Top: John S. Allen's Home Page
Up: Table of Contents
Previous: Letter from L.A.W.
Next: Calculation of the miles/hour figure

Analysis of Bicycle Use and Hazard Patterns

John S. Allen

 I wish to thank Alan Wachtel, John Forester and Philip Graitcer for their helpful comments on draft versions of the following article. I also thank Dr. Gregory Rodgers of the U.S. Consumer Products Safety Commission for corrections which I have incorporated into this revised version. However, I take full responsibility for any errors.


The United States Consumer Products Safety Commission set an ambitious goal for its research project on bicycling: to publish a "comprehensive national study of bicycle use and hazard patterns designed to quantify riding patterns and the rider and environmental factors associated with risk" (Page 2 of the study report, final version -- all page references, except where indicated, are to this version).

The draft was circulated to bicycling organizations including the Bicycle Federation of America and the League of American Bicyclists. The League sent out copies for comment to a number of people including the author of this review.

In March, 1994, League representatives met with CPSC officials to inform them of serious flaws with the report. Nonetheless, the CPSC published the report in final form in June, 1994, with no substantial changes. It is available from the CPSC, Washington DC., 20207 and now (2001) also available on the CPSC's Web site.

The CPSC collected data primarily through telephone surveys. One, the national exposure survey, conducted in June and July of 1991 (p. 21), sampled over 4000 households.

The CPSC also purchased the rights to data from a poll conducted by the Harris organization in 1990 for Rodale Press. After correcting for different age groupings, the two studies give very similar numbers of bicyclists:

 Number of bicyclists

Age *Rodale, 1990
(p. 109 ff)
CPSC, 1991
(p. 44)
18-30 11,348,800 **13,500,000
31-40 11,250,400 12,200,000
41-50 5,543,200 5,800,000
51+ 4,690,400 4,100,000
Total ***32,832,800 35,600,000

*Includes only respondents who bought their most recent bicycle new. These were about 80 percent of respondents.

**Includes 21 to 30 year age group plus 0.5 of 15-20 year age group.

***Number is slightly different from that given in the report, due to calculation from rounded percentage figures.

As this comparison suggests, the two studies agree well with each other and with informal observation whenever conclusions are based on counting heads: for example, both studies show that avid cyclists are a greater percentage of the total on the West Coast, and a smaller percentage in the Southeast.

How Many Hours of Bicycle Use?

The CPSC also attempted to count hours of riding time, and Rodale to count bicycle miles. Let's do a comparison.

For riders 18 years of age and older, Rodale provided data on the monthly mileage of bicyclists (table 7, p. 121), and number of months ridden (Figure B, p. 133), making it possible to calculate an average yearly mileage of 192.3 miles.  Mean riding time for bicyclists 18 and older, as found by the CPSC, is 182.6 hours (p. 44, 45; ages 21+ plus 0.5 of 15-20 year age group).

Comparing these two figures, then, the mean speed for an adult American bicyclist is, then, 1.264 miles per hour.[1]

The CPSC report does not discuss this numerical discrepancy, but rather, states (p. 4) that "the Rodale Press survey findings for adult bicyclists...are generally consistent with the findings of the CPSC exposure survey."

Perhaps the CPSC deserves credit for trying to measure the amount of bicycle use nationwide, but unfortunately, the researchers' technique was not adequate to the task. I suspect that it led respondents to report an entire day's excursion or perhaps an entire vacation as "bicycling." Conducting the telephone survey in the peak cycling months of June and July likely also contributed to an inflated bicycle use figure.

Other comparisons throw light on the inaccuracy of the CPSC data. Comparing it with with 1991 USDOT data, we would count one bicycle for every five passenger cars and light trucks in the U.S. traffic mix -- or about one bicycle for three such vehicles if we exclude limited-access highways, hours of darkness and bad weather. Ah, were it only true! CPSC exposure data would make bicyclists' accident rates far lower per hour of exposure than motorists', hardly consistent with the CPSC's central justification of its involvement with bicycling -- the proposition (p. 1), highly questionable even on the basis of accurate data, that "bicycling is a very risky activity."

The CPSC repeatedly describes its data as representing riding time (for example, see p. 25 and table 5, p. 45). We read (p. 34) that "...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." However, only one person per household was interviewed, and all data on young children was collected second-hand from adults, (p. 21), subjecting it to additional and different sources of error. The problem here is emblematic of those in many other parts of the report.

