Hiles > Essays > Listening >

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

Chapter 3
Car-Bike Crashes 2
A Broader View

Figure 1
Differences in Cross-Fisher fatal and non-fatal distributions by class.

Figure 1: fatal versus non-fatal crashes.


Figure 2
Non-fatal car-bike crash distributions from five studies, grouped by Cross-Fisher class.

Figure 2: five crash studies compared.

Cross-Fisher Problem Class Descriptions

  • Class A: Bicycle ride-out from driveway, alley, and other mid-block location.
  • Class B: Bicycle rideout at controlled intersection.
  • Class C: Motorist turn, merge, drive-through, drive-out.
  • Class D: Motorist overtaking, overtaking threat.
  • Class E: Bicyclist unexpected turn, swerve.
  • Class F: Motorist unexpected turn.
  • Class G: Other.

Sources: Atkinson & Hurst, 1983; Cross, 1978; Hunter, 1994; Ross, 1992; Williams, 1981.

 

Cross-Fisher compared with other studies

Cross and Fisher launched their pioneering report nearly two decades ago. Fortunately, over the years other researchers have used the same crash classification system, or customized versions of it, to study local and regional problems. So we can see if the original study findings hold up over time and in different locations.

Figure 2 compares the Cross-Fisher non-fatal crash statistics with those of four other car-bike crash studies. Several of the studies were too small to have a statistically significant sample of fatalities, so I have only compared non-fatal statistics. Of course, the non-fatals give the best overall picture of crash type frequencies, so this is not a serious shortcoming.

There are some other differences between the original study and those done since. For one thing, not many other researchers besides Cross and Fisher were able to visit the crash sites and interview participants. Relying on police reports alone may inject some errors. Also, most of the studies used classification systems that varied in some ways from Cross-Fisher. In an attempt to simplify and increase the accuracy of the comparison, I have used the seven broader categories that Cross and Fisher called “problem classes,” not the 37 more specific “problem types.”

Five other crash studies

Missoula, Montana (Williams, 1981): This study of 91 accidents from a two-and-a-half-year period found that the median age of bicyclists involved was higher in Missoula than in the Cross-Fisher study, reflecting the large number of adult bicyclists in this college town. Otherwise, the distribution pattern follows the Cross-Fisher study rather closely. Only one of the 19 Missoula cyclists involved in nighttime accidents had a light.

New Zealand (Atkinson & Hurst, 1983): Of the 550 non-fatal collisions studied in New Zealand, 8 percent were Class D overtaking crashes, not too far off from Cross-Fisher’s 10.5 percent. New Zealand’s sample of 142 fatal crashes contained 40 percent of type D, which compares to 37.8 percent in the U.S. study. The authors noted other similarities as well:

It can be seen that the great majority of Type 13 fatalities occur under conditions of bad visibility or at night, often to unlit cyclists. Type 16 accidents (motorist misjudged space required to pass) do not occur at night; the same was true in Cross and Fisher’s study. This suggests that motorists who do see a cyclist at night leave plenty of room.

In nearly all of the Type 13 accidents that occurred in full daylight the motorist’s attention had been diverted by some other activity such as dealing with a fly on the windscreen, picking up a dropped bottle, attending to a misbehaving passenger, or watching something on the other side of the road. The apparently common notion that otherwise-attentive drivers do not see cyclists from behind is not supported by the data. Probably the fact that motorists often fail to see cyclists in other accident situations has been extended in the public imagination to include overtaking accidents.

Class E (bicyclist unexpected turn/swerve) made up an unusually high percentage of crashes in this study. The authors speculate about a number of possible causes, including New Zealand’s narrow, poorly-paved roads; old traffic laws requiring vehicles to wait at the side of the road before turning right (the equivalent of turning left in the U.S., since New Zealanders drive on the left side of the road); and even the growing popularity of BMX bicycles, which have an “image,” the authors say, that could “encourage skylarking and swerving.”

