Hiles > Essays > Listening >

Hiles, Jeffrey A. Listening to Bike Lanes. September 1996.

Chapter 2
Car-Bike Crashes 1
Those Bothersome Bumps From Behind

The kind of separation between cars and bikes provided by bike lanes and next-to-the-road bike paths helps keep motorists from bumping into bicyclists’ back ends. However, these facilities do little to prevent the numerous kinds of collisions caused by the crossing and turning movements of both bicyclists and motorists. Bikeway critics, therefore, question whether overtaking motorists are enough of a threat to justify the effort to separate bikes from cars, considering that there may be side effects, such as hindering bicyclists’ movements and making crossing and turning more difficult, or even more frequent. This chapter will put through the ringer what we know about bump-from-behind collisions in hopes of squeezing out a reasoned understanding of these controversial car-bike crashes. (For convenience, I will use “car” to refer to any kind of motor vehicle.)

John Forester (1994) argues that you can get a clear answer to the overtaking-risk question by looking at crash statistics:

These show that many more car-bike collisions (about 95 percent) are caused by crossing and turning maneuvers from in front of the cyclist than are caused by the car-from-behind-a-lawful-cyclist collisions that worry cyclist-inferiority believers so much. Furthermore, car-bike collisions are only about 12 percent of all accidents to cyclists. This combination makes the car-overtaking-a-lawful-cyclist in urban areas in daylight (which is the type of accident used to justify transportational bikeways) only about 0.3 percent of total accidents to cyclists (pp. 10-11).

Kenneth Cross (1978), whose bicycle crash studies form the foundation for many of Forester’s arguments, paints what seems to be a different picture when he describes what he calls “Problem Type 13,” in which an overtaking motorist fails to see a bicyclist until it’s too late to avoid a collision:

Although seven other problem types occurred more frequently than Problem Type 13, this problem type must be considered one of the most important because it accounted for nearly one-fourth of all fatal accidents in the sample—three times as many as any other problem type (p. 72).

So we have on the one hand an analysis that says the overtaking risk is negligible, and on the other hand an analysis that characterizes the overtaking collision as the most deadly of all car-bike crashes. A clearer picture emerges when we look more closely at Type 13 crashes and, first, at the study from which these statistics came.

The Cross-Fisher study

Cross and Gary Fisher published the results of their landmark car-bike crash study in 1977. They had gathered police reports describing 166 fatal and 753 non-fatal car-bike crashes from four areas in different parts of the United States: Los Angeles, California; Denver-Boulder, Colorado; Tampa-Orlando, Florida; and Detroit-Flint, Michigan. After undertaking the monumental task of visiting crash sites and interviewing participants, the researchers sorted the crashes into 37 different problem types, grouped into seven general classes (Cross, 1978, p. 25).

Class D consists of five motorist-overtaking-bicyclist crash types. Among these, Type 13 makes up a very high portion of bicyclist fatalities, is arguably the most feared of all accident types, and has therefore been studied very thoroughly. Because of that, and since fear of this motorist-overtaking-unseen-bicyclist crash type is both a strong motivation for advocates of separate bicycle facilities and a major target of those who oppose such facilities, this discussion focuses mainly on Type 13.

Fatal versus non-fatal crash reporting

The difference between fatal and non-fatal crash reporting is an important part of the picture. As we will see, fatalities give us a sense of how destructive different crash types tend to be. Non-fatal crash statistics show us the relative frequencies of different crash types.

Another difference between fatal and non-fatal crashes is that we have a complete record of fatalities, but a great many non-fatal crashes go unreported. Cross estimated that at least two thirds of all car-bike collisions are not reported to police, even though more than half of those unreported crashes inflict injuries “severe enough to require some form of medical treatment” (p. 16).

This raises the question of whether the one third of non-fatal crashes that get reported accurately represent the total crash picture. Bicycles with severe injuries, for example, have more incentive to report crashes than cyclists with slight scratches or no injuries at all. So unreported injuries may be less serious on average than reported injuries. It is possible, then, that crash types that tend to produce more serious injuries would be over represented in the crash reports. There is nothing wrong, though, with being more concerned about crashes that do more damage. For practical purposes, then, we should be well served by the picture of car-bike collisions we get from looking at those that do get reported.