The Rodale data included in the CPSC study divides riders into four "cluster groups": "Enthusiast," "Moving Up," Casual" and Infrequent." (Table 1, p. 116). Weighting the average mileage for each cluster group by the percentage of riders in each cluster group and age group gives the following results:[2]

Age Rodale miles/
CPSC hour
18-30 1.079
31-40 1.432
41-50 1.282
51+ 1.915

Cluster group criteria included bicycle purchase price and frequency of bike shop visits as well as mileage, so this analysis is far from definitive; still, the great discrepancy between CPSC and Rodale data, and its wide range, render any conclusions based on the CPSC riding time data invalid.

The Rodale mileage figures agree better with other sources than the CPSC's data, but are nonetheless unreliable in that they are based on an off-the-cuff estimate of a full year's riding, rather than observation, mileage diaries or odometer readings.

Extraordinary Conclusions About Accident Rates

Using its exposure data in combination with an accident survey, the CPSC study reaches the following conclusions, among others:

  • Children ride more (p. 24), and their fatality rate is lower than for older bicyclists (Table 1, p. 108), though their injury rate is higher (p. 7). These conclusions reflect not only the inaccuracy of the CPSC's riding time data, but also the dilution of its definition of "bicyclist" by including the many millions of Americans, mostly adults, who ride as infrequently as once per year (p. 21).

Other studies show adult bike club members riding 2300 miles per year (Kaplan, 1975), children 580 (Chlapecka et al., 1975) and casual, college-affiliated adult bicyclists, 600 (Schupack and Driessen, 1976). Children's mileage has probably declined since the '70s: today's low-saddle, single-speed freestyle/BMX bikes are not as efficient as '70s high-risers. A study published in the September 1994 ITE Journal (Wachtel and Lewiston 1994) unexpectedly finds children's car-bike collision rates lower than adults', but notes, for one thing, that data collection was in Palo Alto, California, which had had bicyclist education in the public schools for several years at the time of data collection..

  • Fatality risk is five times as high for males as for females (p. 11, p. 106). National Highway Traffic Safety Administration figures cited in the CPSC report do show 720 male and only 121 female fatalities for 1991. But risk also depends on exposure. Though both CPSC and Rodale data show about as many male as female riders, neither report shows riding time/mileage or location by gender. A CPSC analysis of Rodale data (p. 142) finds no independent statistical relationship between gender and risk, as does Wachtel and Lewiston 1994. Chlapecka 1975 and Schupack and Driessen 1976 show a 60 percent higher accident rate for females. Kaplan 1975 shows a 64 percent higher rate for all accidents for females, and a 57% higher rate of serious accidents. Though these are not fatality rates, they call the CPSC's conclusion into question.
  • The CPSC found riding on major thoroughfares "2.45 times more dangerous than riding on streets" (p. 74, and Table 10, p. 73). Kaplan found only a 10 percent difference between major arterials and the average of all roads, and a CPSC statistical analysis of Rodale data states that "Somewhat surprisingly, riding on highways has no independent impact on the accident risk" (p. 141), though raw Rodale data shows twice the accident rate for bicyclists who mostly use highways.

Other comparisons of riding environments are even more puzzling. According to the CPSC, riding on neighborhood streets is seven to eight times riskier than riding on bike paths. For children, neighborhood streets are "1.65 times more risky than riding on sidewalks" (p. 72). For adult riders, neighborhood streets are about as safe as sidewalks, and "8.84 times more risky than riding on unpaved surfaces" (p. 14-15, p. 72-74).

However, Rodale data, (p. 152) shows nearly the same accident rate for bicyclists who ride mostly on paths as for those who ride mostly on neighborhood streets. Kaplan 1975 found 2.6 times the accident rate on paths as on the average of all roads, for a population that included mostly experienced, adult cyclsts. Wachtel and Lewiston 1994 find a higher car-bike collision rate for sidewalk riders in 24 of 27 subgroups, including adults as well as children, with an overall ratio of 1.8 times. South et al. 1979 found nearly three times the accident rate on sidewalks as on "signed or striped lanes."

One likely explanation for the unusual CPSC figures is that more skillful riders ride less on sidewalks and paths (see Table 5, p. 65; also Wachtel and Lewiston 1994 and Kaplan 1975), while these same more skillful riders spend more of their CPSC "riding time" actually riding. Since accident rates depend on rider skill and behavior, the CPSC commits a serious error of statistical interpretation in using accident rates for an undifferentiated rider population to compare the "danger" of different riding environments frequented by different subgroups.

John Forester has pointed out some additional serious flaws in the CPSC's analysis of accident locations, including lumping all of a bicyclist's riding into the one facility category on which the bicyclist reported riding most often.

