Chapter 2

[Table of Contents] [Chapter 1] [Chapter 2] [Chapter 3] [Chapter 4]

Chapter 2
Summary of Main Results

This comparative analysis of BL and WCL sites was based on videotapes of almost 4,600 bicyclists in three U.S. cities approaching and then riding through intersections for which the associated bicycle facility was either a BL or WCL. In two of the three cities, the vast majority of bicyclists were traveling to or from college campuses, and the intersections selected were generally in bicycle commuting corridors. The intent was to videotape bicyclists who regularly ride in traffic. The result was a group of sites with varying "real-world" characteristics such as different BL striping techniques (e.g., using a solid or dashed BL stripe all the way to the intersection), presence of parking (e.g., a combination BL and parking lane), and provision of turn lanes at intersections that sometimes narrow the nominal width of the BL or WCL at the intersection proper. What follows is a brief summary of the main operational and safety (conflict) results and some further elaboration of key issues.

Summary of Main Results

Bicyclist Characteristics

The overwhelming majority of videotaped bicyclists were between the ages of 16 and 64. Slightly more than three-fourths were male.
While wrong-way riding on a sidewalk is not necessarily illegal or improper behavior, it can lead to operational and safety problems with motor vehicle traffic. Thus, it has been defined and used in this report as a behavioral characteristic of bicyclists. Overall, 5.6 percent of the bicyclists were riding the wrong way (i.e., facing traffic). This included 1.3 percent in the road and 4.3 percent on sidewalks. However, wrong-way riding was much more prevalent on the sidewalk at WCL sites (7.0 percent) compared with BL sites (2.3 percent). Eliminating sidewalk riding from the comparison, however, still resulted in significantly more wrong-way riding associated with WCL sites (1.7 percent) than BL sites (1.0 percent).
A bicyclist experience oral survey was administered to bicyclists proceeding through the project sites on days when videotaping was not being done. There were no statistically significant differences in the age, gender, and helmet use of bicyclists by type of facility. Higher proportions of Whites and Blacks rode in WCL situations and higher proportions of Asians and Hispanics in BL situations, and the differences were significant.
Bicyclists surveyed at WCL sites tended to ride more days per week, but the miles per week for bicyclists at BL versus WCL sites were equivalent. Overall about one-third of the riders at both BL and WCL sites considered themselves to be experienced bicyclists.
When bicyclists were surveyed, their riding location (i.e., in the street or on the sidewalk) when approaching the survey station was recorded. Surveyed bicyclists showed the same tendency as the videotaped bicyclists in that sidewalk riding was more associated with WCL sites.

Midblock Movements

In the midblock or intersection approach area (between 90 and 150 m from the intersection), significantly more motor vehicles passing bicycles on the left encroached into the adjacent motor vehicle traffic lane from WCL situations (17 percent) compared with BL situations (7 percent). This is in agreement with results from a recent Florida DOT study (Harkey and Stewart, 1997). However, encroach- ments into the adjacent traffic lane very rarely resulted in a conflict with another motor vehicle (figure 5).

Statistical Modeling of Spacing Between Bicycles and Motor Vehicles

Using least squares regression analysis, statistical models were used to examine lateral positioning of the bicyclists with respect to the curb and parked vehicles, as well as separation distance between bicyclists and motor vehicles. The primary purpose of this analysis was to determine which geometric and traffic operational variables influence these measures and to determine if there were differences in these measures that could be attributed to type of facility (i.e., wide curb lane vs. bicycle lane). The results from this analysis are summarized below:

On facilities with no on-street parking, bicyclists tended to position themselves closer to the curb (or gutter pan seam, if present) when the BL widths were less than or equal to 1.6 m compared with WCL facilities with the same traffic volume. When the BLs were greater than 1.6 m in width, bicyclists tended to position themselves further from the curb compared with WCL sites. When motor vehicles were passing bicyclists, the position of the bicyclists tended to be about 0.3 m closer to the curb compared with their position when not being passed. This result was the same irrespective of type of facility.