Perspective: 1993 U.S. car-bike crashes

To be more concrete, in 1993 car-bike collisions in the United States killed 814 bicyclists; this we can say with confidence. Also, there were some 65,000 reported non-fatal car-bike injuries (National Highway Traffic Safety Administration, 1994, p. 129). Perhaps more than 130,000 car-bike collisions went unreported. So fatalities made up 1.2 percent of reported car-bike collisions and perhaps less than 0.4 percent of all car-bike collisions.

The non-fatal statistics, then, represent about 99 percent of all car-bike collisions and give us the best picture by far of the relative frequencies of different crash types. The fatality percentages, when compared with the non-fatal distribution, give us an idea of the relative destructive capacities of the crash types. We can say, for example, that Type 13 crashes in the Cross-Fisher study are relatively infrequent; they make up a small portion of non-fatal crashes. But when they do happen, they are much more destructive than most; they make up a significantly larger portion of fatalities.

If the Cross-Fisher statistics were accurate for 1993 car-bike collisions, we would expect to see about 200 deaths from Type 13 crashes and about 2,600 non-fatal Type 13’s, perhaps 7,800 if you include unreported crashes.

Of all 1993 car-bike collisions, fatal and non-fatal combined, Type 13 would make up 4.3 percent of reported crashes and perhaps 4.1 percent of all crashes including those not reported. Fatal Type 13 crashes would make up 0.3 percent of all reported car-bike collisions and perhaps 0.1 percent of all crashes, counting those not reported.

It is important (and scary) to realize that nearly one fourth of all bicyclists killed by cars were hit by overtaking motorists who did not see them. But it is also important to keep in mind that those overtaking fatalities account for a fraction of a percent of all car-bike collisions. Although non-fatal injuries, especially those involving brain injury, can be as tragic as fatal injuries—perhaps worse.

Table 1
Estimated share of reported 1993 U.S. motor vehicle-bicycle crashes for Cross-Fisher Problem Type 13: motorist overtaking, bicyclist unseen.
  Non-Fatal Fatal Total

Total reported car-bike crashes, all types




Estimated Type 13 (4% non-fatal, 24.6% fatal)




Estimated Type 13 percentage of total reported crashes




Sources: Cross, 1978 ; National Highway Traffic Safety Administration, 1994


Other overtaking crash types

In addition to Type 13, in which the motorist did not see the cyclist, Cross and Fisher named five other types of motorist-overtaking crashes, in which the driver did see the bike before the crash. These included Type 14, where the car was out of control; Type 15, where the cyclist swerved left to escape the overtaking car as the motorist swerved in the same direction in an attempt to avoid the bike; Type 16, where the motorist misjudged the space required to pass; Type 17, where the cyclist’s path was obstructed, forcing the poor pedaler to either swerve into the path of the overtaking car or collide with the obstruction; and a category with no number labeled “motorist overtaking: type unknown.” Collectively, these make up “Problem Class D–Motorist Overtaking/Overtaking Threat.” Table 2 shows the Cross-Fisher statistics for Class D.

Table 2
Cross-Fisher Class D Car-Bike Crashes



Type 13: Bicyclist not observed 24.6% 4.0%
Type 14: Car out of control 4.2% 0.7%
Type 15: Counteractive evasive 2.4% 1.7%
Type 16: Motorist misjudged 1.8% 2.0%
Type 17: Path obstructed 0.6% 2.0%
Type unknown 4.2% 0.1%
Total Class D 37.8% 10.5%


Bike crashes with and without cars

Forester may be close to the mark in saying that cars play a part in only 12 percent of cycling crashes. Studies of emergency room-treated bicycle injuries indicate that motor vehicles are involved in 9.4 or 18 percent of these cases (Clarke & Tracy, 1995, p. 29; Stutts, Williamson, Whitley, & Sheldon, 1990, p. 71).

It does not follow, though, that we should only devote 12 percent of our bike safety concern to car-bike crashes. A Seattle, Washington, study found that half of all serious bike injuries involve motor vehicles. What’s more, 82 to 96 percent of bike-related deaths involve motor vehicles (Cross, 1978, p. 22; Rogers, 1994, p. 10; Bicycle Institute of America, 1993, p. 6). On average, then, car-bike collisions tend to inflict worse injuries than bike crashes that don’t involve motor vehicles.

Unlucky Type 13

So, car-bike collisions are inordinately dangerous among all types of bike crashes, and Type 13 is inordinately deadly among car-bike collisions. It would appear that to wave off these overtaking accidents because they make up a small part of all bike crashes is like saying never mind the cobra as you walk through the snake pit because most of the serpents in there are garter snakes.