In developing its accident data, the CPSC interviewed accident participants or their parents in a random sample of 463 accidents from National Electronic Injury Surveillance System (NEISS) hospital emergency room admissions data. The hazard analysis gives percentages for "hazard patterns" (see below), and describes where accidents occurred using categories such as "street, bikepath, sidewalk." The interview questions are not published in the report, but I strongly suspect that a crosswalk collision is described as a street accident, though it is properly described as a bikepath or sidewalk accident. Respondents or interviewers also may have confused sidewalk, bikepath and bikelane. Such problems may explain the incredible figure for the hazard of riding on streets.

  • The CPSC study concurs well with others that only 10% of bicycle accidents are collisions or near-collisions with motor vehicles (p. 64). However, this figure, which does not depend on exposure data, is wildly inconsistent with the CPSC's reported 7 to 8 times worse accident rate on streets than on bicycle paths, sidewalks and even unpaved surfaces. No factor other than the presence of motor vehicles suggests itself to explain a higher accident rate on streets than in these other environments.
  • The report suggests (p. 15), that "to reduce bicycle injuries and deaths...efforts might focus on improvements in roadway design...The development of bike paths...and bike lanes...should also be considered." This statement retreats somewhat from the corresponding one on page 18 of the draft version that "the separation of bicycles from motor vehicle traffic on roadways, by creating a network of bike paths and bike lanes" could reduce bicycle-related injuries. The retreat may represent a concession to the League representatives. It is the only significant change in the conclusions I found in the final report.

Both the full-strength recommendation and the watered-down one for bike lanes represent only the authors' opinions, since the CPSC collected no data at all on bike lanes. I've already raised questions about the study's bicycle path data. I can excuse the authors' unawareness of the practical difficulties of equipping a million or so miles of neighborhood streets where most riding occurs with bike lanes or supplanting them with bike paths. Please don't infer that I oppose all special facilities for bicyclists -- but I do oppose pie-in-the-sky planning.

To sum up, the researchers failed to hone their investigative technique on the lessons of earlier studies of accident exposure and types, failed to perform cross-checking which would throw light on the credibility of their data, and failed to withdraw their report when confronted with evidence that their data collection technique was deeply flawed.

Other Findings

In the section on "Human Factors Assessment," there's truth in the usual arguments about children's limited perceptual abilities (p. 87 ff.). It is justifiable to withhold drivers' licenses from children and to prohibit them from riding bicycles on busy streets. But in this report, as all too often, these arguments are used to downplay the potential of riding skills education for children. Its effectiveness in building a lifelong pattern of orderly and cooperative behavior in traffic situations is not discussed.

The percentage of riders who never wear a helmet and "had never thought about" wearing one is 21.6 in the final version (p. 29) and 48.0 percent in the draft (draft, p. 42), apparently reflecting a correction of data analysis. The report's figure of 17.6 percent of bicyclists' wearing helmets all or most of the time (p. 11, p. 25, p. 28) is credible, and shows an expected increase over the earlier Rodale study. I think that the figure of 20.6 percent of riding time with helmets is understated, due to avid riders' spending more of their "riding time" riding. Figures for helmet use that accurately reflect riding time, mileage and riding environments might help explain the significant decrease in bicyclist fatalities over recent years.

The CPSC concludes (p. 138) from Rodale data that "[T]hose who ride on highways or other major thoroughfares may be likely to wear helmets because of higher accident risks." The CPSC author jumps to a conclusion here: the decision reflects the opinion of the bicyclists. In any case, bicyclists who have the skill and confidence to ride on major highways are likely to use helmets wherever they ride (as shown in table 18, p. 121; tables 25 and 26, p. 124).

The report's categories of "hazard patterns" (p. 63) add to over 100 percent -- not surprisingly, since many accidents reflect more than one pattern. The analysis does not describe accident types in much detail, though, as noted above, it is more less prone to error than conclusions which involve exposure data.

Percentage Hazard pattern
27 uneven surfaces
22 riding too fast
15 slippery surfaces
15 collision with moving object
13 mechanical failure (but see below)
13 hit non-moving object
11 performing stunts
6 items caught in spokes
29 other


The 13 percent figure for mechanical failure was determined through telephone interviews. Subsequent on-site investigation was carried out only for accidents in this 13 percent. This method could reduce, but not increase, the reported total, and verified only 41 incidents of mechanical failure: 10 percent of the 420 accidents involving operator injury, and 9 percent of the total sample of 463 accidents. The CPSC interviewers' knowledge of bicycles leaves room for considerable error, however: for example, they maintain that a chainguard would have kept the chain from coming off the sprocket and tangling in the rear wheel in one incident. The figures in the current report are not comparable with an early 1970's CPSC study which found mechanical failure in 17 percent of a preselected sample.

