and the remainder wear them rarely or never. There were no
substantial differences in helmet usage rates by age group. However, there were some
differences by gender. Although 13 percent of males always wear helmets, only 7 percent of
females do. A large disparity also exists between the cluster groups. See Figure C. Slightly
more than 50 percent of enthusiasts indicate that they always wear helmets, as compared with
6 percent of infrequent riders, 13 percent of casual riders, and 31 percent of moving-up riders.
Helmet use also varies by region. The Pacific region has the highest rate of helmet use with
23.5 percent using helmets some or all of the time. The South Atlantic region has the lowest
helmet usage rate of 10 percent some or all of the time. See Tables 26 through 29.
Table 30 shows helmet use by places ridden most often. Helmet use is lowest on
neighborhood streets, where 7 percent always wear helmets. Helmet use rises to about 19-24
112

percent on other streets and highways, and is highest on off-road trails, where 39 percent of
riders always wear helmets.
Besides helmet use, certain other behaviors (such as regard for state traffic laws and
headphone use while cycling) may reflect attitudes towards bicycling safety. Approximately 9
percent of adult cyclists (about three million cyclists) wear headphones at least occasionally
while riding their bicycles2. This rate is highest (15 percent) among the moving-up category of
cyclists. See Table 31. In addition, about one-third or over 10 million cyclists report that they
do not always adhere to traffic laws3. See Table 32.
About 9 percent of riders report that they had crashed or fallen off of their bicycles
within 12 months of the survey. Those who ride the most experience more falls and crashes.
Enthusiasts experience more than four times the incidence of falls and crashes than the
infrequent riders do4. See Table 33. However, those who wear headphones or do not always
obey traffic laws are not significantly more likely to have crashed or fallen off of their bicycles.5

Table 34 contains information concerning bicycle safety equipment used within a year
of the survey. Reflectors have the highest rate of usage; over 50 percent of bicyclists have
them. However, this is probably an underestimate since the CPSC mandatory standard for
bicycles requires reflectors on all bicycles sold in the U.S. Headlights are used by 14.5 percent
of bicyclists and taillights are used by 6.6 percent. Reflective clothing is worn by 11 percent of
cyclists; the highest usage rate is among the enthusiasts (34 percent). Seven percent of
bicyclists use whistles, bells, or horns.
Bicycles, Equipment in Use, and Future Purchase Plans
The survey gathered detailed marketing information about bicycles and bicycle
equipment owned by respondents and their households, as well as their future purchase plans.
Results are shown in Tables 35 through 48.

2

Headphone use is defined as a "yes" answer to the question, "Do you ever go
bicycling with headphones on both ears?"
3

This is probably an underestimate. Previous studies indicate a much higher noncompliance with traffic laws. See, Tinkaus, J., Stop Light Compliance by Cyclists: An
Information Look, Perceptual and Motor Skills (61), 814; 1985.
4

A chi-square test for independence concludes that membership in cluster groups
and accident experience are not independent of one another (P2=30.4, p<.01).
5

A chi-square test found that the incidence of falling or crashing was independent
of headphone use (P2=3.39, p>.05) and adherence to traffic laws (P2=6.07, p>.05).
113

Adult bicyclists own an estimated 36.5 million bicycles for personal use which were
purchased new and used. Most bicyclists (76 percent) own one bike. However, almost half of
the enthusiasts own more than one bicycle for personal use. About 67 percent of riders report
that they used general purpose bicycles. Touring bikes and mountain bikes reportedly are
owned by 15.5 and 14 percent of riders, respectively. Racing bikes are owned by 7 percent.
Less than 1 percent of bicyclists own tandem, custom or folding bicycles. Categorized by
speeds, 10 speed bicycles are owned most commonly (54.6 percent) with the next most
popular being 3-speed (16.3 percent) and 1-speed (12.2 percent). Other speeds owned are 15
(4.0 percent), 12 (12.2 percent), 5 (4.8 percent), and other (8.6 percent).6
Slightly more than half (55 percent) of the most recently purchased bicycles were
bought within five years of the survey. In contrast, 83 percent of the bicycles in the CPSC
exposure survey were less than five years old. This difference, close to 30 percent, may be
because the CPSC survey included bicycles for children, which may be purchased more
frequently due to children's growth.
Of the 18.5 percent of riders who expect to purchase a new bicycle within two years,
the mean expected outlay is $334. This represents a 82 percent nominal increase over the
mean price paid by consumers who purchased a bicycle within a year of the survey ($183). By
cluster, enthusiasts expect to pay the most for their next bicycle, $913. The moving-up
bicyclists expect to pay $647. Casual and infrequent bicyclists plan to pay $363 and $224,
respectively. See Figure D. The primary reason cited for purchasing a bicycle was
'replacement' (51.4 percent). Other reasons cited included: 'additional bicycle for self' (28.9
percent) and 'additional bicycle for someone else' (13.7 percent).
One trend that may be deduced from the survey is that mountain bikes are becoming
increasingly popular across all levels of participation. Approximately 44 percent of bicyclists
planning to purchase a new bicycle within two years plan to purchase a mountain bike, an
increase from 14 percent of those bicycles purchased most recently. In contrast, 28 percent of
bicyclists said they plan to buy general purpose bicycles, a decrease from 60 percent of bicycles
purchased most recently. Another 14 percent plan to purchase touring bikes. Most riders (75
percent) who plan to purchase a new bicycle within two years expect to purchase a bicycle
with 10 or more speeds. This is not a substantial change from previous purchases.
Table 49 lists accessories owned and plans for accessory purchases. While 16.4
percent of riders own helmets, another 13.8 percent of riders plan to purchase a helmet within
two years. Of those who plan to purchase a new helmet, 4.7 percent already own helmets.
Therefore, if the plans materialize, an additional 9.1 percent will possibly have acquired
helmets by 1992. This would bring the total helmet ownership to over 25 percent. Also
noteworthy, 14.5 percent of adult bicyclists own child seat carriers.

6

Multiple responses were accepted for owners of more than one bicycle.
114

Problems with Equipment
When asked about current mechanical problems with their bicycles, close to 60
percent of bicyclists indicate that they do not have problems. However, of those who do, 12
percent have problems with shifters, 9 percent with rubbing or dragging brakes, and 8 percent
with stopping. Bicycles that squeak or need lubrication cause problems for 8 percent of
bicyclists. Chains falling off and jamming cause problems for 3 percent of bicyclists. Wobbly
wheels are a problem for 3 percent of cyclists. See Table 50.
Summary and Conclusions
This report summarizes findings of a major survey of adult bicyclists sponsored by
Rodale Press and conducted by NFO. The 1990 survey provides a substantial amount of
information on the demographic characteristics, equipment preferences, and riding patterns of
these bicyclists. Based on the survey results, there are an estimated 32.8 million adults
bicyclists who bought their last bicycle new. These bicyclists have higher household incomes
and education levels than the U.S. population as a whole. They ride predominantly in the
warmer months, and average about 34 miles per warm weather month.
These adult riders own approximately 36.5 million bicycles for personal use. In
addition, an estimated 18.5 percent intend to purchase another bicycle within two years.
Mountain bikes are becoming increasingly popular: 44 percent who say they plan to purchase
a new bicycle within two years expect to purchase a mountain bike.
The survey also provides information about helmet use and other safety related
behavior. Most notable only about 16 percent of bicyclists own helmets, and a smaller
proportion, approximately 10 to 15 percent, wear them all or some of the time. However, 9
percent of riders who don't already own a helmet expect to purchase one within two years.

