U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
202-366-4000
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
| REPORT |
| This report is an archived publication and may contain dated technical, contact, and link information |
|
 |
| Publication Number: FHWA-HRT-13-098 Date: January 2014 |
Publication Number: FHWA-HRT-13-098 Date: January 2014 |
The goal of this work was to explore the factors that influence pedestrians’ roadway crossing locations. Factors intrinsic to both humans and the environment were explored. The following describes the findings from literature and phases 1, 2, and 3 of the present study.
A literature review revealed five traits and factors that influence pedestrians’ decisions on when and where to cross the roadway: age, gender, alcohol, self-identity as a safe person, and perceived control. While these influences are intertwined, each is discussed individually here.
Age not only plays a role in the decision on when and where pedestrians might choose to cross the roadway, but also in the outcome of a collision. Children under the age of 15 make up 25 percent of pedestrian injuries, a value larger than any other age group.(7) Given this fact, it may be worthwhile to target special interventions and educational tools to this age group to reduce pedestrian–vehicle collisions. For example, in areas near schools, crosswalk activation buttons may need to be lowered to child level or made visually more attractive to push. Design modifications, coupled with educational tools that incorporate best crossing practices, may be effective in reducing child pedestrian–vehicle collisions.
On the opposite end of the age spectrum, older adults make up about 19 percent of all pedestrian fatalities, but only 8.5 percent of the total injuries.(7) Further, many studies have found that older adults are less likely to attempt dangerous or potentially dangerous crossings.(8,9,10) These data combined suggest that while older adults are less likely to be involved in a pedestrian–vehicle crash, the outcome is likely to be more severe in terms of pedestrian health impact. As such, the safety countermeasures that may be the most beneficial for older adults may not be effective in all areas. For example, in areas near retirement communities, it may be useful to provide for longer crossing times in marked crosswalks or include an opportunity to request (via push button or other movement technology) a longer light protected crossing.
Pedestrian gender has both been associated with both risk crossing and higher pedestrian fatality rates.(6,7,8) These findings are consistent with the general body of literature suggesting that males are more likely to make risky decisions than females.(41) Given the inherent difficultly in designing crosswalks and safety interventions specifically for one gender, time may be better spent on educational interventions. In other words, specific advertisement and educational campaigns can be used to help males better understand crossing risks and the best way to share the roadway as a pedestrian.
As was previously noted, alcohol is involved in a relatively large proportion of fatal vehicle–pedestrian crashes. Obviously, it is not legal to drive while intoxicated. However, walking is generally thought be to be a safe mode of transportation. It is difficult to design pedestrian crossing countermeasures specifically for the intoxicated pedestrian. However, measures can be taken to reduce the fatalities in areas where pedestrians are likely to be intoxicated (e.g., bar districts). For example, railings or shrubbery that separates the sidewalk from vehicular traffic, and road closures might be used. However, these are not likely to be feasible on a widespread basis. As a result, other tactics, such as educational information, public service announcements, or targeted jay-walking enforcement, may reduce less safe pedestrian crossings.
Those people who identify themselves as safe pedestrians are less likely to make risky crossings and accept fewer vehicle gaps as safe for crossing.(10) However, it seems unlikely that one’s self-identity can easily be manipulated. In other words, this is likely a stable characteristic that is not easily changed.
Finally, perceived control has also been shown to influence when and where pedestrians are willing to cross the roadway.(9) When pedestrians perceive more behavioral control of the situation, they are more likely to cross (or intend to cross) the roadway. As a result, it is logical that roadways that present less perceived control (or predictability) to pedestrians are less likely to have a high proportion of crossings outside the marked intersection during the walk light phase (where, presumably, control would be the greatest). Along the same lines then, it may be possible to modify roadway design to reduce perceived control away from marked intersections and increase perceived control near marked intersections through affordance modifications. For example, a median might increase control because it allows pedestrians to break up crossings, completing one segment, seeking refuge, and completing the crossing when ready. Another way to increase control is to allow pedestrians to activate walk signals via push button (with feedback of activation initiation). However, one must also account for unintended consequences of affordance modification. For example, if barriers are placed between the sidewalk and roadway, people parking vehicles may not be able to access the sidewalk and may be forced to walk unnecessarily close to vehicular traffic.