On roadways with bicycle lanes and on-street parking, bicyclists positioned themselves about the same distance from parked vehicles as they did from the curb on roadways with bicycle lanes and no on-street parking. The small number of observations on WCL facilities with on-street parking prohibited similar analyses for WCL sites.
With respect to separation distance between bicyclists and passing motor vehicles, there were no practical differences between BL sites and WCL sites. Instead, this distance was primarily a function of the total width available (either the WCL width or the BL width and adjacent motor vehicle lane width combined).
(See Harkey and Stewart (1997) for more information about spacing between bicycles and motor vehicles at midblock locations when the bicycle facility is either a BL, WCL, or paved shoulder.)

Intersection Movements

The intersection was defined as starting 90 m upstream from the stop bar and included the intersection proper. Proportionally more bicyclists approached the intersection on a sidewalk when the facility was a WCL (15 percent) than a BL (3 percent).

Overall, 92 percent of bicyclists obeyed the traffic signals that were present, and there were no differences by facility type. When a signal was disobeyed, 16 percent of the actions were considered somewhat unsafe and 2 percent definitely unsafe. There were no differences by facility type.
Overall, 75 percent of bicyclists obeyed existing stop signs. Proportionally more bicyclists obeyed stop signs at BL sites (81 percent) than at WCL sites (55 percent). When a stop sign was disobeyed, 13 percent were considered somewhat unsafe and 2 percent definitely unsafe. The proportion of bicyclists with both somewhat unsafe (19 versus 5 percent) and definitely unsafe (3 versus 0 percent) movements was higher at BL sites. The differences between BL and WCL sites were significant when the somewhat unsafe and definitely unsafe categories were combined.
Seventy-two percent of the bicyclists went straight through the intersection, with another 15 percent turning left and 13 percent turning right. There were no differences by facility type. Nine percent of the bicyclists tended to shy to the right (i.e., move to the right and away from traffic) as they went straight through the intersection (11 percent in BLs and 7 percent in WCLs), and this difference was significant.
Left turns presented a problem for bicyclists and were made in a variety of ways (figure 6). Overall, 44 percent made left turns like a motor vehicle with proper lane destination positioning (41 percent from BL sites and 48 percent from WCL sites)
.

On the other hand, 14 percent of bicyclists at WCL sites made motor vehicle style left turns with improper lane destination positioning compared with 3 percent from BL sites. There were proportionally more pedestrian style left turns from WCL sites (24 percent versus 12 percent from BL sites). Both findings may reflect the generally higher traffic volumes and speeds and greater number of lanes at WCL sites.

Right turns for bicyclists were an easier maneuver, with only 13 percent made in a non-standard fashion (e.g., from a BL or WCL to a wrong way position on the cross street). Nineteen percent of the right turns made at WCL sites were non-standard versus 10 percent of right turns at BL sites, and the differences were significant.

Midblock Conflicts

A conflict was defined as an interaction between a bicycle and motor vehicle, pedestrian, or other bicycle such that at least one of the parties had to change speed or direction to avoid the other.

Of the 188 midblock conflicts, 71 percent were bicycle/motor vehicle, 10 percent bicycle/bicycle, and 19 percent bicycle/pedestrian. Almost all of the bike/bike conflicts occurred in BLs, typically due to one bicyclist maneuvering around a slower moving bicyclist. Compared with BLs, bicyclists in WCLs experienced more bike/pedestrian conflicts (30 percent versus 16 percent, and reflective of the increased sidewalk riding in WCL situations) and less bike/bike conflicts. The differences by facility type were statistically significant.

There were no differences in the bicycle or motor vehicle avoidance response scales by facility type. The scales ranged from no change in riding or driving up to collision or near crash.
Overall, 98 percent of the midblock conflicts were coded as minor, and there were no differences by facility type.
Bicycle actions more associated with BLs in these midblock conflicts included the bicycle having to slow, stop, or swerve for traffic not influenced by the intersection; the bicycle turning or swerving across a lane of traffic (figure 7); encounters with other bikes; and "other" bike actions (such as an improper left turn). The bicycle action more associated with WCLs in these midblock conflicts was encounters with pedestrians.