115

Tables

Table 1. Clustering Characteristics: Median Values of Responses to
the Initial Screening of Adult Bicyclists, by Cluster Group
Enthusiast
No. of Shop
Visits/Yr

Moving Up

Casual

Infrequent

3

2

1

0

Miles/Warm
Month

128

37

15

10

Price Paid
for Last
Bicycle

$642

$369

$213

$122

116

Rider Demographics

Table 2.

Rider Age, by Cluster Group

Years

Enthusiast Moving Up Casual Infrequent
Total
(%)
(%)
(%)
(%)
(%)
18-30
38.8
38.9
38.6
32.7
34.6
31-40
36.1
36.3
35.1
33.8
34.3
41-50
16.5
15.8
15.3
17.4
16.9
51-60
6.0
5.9
6.9
8.5
7.9
61+
2.6
3.1
4.1
7.5
6.4
-----------------------------------------------------------Mean
35.1
35.1
35.3
37.9
37.1

Table 3.

Male
Female

Table 4.

Gender, by Cluster Group
Enthusiast
(%)
76.8
23.2

Moving Up
(%)
67.4
32.6

Infrequent
(%)
44.9
55.1

Total
(%)
50.6
49.4

Marital Status, by Cluster Group

Enthusiast
(%)
Married
48.5
Single
40.8
Widowed
0.6
Sep/Divorced
9.5
Unknown
0.6

Table 5.

Casual
(%)
54.5
45.5

Moving Up
(%)
50.3
39.2
0.9
9.0
0.6

Casual
(%)
61.1
30.0
1.1
7.8
0.0

Infreq.
(%)
66.0
23.8
1.7
8.2
0.4

Total
(%)
63.4
26.7
1.5
8.2
0.3

Educational Level, by Cluster Group
Enthusiast
(%)

High school
or less
Some college
College deg.
Unknown

13.2
33.0
53.8
0.3

Moving Up
(%)
10.5
37.0
52.5
0.5

Casual
(%)

Infreq.
(%)

15.2
36.3
48.5
0.1

26.6
38.4
34.9
0.7

117

Total
(%)
22.6
37.6
39.3
0.5

1990
Census
(%)
58.4
18.4
23.2
--

Table 6.

Employment Status, by Cluster Group

Enthusiast
(%)
Employed
88.9
Not employed 11.1
Table 7:

Managerial/
Professional 46.3
Technical/
Sales/
Admin Support 22.7
Service
9.2
Farming/
Forestry/
Fishing
0.4
Craft/Repair
6.6
Operator/
Laborer
3.9
Other
6.8
Unknown
4.1

< 15
15-29
30-44
45-59
60 +
Unknown

Casual
(%)
83.3
16.7

Infreq.
(%)
76.2
23.8

Total
(%)
78.8
21.2

Occupation of Employed, by Cluster Group
Enthusiast
(%)

Table 8.

Moving Up
(%)
88.6
13.4

Moving Up
(%)

Casual
(%)

Infrequent
(%)

Total
(%)

46.6

42.5

37.2

39.4

21.5
8.3

27.1
8.2

27.2
11.7

26.6
10.6

0.8
7.0

0.8
6.8

1.0
6.1

0.9
6.3

4.5
5.7
5.6

5.5
4.1
5.0

7.3
3.6
5.9

6.6
4.0
5.7

Personal Employment Income (in $1,000s), by Cluster Group
Enthusiast
(%)
13.7
26.4
27.3
15.1
14.1
4.9

Moving Up
(%)
13.8
27.8
26.1
14.2
12.2
3.7

Casual
(%)
12.8
31.4
29.1
11.5
10.6
5.9

Infrequent
(%)
21.2
31.3
24.7
10.2
7.5
4.5

Total
(%)
18.6
31.0
25.9
10.9
8.8
5.0

Table 9. Status of those Not Employed, by Cluster Group

Retired
Disabled
Unemployed
Homemaker
Student

Enthusiast
(%)
23.8
9.4
18.0
24.2
33.7

Moving Up
(%)
22.6
4.3
20.4
28.8
31.1

Casual
(%)
21.3
3.2
9.3
45.7
25.5
118

Infrequent
(%)
29.4
4.7
14.2
45.9
14.4

Total
(%)
27.7
4.5
13.7
44.7
17.3

Unknown

--

0.5

3.5

119

1.8

2.0

Household Demographics

Table 10. Geographic Region, by Cluster Group
Enthusiast
(%)
New England
7.3
Mid Atlantic
13.4
East North Central
14.2
West North Central
5.8
South Atlantic
10.4
East South Central
1.6
West South Central
7.3
Mountain
12.1
Pacific
27.9

Table 11.

Casual
(%)
7.4
15.4
22.1
8.6
12.9
2.3
6.0
6.6
18.6

Infreq.
(%)
4.4
14.8
23.9
8.5
16.2
3.8
10.1
5.6
12.7

Total
(%)
5.4
14.7
22.7
8.3
14.9
3.3
8.9
6.3
15.5

1990
Census
(%)
5.4
15.1
17.0
7.3
17.9
6.2
10.5
5.5
15.1

Household Income (in $1,000s), by Cluster Group
Enthusiast
(%)
6.1
18.6
26.2
17.2
32.1

< 15
15-29
30-44
45-59
$ 60

Table 12.

Moving Up
(%)
8.5
12.0
16.7
6.2
10.2
2.5
5.7
10.1
28.2

Moving Up
(%)
7.3
16.9
27.0
19.6
29.0

Casual
(%)
7.3
17.8
22.1
19.9
32.9

Infreq.
(%)
9.7
20.8
30.5
22.4
16.6

Regional Population Density, All Riders

<50,000
50,000-499,000
500,000-1,999,999
2,000,000 +

Percent
22.1
21.4
16.4
40.0

120

All
Riders
(%)
9.0
19.8
28.3
21.5
21.3

1990
Census
(%)
24.4
25.7
20.3
12.7
16.9

Table 13.

Number of Persons in Household

One
Two
Three
Four
Five or more

Table 14.

All
Riders
(%)
12.3
28.7
21.9
22.9
14.2

Type of Residence, All Riders
Percent

House
Apartment
Mobile Home
Condominium
Twinplex
Other
Unknown

Table 15.

1990
Census
(%)
24.5
32.3
17.3
15.5
10.4

78.6
10.9
3.1
2.8
2.5
1.4
0.6

Home Ownership, All Riders

Own or buying
Rent
Live w/relative
Other
Unknown

Percent
74.7
18.7
2.7
1.2
2.7

121

Bicycle Riding Habits

Table 16.

Frequency of Warm Weather Riding, by Cluster Group

Enthusiast
(%)
60
(%)
11.0
2.3
0.6
81.1
5.0

Total
(%)
10.2
4.8
4.3
79.3
1.4

128

Table 28.

Helmet Use, by Gender
Males
(%)
13.2
5.2
4.1
76.0
1.4

Always
Sometimes
Rarely
Never
Unknown

Table 29.