Hunter study (Hunter, Pin, & Stutts, 1994). This study classified nearly 3,000 bicyclist-motor vehicle collisions from the recent years of 1990 and 1991. The sample came from California, Florida, Maryland, Minnesota, North Carolina and Utah. A project of the University of North Carolina Highway Safety Research Center, this study found that adults make up a larger portion of the bike crash population now than they did during the Cross-Fisher study. Also, Class E is a little lower and Class G (the “other” category) is a little higher than in the Cross-Fisher study. Otherwise, the crash distribution follows the old classic very closely.

Unlike Cross and Fisher, Hunter and company did not seek out a separate statistically significant sample of fatal crashes. Of all the crashes, just 46 (1.6 percent) were fatal. To compare the relative severity of crash types, the researchers looked at the distribution of fatals combined with the 473 “serious” injury crashes in the sample. Once again, the motorist-overtaking class has the highest percentage of the worst injuries, although the differences between classes are not as dramatic as they are in the Cross-Fisher fatalities.

The two most frequent ways in which bicyclists contributed to causing crashes were “failure to yield” at 20.7 percent and “riding against traffic” at 14.9 percent.

Madison, Wisconsin (Ross, 1992). Madison is a college town with a significant network of bike lanes and bike paths. As one might expect, the Class D overtaking accidents among this sample of 774 bicyclist-motorist crashes is the lowest of any of the five studies—just 4.1 percent of the entire class.

Two other classes were unusually high, though: Class C (motorist turn, merge, drive-through, or drive out) and especially Class F (motorist unexpected turn). An on-coming motorist turning lift into the path of a straight-through cyclist made up a whopping 23.3 percent of Madison’s crashes. In the Cross-Fisher study, this type of crash accounted for only 7.6 percent of the sample. In Madison, bicyclists traveling in a contra-flow bike lane on University Avenue made up 36 percent of the victims of this type of crash. A contra-flow lane runs against the direction of traffic. In this case, it runs down the left side of a high-volume, multi-lane, one-way arterial next to the University of Wisconsin. Motorists turning left off University Avenue cross the contra-flow lane. Motorists entering the avenue from side streets turn left across the contra-flow lane; their attention is focused on the motor traffic, which comes from their right, while the bicyclists come at them from the left.

One-eyed folks of either pro-bikeway or anti-bikeway persuasion may be tempted to draw unwarranted conclusions from Madison’s unusual distribution of crash types. Pro-bikeway advocates might point to the fact that classes A, D, and E are all quite low (see Figure 2 on page 22), and that all of these kinds of crashes might be reduced by bike lanes and bike paths. Type A (bicyclist ride-out from driveway, alley and other mid-block location) may be reduced because bicyclists would ride onto a bike path or bike lane instead of into a car lane. Type D, of course, would be reduced because bicyclists would be separated from overtaking motorists. Type E crashes (bicyclist unexpected turn, swerve) would be reduced because bikeways make cyclists ride more predictably, or give cyclists room in which to “swerve” free of threat from overtaking traffic. Pro-bikeway advocates might say that classes C and F appear to be large, but that this is because bike facilities have reduced other classes to relatively small portions of the total. Even if C and F did increase, proponents might add, these two classes are the two least destructive—the fatality percentages are much lower than the non-fatal (see Figure 1). An increase in less destructive crashes may be a fair price to pay for a decrease in the more deadly ones.

Anti-bikeway advocates might point to Madison’s inordinately large Class C and Class F and charge that these are crashes they would expect to see increase because of the bike paths and lanes. These facilities give motorists and bicyclists a tendency to pay less attention to each other, it might be argued, and “hide” bicyclists from view. Worse yet, they complicate crossing and turning interactions between cyclists and motorists. These critics might argue that bike facilities have made these crashes so inordinately common that other types are dwarfed and therefore make up a smaller than normal percentage of the whole, even though they may not be significantly reduced in number. It is not possible to tell from the Madison study whether the city has an unusually low accident rate for classes A, D, and E, an unusually high accident rate for classes C and F, or some combination of the two.