Motor vehicle actions more associated with BLs in these midblock conflicts included illegal parking in the BL and entering/exiting on-street parking or a driver or passenger entering/exiting a parked or stopped vehicle. Motor vehicle actions more associated with WCL conflicts included turning right in front of a bicyclist after overtaking and "other" actions such as failing to yield, improper right turns, and crowding bikes.

Intersection Conflicts

Similar to the midblock area, an intersection conflict was defined as an interaction between a bicycle and motor vehicle, pedestrian, or other bicycle such that at least one of the parties had to change speed or direction to avoid the other.

Of the 198 intersection conflicts, 79 percent were bike/motor vehicle, 10 percent bike/bike, and 10 percent bike/pedestrian. The differences in the BL/WCL distributions were statistically significant. There were proportionally more bike/bike conflicts in BLs (15 percent) and less in WCLs (4 percent). Conversely, there were proportionally more bike/pedestrian conflicts in WCLs (17 percent, and again reflective of sidewalk riding) and less in BLs (6 percent).

The position of the motor vehicle with respect to the bicycle in the intersection conflicts was 66 percent in the same direction, 6 percent in the opposing direction, 5 percent approaching from the left, 15 percent approaching from the right, and 7 percent approaching from some other position. There were no differences by facility type.

There were no differences in the bicycle or motor vehicle avoidance response scales by facility type.

Overall, 93 percent of the intersection conflicts were coded as minor, and there were no differences by facility type.

Bicycle actions more associated with BLs in these intersection conflicts included the bicycle having to slow/stop/swerve for intersection traffic, the bicycle having to slow/stop/swerve for traffic not influenced by the intersection, and the bicycle turning or swerving across a lane of traffic. Bicycle actions more associated with WCLs included passing slow moving or stopped vehicles on the right, encounters with pedestrians, and "other" actions such as improper left turns and merging onto the road from a sidewalk.

Motor vehicle actions more associated with BLs included illegal parking in the BL and "other" actions such as a driver or passenger entering/exiting a parked or stopped vehicle (figure 8) and crowding the BL. Motor vehicle actions more associated with WCLs included having to slow/stop/swerve for intersection traffic and turning right in front of a bicyclist after overtaking.

Statistical Modeling of Conflict Data

Raw frequency bike-motor vehicle conflict rates per entering bicyclist were slightly higher at BL sites than WCL sites when midblock and intersection conflict data were combined (6.7 versus 5.1 bike-motor vehicle conflicts per 100 entering bicyclists).

The rate of midblock bike/motor vehicle conflicts associated with BLs was considerably higher than the rate for WCLs, although the rates were small. Generalized linear models fitted to the data showed that both the presence of a BL and the BL width, along with traffic volume and the presence of driveways, were significant variables in the midblock conflict rate models. The practical effect of such models was that the midblock bike/motor vehicle conflict rate was higher at sites with BLs less than 2.5 m wide than at WCL sites. However, a closer examination of the data revealed that the higher midblock BL conflict rates were attributable to only a few sites. The midblock conflicts at the 10 highest rate sites were thus examined clinically.

An initial model fitted to the intersection conflicts showed no differences in the conflict rate by type of bicycle facility, but higher conflict rates for bicycle left turn movements. A subsequent model was developed that included different intersection types (figure 9) based on the type of BL striping (e.g., solid stripe to the intersection, dashed stripe to the intersection) and whether the typical WCL cross section was maintained through the intersection (or narrowed due to the provision of turn lanes). The model showed lower conflict rates for straight through and right turning bicycles where the BL stripe continued all the way to the intersection and the WCL was not narrowed at the intersection. This is perhaps not surprising, in that bicycles would have more space in these configurations.

As before, a closer study of the data showed that the findings from this model

were mainly attributable to a few sites. The difficulty of statistically interpreting outcomes that seemed so dependent on site-specific characteristics led to clinical analysis of higher conflict rate sites, both at midblock and intersection locations. The results of this clinical examination are described below.