Always
(%)
16.9
12.8
7.8
6.2
8.1
12.8
6.7
9.1
15.2

Yes
No

Sometimes
(%)
2.9
6.2
3.6
5.2
1.9
6.1
4.2
4.7
8.3

Rarely
(%)
4.1
4.4
4.3
2.2
4.5
2.9
2.3
9.0
4.7

Never
(%)
74.5
73.1
83.2
84.2
84.7
78.2
86.0
76.6
70.7

Unknown
(%)
1.6
3.5
1.1
2.2
0.8
--0.9
0.6
1.1

Never
(%)
83.4
63.3
72.2
67.8
76.9
80.2
42.9
79.1

Unknown
(%)
1.2
2.2
2.0
2.1
--0.7
--0.2

Helmet Use, by Places Ridden Most Often

Always
(%)
Neighborhood Streets
7.4
Quiet Sts/Quiet Hwys
24.3
Busy Sts/Busy Hwys
19.1
Bike Paths/Rail Trails 10.6
Bike Routes
13.9
Rural Roads
12.0
Off-Road Trails
39.1
Other
7.4

Table 31.

Total
(%)
10.2
4.8
4.3
79.3
1.4

Helmet Use, by Geographic Region

New England
Mid Atlantic
East North Central
West North Central
South Atlantic
East South Central
West South Central
Mountain
Pacific

Table 30.

Females
(%)
7.3
4.3
4.4
82.6
1.3

Sometimes
(%)
3.6
8.0
3.3
12.6
3.6
4.0
12.9
8.5

Rarely
(%)
4.3
2.1
3.5
7.0
5.6
3.2
5.1
4.8

Headphone Use While Riding Bicycle, by Cluster Group
Enthusiast
(%)
11.0
88.3

Moving Up
(%)
15.1
84.0
129

Casual
(%)
12.6
85.9

Infrequent
(%)
7.6
90.8

Total
(%)
9.2
89.3

Unknown

Table 32.

Always
Sometimes
Rarely
Never
Unknown

0.7

0.9

1.5

1.6

1.5

Adherence to State Traffic Laws, by Cluster Group
Enthusiast
(%)
62.8
34.6
1.1
0.6
1.0

Moving Up
(%)
59.3
35.0
3.6
0.7
1.3

Casual
(%)
61.5
33.8
1.7
1.0
2.0

Infrequent
(%)
66.4
28.6
2.1
0.7
2.2

Total
(%)
64.8
30.3
2.1
0.8
2.0

Table 33. Incidence of Falls or Crashes Within Past 12 Months,
by Cluster Group

Yes
No
Unknown

Enthusiast
(%)
31.6
67.2
1.1

Moving Up
(%)
21.9
77.5
0.6

Casual
(%)
7.6
91.7
0.6

Infrequent
(%)
6.9
92.2
0.9

Total
(%)
8.8
90.4
0.8

Table 34. Bicycle Safety Equipment Used, by Cluster Group
(Multiple Response)
Enthusiast
(%)
Rear reflector
59.1
Pedal reflector
56.5
Wheel reflector
43.6
Reflective
clothing
34.0
Whistle, horn
or bell
7.8
Head light
23.0
Tail light
10.7
Unknown
8.6

Moving Up
(%)
64.9
66.8
56.9

Casual
(%)
63.9
63.4
58.7

Infrequent
(%)
55.6
53.0
47.1

Total
(%)
58.1
56.3
50.1

21.1

15.9

8.2

11.4

8.4
16.1
7.2
12.2

7.4
17.1
6.6
20.4

7.1
13.1
6.3
30.4

7.3
14.5
6.6
26.4

130

Respondent Bicycle Ownership
Table 35.

Bicycles Owned for Personal Use, by Cluster Group

None
One
18.5
19.2
Three or more
Unknown

Enthusiast
(%)
0.5
53.5
Two

Moving Up
(%)
0.4
65.7

Casual
(%)
1.3
77.2
32.0

13.9
--

4.7
1.8

3.1
1.3

Infrequent
Total
(%)
(%)
1.0
1.0
77.3
75.8
27.5
17.2
2.3
0.9

2.9
1.0

Table 36. Number of Bicycles Currently Owned Purchased New
for Personal Use, by Cluster Group

None
One
Two
Three or more
Unknown

Enthusiast
(%)
4.0
51.2
29.9
14.0
1.0

Moving Up
(%)
2.6
64.6
26.8
4.6
1.3

Casual
(%)
4.8
72.3
17.1
4.8
1.0

Infrequent
(%)
6.5
70.9
18.9
3.2
0.5

Total
(%)
5.8
70.2
19.4
3.9
0.7

Table 37. Types of Bicycles Owned for Personal Use, by Cluster Group
(Multiple Response)
Enthusiast
(%)
General Purpose
17.7
Mountain
43.2
Touring
33.3
Racing/Triatholon
38.5
Tandem
2.7
Custom Made
9.4
Folding
0.7
Other
4.5
Unknown
0.1

Moving Up
(%)
21.7
43.4
33.2
20.3
0.8
1.1
0.8
2.7
1.1

Casual
(%)
49.0
21.5
25.1
8.3
0.1
0.9
1.3
2.6
0.4

Infrequent
(%)
79.8
7.5
10.1
4.5
0.4
0.2
0.4
1.2
0.4

Total
(%)
67.5
14.0
15.5
7.3
0.4
0.7
0.7
1.7
0.4

Table 38. Speeds of Bicycles Owned, All Riders (Multiple Response)
(Multiple Response)
Percent
One
12.2
Three
16.3
Five
4.8
Ten
54.6
131

Twelve
Fifteen
Other
Unknown

11.8
4.0
8.6
0.3

132

Information about Last Bicycle Purchased
Table 39.

Time of Acquisition, by Cluster Group
Enthusiast Moving Up Casual Infrequent
(%)
(%)
(%)
(%)
< 12 months
20.6
16.2
10.0
8.3
1-2 years ago
35.2
33.0
21.0
21.7
3-4 years ago
24.9
25.5
25.5
21.5
5 or more years ago 18.4
24.6
42.6
48.3
Unknown
0.8
0.6
0.9
0.2
Price Paid, by Cluster Group
Enthusiast Moving Up Casual
(%)
(%)
(%)
0-$99
1.5
0.5
2.2
$100-$199
26.8
40.3
41.0
$200-$299
7.1
13.4
38.6
$300-$399
9.8
33.4
16.8
$400 or more
69.9
40.3
5.1
Unknown
5.8
6.2
10.4

Total
(%)
9.6
22.7
22.7
44.6
0.4

Table 40.

Mean Price Paid
Median Price Paid

$666
$597

$381
$356

Infrequent
(%)
30.9
5.9

$242
$220

Type of Bicycle, by Cluster Group
Enthusiast Moving Up Casual
(%)
(%)
(%)
General Purpose
8.2
13.7
43.9
Mountain
35.2
40.0
21.7
Touring
18.9
25.0
20.9
Racing/triatholon
28.5
16.2
6.9
Tandem
0.8
0.5
-Custom
4.6
0.5
0.8
Folding
0.2
1.4
1.1
Other
1.1
0.4
1.8
Unknown
0.9
0.4
0.9

Total
(%)
21.9
6.3

6.8
2.3
0.4
9.3

13.9
7.9
6.1
9.2

$124
$100

$183
$130

Table 41.