However, Type 13 crashes, like so many bicycle transportation issues, can’t be accurately summed up with so simple a statement. A recurring theme throughout this paper is that the more you dig below the surface of bicycle issues, the more the picture changes—usually becoming more complex. Another part of the Type 13 story unfolds when you look at when and where this crash type most often occurs: at night and on rural roads.

Overtaking: a rural and nighttime problem

Cross reported that this motorist-overtaking category was the only crash type in his study where nighttime crashes out-numbered daytime ones. Nighttime crashes made up 71 percent of Type 13 fatalities, but only 30 percent of all fatal collisions. Also, 65 percent of non-fatal Type 13 problems were in darkness. In contrast, only 10 percent of non-fatal collisions of all types fell between dusk and dawn (p. 36). In more than 90 percent of the nighttime Type 13 crashes, the cyclists had no lights (Williams, 1993a).

According to Cross, “about 60 percent of the Type 13 accidents occurred on a narrow, ‘rural-type’ roadway with two traffic lanes and no ridable shoulder or sidewalk” (p. 72). This problem type made up half of rural car-bike fatalities, as opposed to just 10 percent of urban ones (Williams, 1993a).

To make a more general statement, two key factors in Type 13 crashes are poor visibility and narrow roads. This means, for one thing, that on some roads the risk of getting hit from behind will be higher than the Cross-Fisher average. It also suggests that a blanket statement which says that a particular crash type makes up X percent of crashes is simplistic and may not apply to a given road.

Speed kills

As we have seen, overtaking collisions make up an inordinately large portion of fatalities, even though they make up a relatively small portion of all collisions. This is easy to understand when you consider that such crashes are apt to involve higher speeds.

In the Cross-Fisher study, more than half of all fatalities were on roads with posted speed limits greater than 35 mph, even though less than 20 percent of all collisions occurred in that fast traffic (Cross, 1978, p. 40). A more recent study of fatal accidents in Victoria, Australia, closely matched these findings (Hoque, 1990, p. 4).

The United Kingdom Department of Transport has provided a more dramatic illustration of the difference speed makes. The department determined that when pedestrians are struck by cars traveling at 20 mph, only about five percent are killed and most injuries are slight, with 30 percent of the walkers left virtually unscathed. At 30 mph, though, 45 percent are killed and many seriously injured. Cars zipping along at 40 mph kill 85 percent of the pedestrians they strike (Bicycle Federation of America, 1993b).

Simply put, cars that are overtaking cyclists are more likely to be at full speed than, say, cars crossing and turning at intersections. The higher the speed the harder the impact, and the more damage done.

Where (not whether) overtaking is a problem

In a quote at the beginning of this chapter, Forester creates the impression that what he calls “bicyclist inferiority believers” base their bicycle facilities planning on a distorted and irrational fear of a type of crash that makes up “only about 0.3 percent of the total accidents to cyclists.” But in using this “0.3 percent” figure, Forester himself distorts the bicycle crash picture.

First of all, Forester waters down the statistics by including bike crashes that aren’t relevant to the bikeway discussion. Bicyclists rounding turns too fast, slipping on wet leaves, running into dogs and having other non-motor vehicle crashes have little to do with the wisdom or folly of bikeways.

The one exception is when cyclists mix with pedestrians on trails, an environment which may foster collisions between the two. Yet, Forester uses his “0.3” figure to argue against “bikeways” in general, which he defines as including not only “bicycle sidewalks or side paths,” but also “bike lanes that are part of the roadway,” a type of facility where car-bike collisions (and, perhaps to a much smaller extent, bike-bike collisions) are virtually the only germane types of crashes.

Second, Forester’s “0.3 percent” figure gives no more weight to a crash that’s likely to kill than to a crash that’s likely to cause a few cuts and scratches. Certainly any fall can be fatal, even a slip in the bath tub. Nevertheless, some ways of falling from a bicycle are more likely to cause serious injuries than others. As we have seen, it appears that the motorist-overtaking collision is on average the most destructive of all.

Third, Forester masks the overtaking problem by restricting his count to “urban” roads. To distinguish between roads they called “urban” and those they labeled “rural,” Cross and Fisher looked at the characteristics of the roads, not whether they were within city limits:

Accidents usually were classified as rural if they occurred in an area where (a) the posted speed limit was 45 mph or more, (b) there were no curbs or sidewalks adjacent to the roadway, (c) street lights were not present at the intersections, and (d) at least 50 percent of the area within one-half mile radius of the accident sites was open. Cases that did not meet all four of these classification criteria were classified as urban.