The one seemingly solid specific problem, the University Avenue contra-flow bike lane, is not so cut and dried either. We might expect bicyclists riding against traffic to have a higher risk of tangling with motorists. Motorists tend to pay most attention to the primary flow of traffic: other motorists. Bicyclists whose movements don’t match the patterns of motorists on a roadway risk eluding motorists’ awareness. Wachtel and Lewiston (1994) found that bicyclists crossing intersections while riding on the wrong side of the street were twice as likely to collide with cars as right-way cyclists. They also found that wrong-way cyclists riding on sidewalk-like bike paths were about four times more likely to clash with cars while crossing intersections than were those riding with traffic.

A contra-flow lane would seem to be a mistake. University Avenue, though, is a major route to the university. Many bicyclists would have to expend extra time and energy if they took an alternative route. As a result, Madison would probably see a lot of wrong-way bicycling on that road, even without the contra-flow lane. Those bicyclists might be at an even higher risk without the bike lane. Moreover, University Avenue has a high volume of both motor and bicycle traffic, so there are more opportunities for car-bike collisions than on most streets, and we would expect higher numbers of crashes there than on less-traveled roads. Once again, we can't tell from Cross-Fisher-style studies whether bicyclists have higher or lower risks per mile in different locations.

The one thing we can say with confidence is that crash types in Madison appear to differ from the typical American city’s. Without more information, we cannot say for sure why there is a difference. It could be from bike facilities (for better or worse), from the city’s unusually large number of adult cyclists, from peculiarities in the street patterns of the city, or from any combination of these or other factors. We can’t even say if the difference we see is for better or for worse. If we saw Madison-like distributions in other towns with similar networks of bike lanes and paths, we might conclude that the facilities were a factor, but even this would not tell us if they were a good factor or bad, only that they had made a change in the relative distribution of crash types.

Table 3
Five crash studies compared
  A B C D E F G
Cross-Fisher 13.9 17.0 18.7 10.5 14.2 14.5 11.2
Missoula 8.9 10.0 23.3 13.3 8.9 20.0 15.6
New Zealand 8.0 13.0 17.0 8.0 27.0 13.0 13.0
Hunter 11.8 16.0 22.3 8.5 8.4 12.1 20.9
Madison 3.8 8.3 29.8 4.1 5.1 33.4 23.3
Average 9.3 12.9 22.2 8.9 12.7 18.6 16.8

Sources: Atkinson & Hurst, 1983; Cross, 1987; Hunter, 1994; Ross, 1992; Williams, 1981.

 

The overall pattern

The slight differences in the results of these five studies provide material for some interesting speculation. But it’s the similarities that are most significant. All of the studies reveal the same general crash patterns. Most notably, overtaking crashes are relatively infrequent. One possible explanation is that the motorists who come from behind you when you’re bicycling usually have plenty of time to see you long before they reach you, so they have plenty of time to avoid you. What’s more, an overtaking motorist is following a path parallel to yours, several feet to the left in most cases—and parallel lines don’t meet.

Most often, the danger lies where the bicyclist’s and motorist's paths cross suddenly, catching both parties by surprise, and leaving too little time for either party to avoid impact.

Another possible explanation for the low number of overtaking collisions is that, as noted earlier, bicyclists fear and avoid roads where the overtaking threat seems greatest. Given that overtaking crashes are the most destructive types, this fear would not seem totally unfounded. If this is a factor, it may mean that bicyclists are not making full use of the road system. To put it another way, it may mean that the road system is not fully accessible to bicyclists.