Table 42.
One
Three
Five
Ten
Twelve
Fifteen

Infrequent
(%)
71.7
7.9
10.1
3.5
0.2
0.2
0.3
0.8
1.0

Number of Bicycle Speeds, All Riders
Percent
9.4
13.5
4.2
49.6
11.4
3.5
133

Total
(%)
60.0
13.8
13.7
5.8
0.2
0.5
1.5
1.0
1.0

Other
Unknown

7.9
0.6

Future Purchase Plans

Table 43.
Within 6
Within 1
Within 2
No plans
Unknown

Plan to Purchase Bicycle within Two Years, All Riders
Percent
months
2.3
year
6.2
years
10.0
81.2
0.4

For Those Who Plan to Purchase a Bicycle within Two Years
(18.5% of respondents)
Table 44. Plan to Purchase New vs. Used Bicycle, All Riders
Percent
84.2
1.3
11.4
3.1

New
Used
Don't know
Unknown

Table 45.

Expected Price, by Cluster Group
Enthusiast
$913
$677

Mean
Median

Table 46.

Moving Up
$547
$406

Casual Infrequent
$354
$224
$277.5
$188

Total
$334
$245

Type of Bicycle Expected to Purchase, by Cluster Group

General purpose
Mountain
Touring
Racing

Enthusiast
(%)
7.1
41.2
19.8
19.8

Moving Up
(%)
10.2
52.6
14.9
13.7
134

Casual
(%)
18.8
53.2
18.9
5.7

Infrequent
(%)
35.4
42.0
11.3
5.9

Total
(%)
27.4
44.4
13.4
7.1

Tandem
Custom
Folding
Stationary
Other
Unknown

2.9
3.8
-0.5
1.9
2.9

2.8
1.5
-0.4
2.9
0.8

135

-0.5
0.5
0.9
0.9
0.5

0.6
-0.5
2.4
-1.9

0.8
0.4
0.4
1.8
0.6
3.7

Table 47. Number of Speeds on Bicycle Expected to Purchase,
All Riders
Percent
4.6
8.9
6.5
35.8
13.2
13.9
12.0
5.1

One
Three
Five
Ten
Twelve
Fifteen
Other
Unknown

Table 48.

Reason for Future Purchase, All Riders

First purchase
Replacement
Additional bicycle for self
Additional bicycle for someone else
Unknown

136

Percent
3.2
51.4
28.9
13.7
2.8

Bicycle Equipment

Table 49.

Accessories, All Riders (Multiple Response)

Indexed derailleur system
Helmet
Lights
Toe clips
Mirror (eyeglass/helmet)
Mirror (handlebar)
Child seat carrier
Handlebar tape
Bicycle trailer
Unknown

Table 50.

Owned
(%)
5.9
6.4
17.1
12.5
4.6
9.2
14.5
13.6
0.6
2.8

Purchase Plans
(%)
1.2
13.8
5.6
2.6
3.0
2.0
3.8
3.1
0.6
29.8

Problems With Bicycle, All Riders (Multiple Response)
Percent
60.2
15.1
2.7
8.8
7.8
11.8
3.2
7.9
3.1

No problems
Flat tire
Wheels wobble
Brakes rub or drag
Brakes do not stop bike quickly
Shifters need adjustment
Chain falls off or jams
Squeaks/needs lubrication
Unknown

137

Part VIII.
The Risk and Helmet Use Patterns of Adult Bicyclists:
An Analysis of the 1990 Rodale Press Survey
____________________
Gregory B. Rodgers, Ph.D.
Directorate for Economic Analysis
April 1993

Introduction
The CPSC purchased the results of a 1990 national survey of adult bicyclists conducted
by the Rodale Press, the publishers of Bicycling magazine, to enhance the staff analysis of
bicycle use and risk patterns. The Rodale Press survey did not gather information on bicycle
use by children, a major focus of the CPSC bicycle project. However, it did gather information
on a wide range of topics relevant to an analysis of the risk and safety behavior of adult
bicyclists.
This report uses the Rodale Press survey results to model the accident risk and safetyrelated behavior of adult bicyclists in the United States. Safety behavior on the part of
individuals is measured by the use of bicycle helmets, which has been shown to reduce both the
likelihood and severity of head injury. The analysis shows that the risk and helmet use patterns
of adult bicyclists are predictable: they are related to personal rider characteristics, bicycle use
patterns, and demographic factors. The analysis also finds that risk and helmet use patterns are
related: factors associated with higher bicycle accident risks tend to be associated with higher
expected rates of helmet use.
The Model
The risk and safety-related behavior of adult bicyclists may be represented by the
following two equation model.
(1) RISK = f(SAF-EFF, RIDER, USE, TYPE)
(2) SAF-EFF = g(RIDER, USE, TYPE, DEMO)
where,
RISK
SAF-EFF
RIDER
USE
TYPE

=
=
=
=
=

the risk of injury accident,
individual safety efforts,
the personal characteristics of bicyclists,
patterns of bicycle use,
bicycle type,
137

DEMO

=

demographic characteristics.

The first equation attempts to explain the bicycle-related injury risk (RISK) in terms of
several general explanatory variables. Individual safety efforts, represented by the variable
SAF-EFF, are assumed to affect the injury risk directly, by affecting the likelihood of accident
or the severity of injury given that an accident has occurred.
Rider use patterns (USE) and the personal characteristics of bicyclists (RIDER) are also
assumed to affect the injury risk. Risk is likely to increase with riding distances and to be
higher on certain riding surfaces, such as highways or various unsafe terrains. Risk may also
be affected by rider characteristics such as age and gender. Teenagers, as a group, tend to
exhibit risk-taking propensities (Hodgdon, Bragg, and Finn, 1981; Noe, McDonald, and
Hammit, 1983), and males may be more likely than females to take risks, as evidenced by
automobile and other accident rates (Hodgdon et al., 1981; Holinger, 1979; National Safety
Council, 1992; Rodgers, 1990).
The type of bicycle used (TYPE) may also affect the injury risk. There is no evidence
that certain bicycle model types are inherently more or less safe than others. However,
Mortimer et al. (1976) conducted performance tests designed to measure the relationship
between bicycle maneuverability and handlebar configuration. They found, in several tests,
that the maneuverability of bicycles with "high rise" and "standard" handlebars was better than
that of bicycles with the "dropped" (i.e., C Bend) handlebar configuration found on lightweight
racing style bicycles.
Individual safety efforts (SAF-EFF), an explanatory variable in equation (1), is itself
determined endogenously in equation (2). That is, safety efforts are assumed to be determined
within the model as a function of a set of explanatory variables. Although there are many types
of safety efforts that might be exhibited by bicyclists, helmet use will be taken as the measure
of safety effort in this analysis.1 As shown by Thompson et al. (1989) and Dorsch et al.
(1987), helmet use can substantially reduce both the injury risk and severity of head injury,
given that an accident has occurred.
Just as use patterns and rider characteristics affect the injury risk, they may influence
the helmet use decision. For example, riders who use their bicycles on off-road trails may wear
helmets to protect their heads from tree limbs, brush, or falls. Those who ride on highways or
other major thoroughfares may be more likely to wear helmets because of higher accident
risks, or because of the likely greater severity of injury in accidents involving collisions with
motor vehicles.