By restricting the car-bike portion of his crash count to roads classified as “urban,” Forester is excluding many of the kinds of roads on which overtaking crashes are most likely to occur. He is even excluding some roads that are actually within city limits. With such an unusually dangerous crash type, it’s important to understand the dynamics of where it happens and why. Cross describes the setting of Type 13 crashes:

About three-fifths of the rural accidents and about one-half of the urban accidents occurred on a narrow, two-lane roadway with no ridable shoulder. Thus, about 60 percent of the Type 13 accidents occurred on a narrow, “rural-type” roadway with two traffic lanes and no ridable shoulder or sidewalk (p. 72).

Fourth, Forester says that “practically all Americans” are “totally misinformed about cycling in traffic” and as a result have phobic fears of the cars from behind (1994, p. 8). Assuming that is true, we would expect practically all Americans to do everything possible to avoid cycling on narrow, fast, hilly, winding busy streets. For the sake of argument, suppose that those streets pose a higher overtaking threat. Then we must ask this: Does the relatively low number of overtaking collisions mean only that there is little threat of that kind of accident in the street system, or do cyclists themselves keep the number of overtaking collisions low by staying away from streets where the threat is strongest? Fifth, to use Forester’s own words, “there is no reasonable way to rank car-bike collision types in order of importance, because the order depends upon what kind of cyclist you are and where you are riding…” (1993, p. 268).

The Cross-Fisher statistics are stacked by the preponderance of crashes by children, whose cognitive skills are not as well developed as older cyclists’. The top children’s crashes are caused by kids running stop signs and riding out of driveways (Forester, 1993, p. 269).

Moreover, few children in the Cross-Fisher study were involved in Type 13 crashes. The median age of bicyclists in these unlucky crashes was higher than it was for any but one other crash type. “Apparently,” Cross wrote, “bicyclists younger than 11 or 12 years of age are not permitted to ride during darkness and in types of areas where Type 13 accidents occur” (p. 73).

The top crash types for adults, though, are not loaded with the kinds of simple, careless errors that children commit. Arguably, adult crashes better represent the inherent hazards of the cycling environment—as opposed to hazards of bad cycling. For adults on roads classified as “urban,” Type 13 ranked seventh out of the 37 crash types, followed by Type 16, in which overtaking motorists saw the bicyclists, but misjudged and passed too closely.

On “rural” roads, Type 13 was the number-one crash type for adults, followed again by Type 16 (Forester, p. 269). Granted, in the lion’s share of Type 13 crashes, the bicyclists had failed to make themselves visible with lights at night. Type 16, however, is also an overtaking problem, is number two for rural roads, and can’t be blamed on anything as simple as lightless night riding.

So, although overtaking motorists are involved in a relatively small portion of bike crashes in general, some stretches of roadway have characteristics that contribute to a higher than average overtaking threat. When it comes to saying yea or nay on plans for bicycle facilities, it seems more effective to address problems of specific places than to adopt blanket policies as if all streets were the same.

A number of factors aggravate the overtaking threat:

High speeds

As we have seen, high speeds mean high impact, which means more severe injuries. In some cases speed might also increase the likelihood of overtaking collisions by decreasing the amount of time motorists have to see, recognize, and maneuver around bicyclists. Although this should not be a factor if motorists are traveling at speeds that suit the conditions, small hills and bends in otherwise straight roads can create blind spots.

Narrow lanes

Cyclists are more likely to be in the path of overtaking motorists when the lanes are narrow. With wide lanes, cyclists are more likely to be to the right of the danger zone.

Sight obstructions

Again, hills and turns limit how far motorists see down the road and decrease the amount of time motorists have to recognize and avoid bicyclists.

Night riding

This may not appear to be a design factor, since you can’t control the sun. It may seem like more of a law enforcement problem: getting cyclists to use lights at night. But just a small red light, and in some places just a small red reflector, can satisfy the law without making a bicyclist all that conspicuous on the roadway. Also, bicyclists are not required to have lights on their bikes in the daytime. Few cyclists would want a law that requires them to load their bikes with lighting equipment at all times. But this means that bicyclists who never anticipated riding at night are not likely to have lights when the need for a night ride arises, such as when a meeting unexpectedly runs past dusk.