In 18 incidents of mechanical failure, the CPSC determined poor maintenance to be a factor; reading through the detailed descriptions of the incidents suggests that the figure should be considerably higher. Outright structural failures occurred in only 8 incidents, by my count; most of these involved loose parts, and could be described as maintenance-related. Some of the resulting accidents could be avoided by stiffer design requirements, such as a second brake on a coaster-brake bicycle.

On the basis of its analysis, the CPSC found "no evidence of systematic mechanical hazards that would warrant amendments of revisions to the existing mandatory standard for bicycles" (p. 14). Yet design changes and warnings can reduce the maintenance-related problems, as well as emerging accident patterns like the following:[3]

  • Quick-release hubs in the hands of people who do not know how to operate them have led to an epidemic of injuries and lawsuits.
  • Pinching fingers between the chainwheel and chain while riding is unique to today's BMX or freestyle children's bicycle with its low saddle. The report misses this point, instead making the bizarre statement (p. 96) that "[I]t is ill-advised for a rider to make adjustments to a bicycle chain while riding the bike."
  • The CPSC finds no design problems with brakes (p. 97-98), but here are two I've begun to see recently: 1) A range of coaster-brake arm straps to fit different chainstay diameters has given way to a multi-hole, flexible strap. This should be cinched tightly but is almost always looped loosely over the chainstay. There would be no problem if the brake were only used to resist forward motion -- but the brake is also used to hold the bicycle from rolling backwards when stopped and facing uphill. The assembly works back and forth and eventually fails. 2) On a mountain bike with cantilever brakes, front brake cable failure can snag the transverse cable on the knobs of the front tire and cause a pitchover accident. One such accident that left the bicyclist a quadriplegic has led to a multi-million dollar jury verdict.

The final report concedes (p. 77-78) that many cyclists trust reflectors too much; and does back off from a suggestion of the draft report: "improving the lighting in areas where nighttime bicycling is known to be prevalent (p. 17 of draft report)." In fact, many American bicyclists neglect to purchase the headlamp they need to be seen because streetlight light the way in urban areas. Other reasons for failure to use appropriate nighttime equipment are lack of law enforcement and the CPSC's own reflectors-only nighttime conspicuity policy. Unfortunately, only the Rodale study includes detailed data on nighttime equipment, and this data is based on ownership, not nighttime use rates.

All in all? The words "U.S. Government statistics show..." can only too easily choke off reasoned discussion. If the CPSC had sought out advice from experts on bicycling research, the funding would have been sufficient to determine answers to many intriguing questions, and to inform action. Instead, this study is more likely to promote the status quo in the CPSC's design regulations, and to lend fuel to the fire of public enthusiasm for ill-considered and badly-engineered bicycle facilities projects, a fire which all responsible bicycling advocates and planners struggle long, hard and with little success to put out.


1) In my article as initially published, I assumed 8 months of bicycle use per year for all rider groups. Dr. Rodgers, in his response, corrected a mathematical error in my calculations, but also assumed 8 months. My new calculation takes into account the Rodale figures on months of use given on page 133 of the CPSC report. The next page in this series describes how I worked through the calculation.

2) As in my calculation of the overall average speed above, the numbers in the table account for the Rodale figures on months of use.

3) As of September, 2003, I have become aware of additional problems resulting from new designs: failure of wheels using unconventional and unsound materials and designs, failure of poorly designed suspension front forks, and high forces excerted by disk brakes tending to pull the front wheel out of the fork.



Chlapecka et al.,  (1975). Bicycle Accidents and Usage Among Elementary School Children in the United States; National Safety Council.

Kaplan (1975). Characteristics of the Regular Adult Bicycle User, FHWA (now available online on this Web site).

South et al. (Regional Consultants, Inc.) (1979). Evaluation of the Eugene Bikeways Master Plan; City of Eugene, Oregon (now available online on this Web site).

Schupack and Driessen (1976). Bicycle Accidents and Usage Among Young Adults: Preliminary Study; National Safety Council (now available online).

Wachtel and Lewiston  (1994) Risk Factors for Bicycle-Motor Vehicle Collisions at Intersections; ITE Journal, September, 1994 (now available online).


Top: John S. Allen's Home Page
Up: Table of Contents
Previous: Letter from L.A.W.
Next: Calculation of the miles/hour figure

Contents 2001, John S. Allen
Last modified October 21, 2001