1

Other types might include, for example, being careful to stop at stop-signs or being
cognizant of surrounding automobile traffic. However, information on these types of safety
efforts are difficult to obtain from telephone or mail surveys.
138

Demographic characteristics are also likely to affect individual safety efforts, such as
helmet use. Under the assumption that safety is a normal economic good, helmet use should
increase with household income. Riders in households with higher education levels may also
wear helmets more frequently since they may be more aware of, or more capable of evaluating,
the benefits of helmet use.
Data
Data for the analysis are from a 1990 survey of adult bicyclists conducted by National
Family Opinion, Inc. (NFO) from its national consumer mail panel. The NFO consumer panel
is not a probability sample, but is balanced to match U.S. statistics on five demographic
variables: geographic region, population density, household income, household size, and age of
panel member.2
The survey, which was sponsored by Rodale Press, the publishers of Bicycling
magazine, elicited information on a wide range of topics relevant to bicycle riding. Topics
included, among others, riding patterns and habits (e.g. riding distances and where bicycles are
ridden), the physical characteristics of riders (e.g. age and gender), the types of bicycles used,
future purchase plans, bicycling safety (e.g. information on accidents and helmet use), and rider
and household demographic characteristics.3
Screening questionnaires were initially sent to a sample of 150,000 panel households
from around the nation. About 104,000 households (69 percent) responded and about 40,300
indicated that they had at least one adult bicycle owner. In total, there were about 61,600
individual adult bicycle owners in these households.
To qualify for the survey, respondents had to be age 18 or older, they had to own a
bicycle intended for an adult, and they had to have purchased their last bicycle new. Since a
major goal of the survey was to gather information of interest to the bicycle industry,
qualifying respondents also had to answer three questions on the screener. The questions
included: 1) the number of miles they ride in an average warm month; 2) the number of times
they visited a bike shop in the last year; and, 3) the price they paid for the last bicycle
purchased. About 37,900 respondents, out of the 61,600 individual bicycle owners (61
percent), qualified. The respondents who did not qualify were, for the most part, those who

2

The NFO panel has also been used by the CPSC in the analysis of hazards associated with
ATVs.
3

For a detailed description of the survey methodology and results, see Donaldson (1993),
in Part VII of this report.
139

had purchased their last bicycle used or did not answer one or more of the questions on miles
ridden, bike shop visits, or price paid (Rodale Press, 1990).4
NFO separated qualifying respondents into four groups by means of a cluster analysis
applied to the screening questions. The cluster analysis was intended to identify groups of
bicyclists who ranged from those who hardly ever ride to those who might be considered
biking "enthusiasts." In-depth questionnaires were subsequently mailed to about 4,200
households, including at least 1,000 households from each of the four clusters. Each cluster
sample was selected by NFO to be representative of the households in the cluster.
Questionnaires were returned from 3,248 respondents, for a response rate of about 77 percent.

The purpose of sampling the various clusters was to provide representative samples of
bicyclists in several "marketing" categories. However, NFO also provided weights for the
observations from each of the clusters so that the entire sample could be used to make
projections representative of all U.S. adult bicyclists.
Statistical Analysis
Analytic Techniques
The risk and helmet use functions are estimated in reduced form with probit regression
models.5 Probit analysis, like multiple regression analysis, is a statistical procedure in which
variation in the dependent variable is explained by variation in the explanatory variables. The
multiple regression approach allows us to determine and quantify the factors associated with
changes in the dependent variable and to sort out the potentially complex interrelationships
between the dependent and independent variables.
The probit specification of the regression model is used to examine the relationship
between the independent variables and a dependent variable that represents two distinct
alternatives (Pindyck and Rubinfeld, 1991). The dependent variable in the risk model
(equation 1) represents whether or not the respondents had "crashed or fallen off [their]
bicycle" in the 12 month time span prior to the survey. It is set equal to one if there had been a

4

Out of the 61,600 adult bicyclists, about 20 percent purchased their last bicycle used, and
about 18 percent did not answer all of the questions.
5

Estimating equation (1) directly, with the endogenous explanatory variable on the right
hand side of the equation, produces inconsistent estimates of the parameters (Pindyck and
Rubinfeld, 1991). However, it can be estimated consistently in reduced form, an approach
which is sufficient for our purposes because we are primarily interested in the net effects of the
exogenous explanatory variables on the accident risk and the likelihood of helmet use.
140

crash, zero otherwise. It should also be noted that it represents the accident risk, rather than
the injury risk, since some crashes or falls may not result in physical injury. The dependent
variable in the safety efforts model (equation 2) is a dichotomous variable based on the
frequency of helmet use.6 The helmet use variable is set equal to one for bicyclists who always
wear helmets, and is set equal to zero for bicyclists who wear helmets only sometimes, rarely,
or never.7
The specific independent variables to be used in the analysis are defined in Table 1.
(The tables begin on page 149.)
Results
Tables 2 and 3 present, respectively, the accident risk and the helmet use regression
models. In the first model of each table rider age is entered as a series of dummy variables. In
the second, age is entered as a continuous variable. Since the equations were estimated in
reduced form, the explanatory variables in both the risk and helmet use models are the same.
All of the equations in Tables 2 and 3 are statistically significant. About 15 percent of
the observations were lost because of missing information on the variables. A sensitivity
analysis, conducted by replacing missing values with the mean value of the variable in question
(Pindyck and Rubinfeld, 1991), indicated that the models were not substantially affected by the
missing data.
Accident Risk. First consider the risk models of Table 2. The results show several
strong relationships between the accident risk and rider use patterns. As expected, risk
increases with miles ridden per month. This is indicated by the positive and significant
coefficient for the variable MILES.
The relationship between risk and riding surface is measured with a series of dummy
variables representing the various riding surface types, relative to neighborhood streets. The
accident risk increases for bicyclists on off-road trails (TRAILS), and decreases for riders on
bike paths (BIKEPATH).

6

Helmets are intended to reduce the likelihood or severity of injury, given an accident,
rather than the accident risk itself. Nevertheless, the structural equation for the accident risk
should include safety efforts as an explanatory variable (as in equation 1) since helmet use is
but one type of safety effort: other types are likely to affect the risk of accident.
7

Alternatively, the helmet use variable could have been set equal to one for bicyclists who
always or sometimes wear a helmet. The results of the analysis were, however, virtually
identical when "sometime" wearers (about 4 percent) were included in the category of helmet
users.
141

Somewhat surprisingly, riding on highways (HIGHWAY) has no independent impact on
the accident risk. However, this may be because responses to the survey question on riding
location combined highways with streets (i.e., busy highways/busy streets and quiet
highways/quiet streets), and may therefore have diluted distinctions based on traffic volume.
The relationship may also have been confounded by the significant correlation between the
highway and distance variables (r=0.21, p<0.0001).8 Riding on rural roads (RURAL) has no
independent impact on the accident risk.
The accident risk is also affected by bicycle type. Risk is higher for riders of both allterrain (ATB) and racing style (RACING) bicycles, relative to the more general purpose
models. This result does not necessarily imply that these types of bicycles are more dangerous
than general purpose models, and probably reflects in part the relationship between risk and
riding patterns not picked up by the other use pattern variables.9
While there is no independent statistical relationship between gender (GENDER) and
risk, the accident risk is related to age. Age is entered into Model 2 as a quadratic variable
(AGE and AGESQ), and captures the apparently nonlinear relationship between age and risk;
that is, risk initially declines with age, but then rises for older riders.10 Based on the results
from Model 1, the risk for riders over age 64 is significantly higher than for riders 25-64 years
of age.
While the higher risk for the younger adults was anticipated and may be related to risktaking propensities, the reason for the increased risk of older bicyclists is unclear. It might be
explained by a deterioration in reaction time, a characteristic that is important in avoiding
accidents. Maring and van Schagen (1990), who also found an increased accident risk for
bicyclists over the age of 60 in the Netherlands, suggested that the higher risk might be related
to changing cognitive and perceptual processes that tend to reduce the flexibility of older riders
in responding to unforeseen situations. Older persons involved in accidents may also tend to
suffer adverse outcomes because of medical complications and other factors associated with
postinjury homeostasis.