Clinical Examination of High Conflict Rate Sites

The 10 highest conflict rate sites for both the midblock and intersection areas were examined clinically to determine if any typical conflict patterns existed. In the midblock area, there were seven BL and three WCL sites. The predominant motor vehicle actions in the midblock conflicts pertained to motor vehicles entering or exiting on-street parking (there were several sites where

parking was part of the facility), parking or stopping in the bicycle facilities to let a passenger enter or exit the vehicle, and pulling across the BL or WCL into an intersecting street or driveway. The predominant bicycle actions were turning or swerving across a lane of traffic (usually to avoid making a left turn at the intersection ahead) and interacting with pedestrians when riding on the sidewalk. If "fault" in the conflicts had been assigned, the large majority of the fault would have been due to motor vehicle actions.

In the intersection area, there were four BL and six WCL sites. The predominant motor vehicle actions again pertained to entering or exiting on-street parking and parking or stopping in the bicycle facility to let a passenger enter or exit the vehicle. The predominant bicycle actions were turning or swerving across a lane of traffic, passing slow or stopped motor vehicles on the right, and interacting with pedestrians. Some of the conflicts resulted simply from the typical maneuvering that might occur when bicycles and motor vehicles position themselves to make turns at intersections. If "fault" in the conflicts had been assigned, the majority would have been due to bicycle actions.

Identifiable situations leading to conflicts from this clinical analysis were presence of parked motor vehicles (either entering/exiting legal parking or illegal parking/stopping) in the BL or WCL, presence of driveways or intersecting streets, and provision of turn lanes at intersections that typically (but not always) resulted in a narrowing of the BL or WCL at the intersection proper (normally in the last 30 to 50 m before the stop bar). Except for combined BL and parking facilities, these situations did not appear to be related to whether a BL or WCL was present. In other words, the conflicts that resulted were site-specific and likely would have occurred whether a BL or WCL was present.

Clinical Examination of Serious Conflicts

Seventeen conflicts were coded as serious, 10 at WCL and 7 at BL sites. If "fault" had been assigned, 11 would have been the fault of the motorist and 6 the fault of the bicyclist. The motorist turned right soon after overtaking the bicyclist in six of the conflicts (figure 10), pulled from a

driveway to the street in three conflicts, and was involved in a parking situation in the other two cases. The bicyclist turned or

swerved across a lane of traffic in three conflicts, disobeyed a traffic signal in two cases, and shifted in front of a motor vehicle in the process of avoiding rough pavement in the other (figure 11). Examining these situations clinically, there appeared to be no differences between BL and WCL serious conflicts.

Comparisons with Crash Data

One year (1995) of police-reported crash data were "typed" using the NHTSA methodology for all three of the project communities. There were parallels to the videotape data.

In Santa Barbara, one of the two most frequently occurring crash types was the bicyclist striking a parked vehicle. Santa Barbara had a number of individual intersections where parking was part of the bike facility (figure 12), and, overall, 41 percent of the bicyclists were recorded as riding next to parked vehicles, as compared with 21 percent of the Austin bicyclists and none of the Gainesville bicyclists.

In Gainesville, the most frequently occurring crash type was Motorist Drive Out at Stop Sign. In three out of four of these crashes, the bicyclist was riding the wrong way (facing traffic) on the sidewalk. Seventeen percent of the Gainesville bicyclists were observed approaching the targeted intersections on the sidewalk, as compared with less than 3 percent of the Austin bicyclists and less than 2 percent of the Santa Barbara bicyclists. In addition, 9 percent of the Gainesville bicyclists were observed riding the wrong direction on a sidewalk, compared with 1 percent of both the Austin and Santa Barbara bicyclists. Wrong-way sidewalk riding was also a factor in 87 percent of Gainesville's Right Turn on Red crashes, and 75 percent of its Drive Out at Midblock crashes.

In Austin, 11 percent of the bicyclists made "advance crossovers" to the left prior to the intersection (figure 13), as compared with 3 percent in both Gainesville and Santa Barbara. Nearly 6 percent of the crashes reported for Austin were Bicyclist Left Turn in Front of Motorist. None of these types of crashes were reported for Gainesville or Santa Barbara.