Yet another factor that could contribute to the relatively low percentage of overtaking collisions is that the overall statistics are child-heavy. That is, two thirds of the Cross-Fisher non-fatal crashes involved cyclists 15 years old and younger (Cross, 1978, p. 28). In general, children make more mistakes than adult bicyclists. Youngsters are especially prone to ride out of driveways or run stop signs without looking, crash types that stack the deck against overtaking types. Children also tend to stay more within residential areas where traffic is slower, so the overtaking risk is smaller. These neighborhood streets may also have frequent intersections, which may raise the likelihood of crossing and turning crashes. In contrast, for adults on rural roads, Type 13 overtaking crashes are the number one crash type, and Type 13 is number six among urban adult crashes. It should be said again, though, that this does not mean that adults on rural roads are more prone to overtaking crashes than kids on city streets. There are just fewer intersections in the country, so fewer intersection-type crashes. Also, keep in mind that most overtaking crashes, regardless of location, happen at night.

In any case, the crash report patterns show that it’s the motorists in front of cyclists, primarily at intersections, who are most likely to be involved in car-bike collisions—except, of course, when the bicyclists are riding without good lighting on narrow, high-speed roads at night. However, we cannot unequivocally conclude from this that overtaking crashes are inconsequential.

Table 4 shows the ten most frequent crash types in the Cross-Fisher study. The top seven, which together make up 52.9 percent of the crashes, are all crossing and turning types. But number eight out of the 37 crash types is unlucky Type 13, the leading cause of car-bike fatalities. So although the statistics tell us that overtaking crashes make up a relatively small part of all crashes, we can hardly tell planners and engineers to just disregard these bumps from behind.

Table 4
Top 10 Cross-Fisher crash types, all age groups
Class/Type Description percent cummulative percent
B5 Bicyclist ride-out: intersection controlled by stop sign or yield sign 10.2 10.2
C9 Motorist failure to yield at stop sign or yield sign 10.2 20.4
E18 Bicyclist unexpected left turn, parallel paths, same direction 8.4 28.8
F23 Motorist unexpected left turn, parallel paths, same direction 7.6 36.4
A1 Bicyclist ride-out from residential driveway or alley 5.6 42.0
F24 Motorist unexpected right turn, parallel paths, same direction 5.6 47.6
C8 Motorist drive-out from commercial driveway 5.3 52.9
D13 Motorist overtaking, bicyclist unseen 4.0 56.9
A2 Bicyclist ride-out from commercial driveway 3.2 60.1
E19 Bicyclist unexpected left turn, parallel paths, opposite direction 3.3 63.3

Source: Cross, 1978.

 

Education and engineering: different needs, different outlooks

The bottom line is that a cycling instructor teaching a group of bike club members how to ride in city traffic can confidently say that by far the most frequent collisions are crossing and turning types and that urban cyclists have a relatively slim chance of getting rear ended as long as they don’t ride at night without lights. On the other hand, planners and engineers can point with equal confidence to overtaking crashes as one of their main concerns. Overtaking crashes are the number one cause of bike-related fatalities—and fatalities make the news, spur people to action, and bring demands onto city officials far more than non-fatal crashes. Moreover, we must keep in mind that when traffic planners and engineers are pressured to take action to help prevent bike crashes, of all their engineering choices none are as visible to the public as bike lanes and paths, and these facilities are capable of reducing both overtaking collisions, which are the most deadly types, and bicyclist ride-out collisions, the most common type of car-bike crash for younger children (Wilkinson, et el., 1994b, pp. 24-25). In a later chapter we will see more clearly that crash statistics alone do not give us a complete picture of why planners, engineers, and bicycle advocates have reason to shape environmental design with the overtaking threat in mind.

Table 5
Cross-Fisher Type 13 crashes summarized
  • 4.0% of non-fatal crashes
  • 24.6% of fatal crashes
  • 10% of urban fatal crashes
  • 50% of rural fatal crashes
  • Two thirds happened at night, 90% to cyclists without lights
  • Nearly one third involved a drinking driver
  • Half of urban Type 13’s and three fifths of rural Type 13’s were on narrow, two-lane roadway with no ridable shoulder.
  • Number six adult urban crash type, out of 37 types.
  • Number one adult rural crash type

Sources: Cross, 1978, pp. 71-72; Forester 1993, p. 269; Williams, 1993a.

 


Previous | Contents | Next