8

It is also possible that some individuals ride more carefully on highways than on
neighborhood streets because of the potentially greater injury severity that might be expected
in highway accidents.
9

RACING was significantly correlated with riding on highways, and ATB was significantly
correlated with riding on trails. In addition, both the RACING and ATB variables were
significantly correlated with MILES.
10

A similar relationship was found between age and the fatality risk for drivers of allterrain vehicles (Rodgers, 1990).
142

There is no evidence that the accident risk is directly correlated with rider demographic
characteristics such as household income (INCOME) and education level (EDUC). However,
there are apparently some regional variations in the accident risk, with lower risks in the
Northeast and Mid-Atlantic States (represented by the variable EAST) and higher risks in the
Pacific Coast States.
Helmet Use. The likelihood of helmet use is affected by a number of factors that also
affect the accident risk. Helmet use increases with greater monthly riding distances, and is
higher for bicyclists who ride most often on off-road trails. The coefficient for the bike path
variable (BIKEPATH) is negative (as in the accident risk models), but not significant at the
usual 5 percent significance level (p=0.11). The impact of riding on rural roads on helmet use
patterns is also nonsignificant.
In contrast to the nonsignificant relationship between risk and the HIGHWAY variable
in Table 2, helmet use is significantly higher for bicyclists who ride on highways. This may
suggest that bicyclists who ride primarily on highways are more likely to wear helmets because
of the potentially greater injury severity that might be expected in highway accidents involving
motor vehicles (rather than simply because of a higher accident risk), a factor that would not
necessarily be picked up in the risk equation.
Helmet use, like the accident risk, is also affected by bicycle type. Riders are more
likely to wear helmets if they ride racing or all-terrain bikes. On the other hand, the
coefficients for the age and gender variables are both nonsignificant, suggesting that age and
gender have no independent effect on the likelihood of helmet use.
In contrast to the accident risk models, rider demographic characteristics have a
substantial impact upon helmet use patterns. As expected, helmet use increases with
household income and with rider education. There are also some regional variations in helmet
use: bicyclists in the Central, Southern, and Mountain States were less likely to wear helmets
than bicyclists in the Pacific Coast States.
Risk and Helmet Use Estimates
This section examines the sensitivity of changes in the accident risk and the likelihood
of helmet use to changes in the independent variables. One way to do this is to estimate and
compare differences in the average accident and helmet use probabilities for various population
subgroups, such as female bicyclists or bicyclists between the ages of 18 and 24. Although
such estimates do not statistically hold other variables constant, they provide consistent group
estimates of the proportion of individuals who are likely to have accidents and wear helmets
(Train, 1986).
The expected accident and helmet use probabilities for selected population subgroups
are presented in Table 4. The expected accident probabilities are based on Model 2 of Table 2;
143

the helmet use probabilities are based on Model 2 of Table 3. A major pattern that emerges is
that differences in the average accident risk from one population subgroup to another tend to
be matched by similar changes in the average likelihood of helmet use.11
A number of the results follow directly from the discussion of regression results. Both
the average accident risk and the expected rate of helmet use are higher for bicyclists who ride
greater distances, for bicyclists who ride most often on trails, and for bicyclists who use racing
and ATB style bicycles. On the other hand, the expected risk and helmet use rates are lower
for bicyclists who ride most often on neighborhood streets.
Notwithstanding the nonsignificant regression findings for the gender variable, the
expected risk and helmet use rates are significantly higher for male riders. In addition, the
accident risk (as well as the expected rate of helmet use) is higher for bicyclists who ride most
often on highways. This is because males and those who ride on highways ride longer average
distances than females and others who do not often ride on highways, and distance is positively
related to risk and helmet use. Males ride about twice the distance of females (47 versus 24
miles in an average warm weather month). Similarly, bicyclists who ride most often on
highways ride about 2.5 times the distances of other bicyclists (76 versus 31 miles per month).
Finally, as indicated by the regression results, helmet use varies systematically with
income and education. The expected rate of helmet use increases monotonically from about
7.1 percent of those bicyclists with household incomes of less than $15,000 per year to almost
18 percent for those with household incomes of $100,000 or more per year. Similarly,
expected helmet use rates increase from 4.4 percent of bicyclists with a high school education
or less to almost 17 percent for college graduates.
Discussion
This study examined the risk- and safety-related behavior of adult bicyclists, based on
the results of a national cross-section survey conducted by Rodale Press in 1990. The results
show that the accident risk and helmet use patterns of adult bicyclists are predictable and
depend upon the characteristics and riding patterns of bicyclists.
The expected accident risk for adult bicyclists increases with greater riding distances.
Risk also varies by riding surface. Relative to riding on neighborhood streets, riding primarily
on off-road trails increases risk. In contrast, riding primarily on bike paths lowers risk. There
was no evidence that riding on highways independently increases the accident risk, relative to
11

For example, when comparing bicyclists who ride less than 11 miles per month with
those who ride more than 50 miles per month, the average expected accident risk rises from
4.5 to 22.5 percent, and the expected rate of helmet use rises from 3.7 to 30.4 percent.
Similarly, the average expected accident risk and helmet use rates rise from 7.1 and 7.3
percent, respectively, for females, to 12.0 and 13.8 percent, respectively, for males.
144

riding on neighborhood streets, but this finding may have been due to combining the highway
variable with other "streets." Risk is also related to rider characteristics. Risk declines with
age, but then increases for riders over the age of 64. Although rider gender has no
independent effect on risk when other factors are held constant, the average risk is higher for
males because, on average, they ride greater distances than females.
In addition, factors associated with higher bicycle accident risks tend to be associated
with higher expected rates of helmet use, and factors associated with lower bicycle accident
risks tend to be associated with lower expected rates of helmet use. For example, the expected
rate of helmet use increases with riding distances, tends to be higher for those who ride most
often on trails or highways, and tends to be lower for those who ride most often on
neighborhood streets and bike paths. Neither rider age nor rider gender has an independent
effect on the likelihood of helmet use. However, the average rate of helmet use is higher for
males, who on average ride greater distances than females. In addition, as anticipated, helmet
usage rates are affected by demographic characteristics -- helmet use increases with rider
education levels and household income.
These results are generally consistent with a theory of compensatory behavior in risky
activities, which has been discussed by Peltzman (1975) and others.12 This theory hypothesizes
that in familiar risky activities, such as driving, individuals tend to compensate for changes in
the risk environment. Individuals are expected to increase safety efforts in response to
exogenous increases in risk, and to reduce safety efforts in response to reductions in risk.13
The finding, for adults, that higher (lower) rates of helmet use tend to be associated
with higher (lower) accident risks does not imply that helmet usage rates are high enough.
Individuals may systematically underestimate the value of helmet use and therefore use them
less than they should. The 1991 CPSC bicycle exposure survey found, for example, that
almost half of the bicyclists who never wear helmets said they never thought about doing so
(Rodgers, 1992b). Nevertheless, the results suggest that adult bicyclists tend to increase
helmet use when they perceive that they are at greater risk.14 Providing the public with

12

See, e.g., O'Neill, 1977; Blomquist, 1986, 1988; Evans, 1985; Orr, 1982; and, Viscusi,
1984. For a non-technical discussion of the hypothesis, see Rodgers (1992a). For examples of
recent empirical analysis involving the behavior of automobile drivers and motorcyclists, see
Blomquist 1991; Crandall and Graham, 1984; Graham, 1984; McCarthy, 1986; Evans and
Graham, 1991; Winston, 1987; and Graham and Lee, 1986.
13

As an example, drivers might be expected to increase seat belt use when driving in
rainstorms or on congested highways.
14

In fact, the Rodale Press survey results also indicate that helmet usage rates have
increased substantially in the last couple of years; more than half of current helmet users
reported that they began wearing helmets in the last two years. This may suggest that the
145

information describing the bicycle-related accident risks and advantages of helmet use in
reducing head injuries may therefore be an effective strategy in efforts to reduce injuries to
adult bicyclists.