It would be great if every bicyclist who rode at night looked like a rolling Christmas tree. Certainly, the requirement to have lights at night is one of the most important bike laws that police can enforce for bicyclists’ own good. But the advocate or bicycle planner whose town has a lot of nighttime cycling—a university town, for example—may find that facilities such as bike lanes, for example, reduce serious injuries on some roadways more effectively than trying to create a utopian society.

Alcohol/drug use

Alcohol was identified as a factor in a third of Cross and Fisher’s rural Type 13 fatalities (Cross, 1978, p. 73). In a 1990s update of the Cross-Fisher study, “alcohol/drug use” was found to be an “over representation” for crashes involving bicyclists in adult age groups (Hunter, Pein, & Stutts, 1994, p. 10). The National Center for Statistics and Analysis (1994) also found that “alcohol involvement—either for the driver or the cyclist—was reported in more than a third of pedalcyclist fatalities in 1993” (p. 3).

Traffic volume?—Not necessarily

Heavy traffic would seem to be an obvious risk factor. It appears reasonable to assume that the more frequently a bicyclist gets passed, the more that bicyclist is exposed to the overtaking threat. But the relationship between traffic volume and crashes is not that straight forward. Wachtel and Lewiston (1994), for example, found no significant relationship between traffic volume and the risk to bicyclists crossing intersections (p. 32). Studies of motor vehicle crashes not involving bikes have even shown that accident rates are sometimes lower with increased traffic (Hall & Pendleton, 1990). This makes sense if you imagine a bicyclist riding urban streets at rush hour when both bicyclists and motorists are extra vigilant and careful because of the heavy traffic. In contrast, picture a cyclist without good lights and reflectors who is pedaling in darkness on a narrow country road. Because of the scarce traffic, a motorist would not anticipate encountering a bicyclist there and could crest a hill and find the cyclist just ahead with no time to react.

Garder, Leden, and Thedeen (1994) point to two more studies that support the notion that the number of “bicycle accidents at an intersection is proportional to the bike volume, but not very dependent on the motor-vehicle volume.” They speculate that “increased vehicle volumes make the cyclists more careful. At least the risk to cyclists would not increase in proportion to the vehicle-volume increase…” (p. 433).

Seat-of-the-pants profile

Putting this all together, we might expect an unusually high overtaking-crash problem on a road with speeds of 45 mph or faster that is narrow, two-lane, hilly, winding, and that connects university student housing with popular night spots in a community that has a depressed economy and therefore high alcoholism.

But “unusually high” is a vague assessment. How bad of a problem would overtaking crashes be on the rare road that fits that profile? How about on roads that fit parts of the profile? There are too many variables and there is too little information to give anything but an intuitive impression.

It is not even possible to clearly define the overtaking risk on “rural” versus “urban” roads. Cross-Fisher tells us that Class D accidents made up 10.5 percent of the total non-fatal sample, but 31 percent of the rural portion of that sample. They made up 37.8 percent of the total fatal sample, but 56 percent of fatalities on roads classified as rural (Cross, 1978, p. 71). This does not, however tell us about risk. Rural-type roads have far fewer intersections per mile than urban roads. So there are fewer opportunities for crossing and turning movements and we would expect these non-overtaking crashes to make up a smaller portion of the picture, even if the risk per mile of getting hit from behind was the same for both urban and rural roads.

Given two randomly-selected car-bike crashes, one rural and one urban, we can say that the chances of the rural one being an overtaking collision is higher than it is for the urban. But to determine the relative risks of riding on different roads, we would have to know how much bicycle traffic there was on the road for each accident in the sample. The data is not there to either confirm or deny differences in overtaking threat from one road to another.


We can say that motorists overtaking bicyclists accounted for 10.5 percent of the non-fatal crashes in the Cross-Fisher study, so that about nine out of 10 car-bike collisions involved crossing and turning movements. We can say that the dangerous Type 13, where the motorist didn’t see the cyclist, made up only four percent of the non-fatal crashes. Furthermore, urban daytime Type 13’s accounted for only one percent of the non-fatals. But how much does that tell us about the problems on a particular stretch of road? And how do we weigh the severity of crashes? How many cuts and scrapes equal a death?

The purpose of this chapter is neither to disparage Forester nor to exaggerate the overtaking threat. Rather it is to chip away at the illusion of certainty that numbers can create. We must use aggregate statistics with caution; they may mislead us when we make decisions about specific local problems.

Previous | Contents | Next