Further Comment

Level of Experience

Many in the bicycling community have assumed that more experienced bicyclists tend to use WCLs and that lesser experienced bicyclists use BLs. This issue was explored in this project by use of an oral questionnaire, where each surveyed bicyclist was asked to read or listen to a statement being read to them about their experience or comfort level on certain types of facilities. Overall results showed that 34 percent of the bicyclists considered themselves to be experienced, and there were no differences by type of facility.

Wrong-Way Riding

Wrong-way riding, or riding facing traffic, was present for approximately 6 percent of the videotaped bicyclists. There seems to be a prevailing feeling that this practice is more widespread in BLs, but in this study a higher proportion of the wrong-way riding tended to occur at WCL sites, whether in the roadway or on the sidewalk (figure 14). Proportionally more of the WCL wrong-way riding took place on the sidewalk; however, eliminating sidewalk riding from the tabulation still showed significantly more wrong-way riding in the street associated with WCL sites. This may be related to the fact that WCLs are often associated with higher volume roadways and that maneuvering through intersections on these roadways can be a complex task. Thus, the bicyclist may choose what seems to be a safer route by riding the wrong way on an adjacent sidewalk or in the street. It may not be safer in actuality, as wrong-way riding either in the street or on a sidewalk is a frequent factor in bicycle-motor vehicle crashes (See Hunter, Stutts, Pein, and Cox, 1996).

Turning and Other Maneuvers at Intersections

Besides the sidewalk riding mentioned above, complexity of traffic at the WCL intersections in this study may also be related to the operational findings that more incorrect left-turn destination positioning and pedestrian-style left turns were associated with WCL intersections. In addition, WCL sites had proportionally more non-standard right turns than BL sites. Left turns presented problems at BL sites as well. An intersection conflict model showed higher conflict rates for straight and right turning bicycles where the bike lane was terminated prior to the intersection, dashed to the intersection, or the nominal width of the BL or WCL was narrowed due to the provision of turn lanes. A prevalent conflict in these situations, whether at a BL or WCL site, is for a motor vehicle to pass a bicyclist and then turn right soon after the overtaking maneuver is made. Experienced bicyclists can prevent some of these conflicts by taking control of the lane with their positioning, particularly within the intersection, so that the motor vehicle cannot pass. More bicyclists need training related both to turning maneuvers at intersections and to safely negotiating these areas if merely going straight through. Intersections continue to account for about half of all bicycle-motor vehicle crashes (Hunter, Stutts, Pein, and Cox, 1996).

Conflicts

There were nearly 400 midblock and intersection conflicts noted, but the vast majority were minor in nature. There was no difference in the severity level of the conflicts for BL versus WCL sites as measured by bicycle or motor vehicle response scales to conflicts. Bike/bike conflicts were more associated with BLs, while bike/pedestrian conflicts were more associated with WCLs. Unadjusted conflict rates showed BL sites to have slightly higher rates per entering bicyclist than WCL sites.

Many midblock and intersection conflict models were attempted to identify significant variables related to the occurrence of conflicts. A midblock conflict model showed that presence and width of a BL were significantly related to conflicts, along with traffic volume and presence of driveways. Conflicts increased with traffic volume, number of driveways, presence of a BL, and narrower BLs. The interpretation question was whether the higher conflict rates were really attributable to these variables, particularly narrower BLs, or to site-specific characteristics for a few locations. Further analysis showed that a few sites with narrower BLs and high conflict rates tended to greatly affect the results. This led to a clinical analysis of high conflict rate sites.

Results of this clinical analysis showed several factors to be consistently related to the occurrence of the conflicts: (1) presence of parked motor vehicles (either entering or exiting legal parking or illegal parking or stopping) in the BL or WCL, (2) presence of driveways or intersecting streets, and (3) provision of additional (usually turn) lanes at intersections that typically (but not always) resulted in a narrowing of the BL or WCL. Fortunately, these are factors for which some countermeasures are available.

[Table of Contents] [Chapter 1] [Chapter 2] [Chapter 3] [Chapter 4]