References
Beyers, Dan. Helmet Laws Gain a Track Record. The Washington Post, December 13, 1992,
p. B1.
Blomquist, Glenn C. A Utility Maximization Model of Driver Traffic Safety Behavior.
Accident Analysis and Prevention 18(5), October 1986, 371-5.
Blomquist, Glenn. The Regulation of Motor Vehicle and Traffic Safety. Boston: Kluwer;
1988.
Blomquist, Glenn C. Motorist Use of Safety Equipment: Expected Benefits or Risk
Incompetence? Journal of Risk and Uncertainty, 4, 1991, 135-152.
Crandall, Robert W., and John D. Graham. Automobile Safety Regulation and Offsetting
Behavior: Some New Empirical Estimates. American Economic Review 74(2), May 1984,
328-31.
Donaldson, Mary. Characteristics of Adult Bicyclists in the U.S.: Selected Results from a
National Survey of Bicycle Riders. U.S. Consumer Product Safety Commission, April 1993.
Dorsch, Margaret M., Alistair J. Woodward, and Ronald L. Somers. Do Bicycle Safety
Helmets Reduce Severity of Head Injury in Real Crashes? Accident Analysis and Prevention
19(3), 1987, 183-190.
Evans, Leonard. Human Behavior Feedback and Traffic Safety. Human Factors 27(5),
October 1985, 555-76.
Evans, William N., and John D. Graham. Risk Reduction or Risk Compensation? The Case of
Mandatory Safety-Belt Use Laws. Journal of Risk and Uncertainty 4(1), 1991, 61-73.
Graham, John D. Technology, Behavior, and Safety: An Empirical Study of Automobile
Occupant-Protection Regulation. Policy Sciences 17, 1984, 141-51.

demand for helmets has increased as consumers have become more aware of the safety
benefits, which have been widely publicized in the safety literature and in the media, (see, e.g.,
Beyers, 1992).
146

Graham, John D., and Younghee Lee. Behavioral Response to Safety Regulation: The Case of
Motorcycle Helmet-Wearing Legislation. Policy Sciences 19, 1986, 253-73.
Hodgdon, J.D., B.W.E. Bragg, and P. Finn. Young Driver Risk-Taking Research: The State of
the Art. Technical Report, Contract DTNH22-80-C-07360, Washington, DC: Department of
Transportation, 1981.
Holinger, P.C. Violent Deaths Among the Young: Recent Trends in Suicide, Homicide, and
Accidents. American Journal of Psychiatry 136, 1979, 1144-7.
HDR Engineering, Inc. National Bicycling and Walking Study: Interim Report. Report No.
FHWA-PD-92-003. Washington, DC: Federal Highway Administration, November 1991.
Maring, Wim, and Ingrid van Schagen. Age Dependence of Attitudes and Knowledge in
Cyclists. Accident Analysis and Prevention 22(2), April 1990, 127-136.
McCarthy, Patrick S. Seat Belt Usage Rates: A test of Peltzman's Hypothesis. Accident
Analysis and Prevention 18(5), October 1986, 425-38.
Mortimer, R.G., Patricia A. Domas, and Robert E. Dewar. The Relationship of Bicycle
Manoeuvrability (sic) to Handlebar Configuration. Applied Ergonomics 7(4), 1976, 213-9.
National Safety Council. Accident facts, 1992 Edition. Itasca, IL: Author; 1992.
Noe, P.F., C.D. McDonald, and W.E. Hammitt. Comparison of Perceived Risk Taking in
Groups and Implications Drawn from the Risky-Shift Paradigm. Perceptual and Motor Skills
56, 1983, 199-206.
O'Neill, Barry. A Decision-Theory Model of Danger Compensation. Accident Analysis and
Prevention 9, 1977, 157-65.
Orr, Lloyd D. Incentives and Efficiency in Automobile Safety Regulation. Quarterly Journal
of Economics and Business 22(3), Autumn 1982, 43-65.
Peltzman, Sam. The Effects of Automobile Safety Regulation. Journal of Political Economy
83(4), 1975, 677-725.
Pindyck, Robert S. and Danial L. Rubinfeld. Econometric models and economic forecasts.
New York: McGraw-Hill, 1991.
Rodale Press. The Cycling Consumer of the 90's: A Comphrehensive Report on the U.S.
Adult Cycling Market. Emmaus, PA: Rodale Press, 1991.

147

Rodgers, Gregory B. Evaluating Product Related Hazards at the Consumer Product Safety
Commission: The Case of All-Terrain Vehicles. Evaluation Review 14(1), February 1990, 321.
Rodgers, Gregory B. The Safety Effects of Child-Resistant Closures. U.S. Consumer Product
Safety Commission, May 1992a.
Rodgers, Gregory B. Bicycle and Bicycle Helmet Use Patterns in the United States: A
Description and Analysis of National Survey Data. U.S. Consumer Product Safety
Commission, December 1992b.
Singh, Harinder, and Mark Thayer. Impact of Seat Belt Use on Driving Behavior. Economic
Inquiry 30(4), October 1992, 649-58.
Thompson, Robert S, Frederick P. Rivara, and Diane C. Thompson. Case Control Study of
the Effectiveness of Bicycle Safety Helmets. The New England Journal of Medicine 320(21),
May 25, 1989, 1361-7.
Tinsworth, Deborah K., Curtis Polen, and Suzanne Cassidy. Bicycle-Related Injuries Treated
in U.S. Hospital Emergency Rooms. U.S. Consumer Product Safety Commission, January
1993.
Train, Kenneth. Qualitative choice analysis: Theory, econometrics, and an application to
automobile demand. Cambridge, MA: The MIT Press; 1986.
Viscusi, Kip W. The Lulling Effect: The Impact of Child-Resistant Packaging on Aspirin and
Analgesic Ingestions. American Economic Review 74(2), May 1984, 323-27.
Winston, Clifford and associates. Blind Intersection: Policy and the Automobile Industry.
Washington, DC: Brookings, 1987.

148

Table 1: Variable Definitions
________________________________________________________________
Rider Characteristics (RIDER)
-AGE
rider age;
-AGESQ
the square of rider age;
-AGE(X1-X2)
1 if rider is of the age described in the
brackets (i.e., age X1 to age X2), 0
otherwise;
-GENDER
1 if the rider is male, 0 if the rider is
female.
Use patterns (USE)
-MILES
-NEIGH
-HIGHWAY
-BIKEPATH
-RURAL
-TRAIL
-OTHER SURF.
Bike Types (TYPE)
-ATB
-RACING

The natural logarithm of the number of miles
traveled in an average warm month;
1 if the bicyclist rides most often in
neighborhood streets, 0 otherwise;
1 if the bicyclist rides most often on quiet
highways/quiet streets or busy highways/busy
streets, 0 otherwise;
1 if the bicyclist rides most often on bike
paths or bike routes, 0 otherwise;
1 if the bicyclist rides most often on rural
roads, 0 otherwise;
1 if the bicyclist rides most often on offroad trails, 0 otherwise;
1 if the bicyclist rides most often in "other
places," 0 otherwise.
1 if all-terrain or mountain bicycle, 0
otherwise;
1 if racing/triathlon bicycle, 0 otherwise;

Demographic Characteristics (DEMO)
-INCOME
Household income, in thousands of dollars;
-EDUC1
1 if the rider has a high school education or
less, 0 otherwise;
-EDUC2
1 if the rider attended college but did not
graduate, 0 otherwise;
-ECUC3
1 if the rider is a college graduate, 0
otherwise;
-EAST
1 if the rider resides in a Northeastern or
Mid-Atlantic State, 0 otherwise;
-MIDWEST
1 if the rider resides in a East Central or
West Central State, 0 otherwise;
-SOUTH
1 if the rider resides in a Southern State, 0
otherwise;

149

-MOUNTAIN
-PACIFIC

1 if the rider resides in a Mountain State, 0
otherwise;
1 if the rider resides in a Pacific Coast
State, 0 otherwise.

150

Table 2: Regression Results: Risk of Accident

Variable

Model 1:
Coefficient

INTERCEPT
MILES
GENDER
AGE
AGESQ
RACING
ATB
HIGHWAY
BIKEPATH
RURAL RD.
TRAIL
OTHER SURF.
INCOME
EDUC2
EDUC3
EAST
MIDWEST
SOUTH
MOUNTAIN
AGE (18-24)
AGE (25-34)
AGE (35-44)
AGE (45-54)
Age (55-64)

-1.7092
.2804**
.0651
--.3266**
.2950**
-.0671
-.3065*
-.0101
1.1348**
.1147
-.0005
-.1482
.1472
-.3113**
-.1927
-.2669*
-.3341*
.0637
-.3551*
-.4436**
-.4791*
-.8456**

SE
.1993
.0288
.0741
--.1239
.0935
.1095
.1243
.1127
.2404
.2385
.0013
.0985
.0977
.1156
.1045
.1098
.1607
.1711
.1605
.1647
.1894
.2495

Model 2:
Coefficient
-.4323
.2763**
.0765
-.0795**
.0008**
.3389**
.2914**
-.0668
-.3261**
-.0211
1.1387**
.1591
-.0003
-.1259
.1757
-.2981**
-.1923
-.2543*
-.3221*
------

SE
.3095
.0287
.0737
.0145
.0002
.1239
.0933
.1092
.1240
.1122
.2403
.2358
.0013
.0981
.0973
.1151
.1043
.1096
.1601
------

* significant at p < 0.05, two-tailed test.
** significant at p < 0.01, two-tailed test.
N (Accident)
N (non-Accident)
c
Score
Model Chi-square

452

452

2,499.5
.760
315.5
270.1

151

2,499
.764
313.6
264.1

Table 3: Regression Results: Likelihood of Helmet Use

Variable

Model 1:
Coefficient

INTERCEPT
MILES
GENDER
AGE
AGESQ
RACING
ATB
HIGHWAY
BIKEPATH
RURAL RD.
TRAIL
OTHER
INCOME
EDUC2
EDUC3
EAST
MIDWEST
SOUTH
MOUNTAIN
AGE (18-24)
AGE (25-34)
AGE (35-44)
AGE (45-54)
AGE (55-64)

-2.6424
.3752**
-.0186
--.6876**
.3203**
.2560*
-.1794
.1239
.9476**
-.4237
.0026*
.2039
.6866**
-.0015
-.2780*
-.2067
-.3569*
-.4064*
-.2785
-.3619*
-.2606
-.2596

SE
.2209
.0298
.0753
--.1180
.0941
.1036
.1173
.1160
.2499
.3497
.0012
.1173
.1142
.1113
.1091
.1120
.1677
.1956
.1731
.1763
.1953
.2212

Model 2:
Coefficient
-2.6739
.3746**
-.0233
-.0173
.00024
.6764**
.3157**
.2554*
-.1780
.1161
.9135**
-.4278
.0024*
.2155
.7189**
-.0061
-.2746*
-.2041
-.3570*
------

SE
.3441
.0298
.0751
.0158
.00018
.1177
.0941
.1036
.1169
.1152
.2485
.3480
.0012
.1178
.1144
.1111
.1089
.1119
.1676
------

* significant at p < 0.05, two-tailed test.
** significant at p < 0.01, two-tailed test.
N (Helmet User)
N (non-Helmet User)
c
Score
Model Chi-square

724

724

2,227

2,227

.827
489.3
448.5

152

.827
483.8
445.9

Table 4.

Average Accident Risk and Helmet Use Probabilities,
for Selected Population Subgroups

Variable
_
X

Group

Accident
Risk
%

Helmet
Use
%

% of
observations
(weighted)

All

9.5

10.6

100.0

Miles
per month

# 10
11-25
26-50
> 50

4.5
9.3
13.8
22.5

3.7
9.2
15.2
30.4

50.6
19.3
15.1
15.0

Age

18-24
25-34
35-44
45-54
55-64
$ 65

17.9
9.9
6.8
6.9
6.5
12.6

11.4
10.0
10.7
11.6
9.1
11.2

12.4
36.9
29.9
10.4
6.0
4.2

Gender

Female
Male

7.1
12.0

7.3
13.8

50.0
50.0

Surface Type Ride On Most Often:
Neigh.
Streets

No
Yes

12.0
8.2

15.4
7.9

35.6
64.4

Highway

No
Yes

9.0
14.4

9.1
22.9

89.6
10.4

Bikepath

No
Yes

9.8
7.5

10.5
11.3

89.3
10.7

Rural Rd.

No
Yes

9.5
9.9

10.4
11.5

88.6
11.4

Trail

No
Yes

9.0
53.8

10.2
43.4

98.9
1.1

Racing

No
Yes

8.9
18.7

9.2
30.2

93.6
6.4

ATB

No

8.2

9.2

85.7

Bike Type:

153

Yes

17.4

154

18.8

14.3

Table 4 (continued)
Accident
Risk
%

Helmet
Use
%

% of
observations
(weighted)

Variable

Group

Income
(in $1000s)

< $15
$15-$29.9
$30-$44.9
$45-$59.9
$60-$74.9
$75-$99.9
$ $100

9.7
10.0
9.2
8.9
9.9
10.0
10.3

7.1
8.5
9.9
10.2
12.9
16.4
17.8

7.7
19.5
29.1
21.7
9.8
6.2
4.9

School

H. Sch. or less
Some College
College Grad.

8.7
8.2
11.2

4.4
7.7
16.8

22.6
37.7
39.6

Region

East
Midwest
South
Mountain
Pacific

8.9
8.9
7.8
9.7
14.4

14.2
7.7
8.7
9.3
15.4

20.4
31.4
26.4
6.4
15.4

155