The remainder of this study focused on which specific environmental factors influence where and when pedestrians cross the roadway. Specifically, pedestrian crossings at 20 different locations in the Washington, DC, metropolitan area were recorded. The area of the crossing (marked crossing or unmarked non-intersection crossing), timing of the crossing (with traffic or against traffic), and other crossing circumstances were recorded (e.g., yielding and evasive actions). For these 20 locations, more than 70,000 crossings were coded. Two different methodologies were used to collect these data. One methodology involved two researchers manually counting and coding pedestrian crossings while observing people in vivo. One advantage of this methodology is that the observers can move about and gain better vantage points as needed. However, in many of the high pedestrian volume areas, it is extremely difficult to obtain accurate counts of pedestrian crossings through in-person observations. To be better able to more accurately code and flexibly collect pedestrian data, a second methodology of video recording was used. This allowed the video to be coded in both slowed and speed motion to maximize accuracy and efficiency. The use of the DDOT traffic management cameras provided a simple and unobtrusive way to capture video. Along with the benefits of this data collection methodology, there were also some challenges. The first challenge is that the primary use of the cameras is for traffic management. As a result, the cameras can be moved at any time and may not be capturing footage from the desired areas. Furthermore, there can be service disruptions with the video equipment (e.g., recording failure, camera failure). Despite these issues, video-based data collection proved to be both fruitful and reliable.
Coded data from all 20 locations were combined to create two separate crossing prediction models.
The first model predicts which area of the road pedestrians will cross (marked intersection or unmarked non-intersection) based on the features of the roadway environment. Although many factors were examined, not all were incorporated into the model. The roadway environment variables that were included in the model fell into three categories:
1. Travel Pace and Phasing.
a. Length of the walk phase.
b. Length of the do not walk phase.
c. Travel pace.
2. Traffic Throughput.
a. AADT.
b. Traffic directionality (one- or two-way street).
c. Curb-to-curb distance (i.e., the width of the street).
3. Distance to Safety.
a. Distance to the next marked crosswalk.
b. Presence and type of median.
c. Presence of cross streets between marked crosswalks.
The mean probability of the model correctly predicting a crossing in the marked intersection was .9079. In other words, across the 20 locations, the model correctly predicted an average of 90.79 percent of the crossings. The median prediction accuracy was 91.28 percent with a range of 80.55 percent at Location 8 to 95.22 percent at Location 18. Overall, the model was able to successfully predict crossing location (marked intersection versus unmarked non-intersection) using the features of the roadway environment.
It is likely that many of the pedestrians observed making crossings who did not cross in the unmarked non-intersections did not do so simply because the trip did not require a crossing in that specific direction. In other words, pedestrians may not have made an east/west non-intersection crossing because their travel origination and destination only required a north/south path. As a result, general rule-breaking was also examined. Rule-breaking is any crossing that does not take place entirely during the walk light phase in the marked intersection.
Although many factors were examined, not all were incorporated into the model. The roadway environment variables that were included in the model fell into five categories:
1. Travel Pace and Phasing.
a. Length of the walk phase.
b. Length of the do not walk phase.
c. Travel pace
2. Traffic Throughput.
a. AADT.
b. Traffic directionality (one or two way street).
c. Curb-to-curb distance (i.e., the width of the street).
3. Distance to Safety.
a. Distance to the next marked crosswalk.
b. Presence and type of median.
c. Presence of cross streets between marked crosswalks.
4. External Objects (barriers/vehicles) in Center of Road.
a. Presence of physical barriers that might prevent a pedestrian from crossing the roadway.
b. Presence of a center turning lane.
5. External Objects (vehicles) on Sides of Road.
a. Presence of parking along the roadway.
b. Presence of a right turn only lane.
c. whether or not the far marked crosswalk is light controlled.
The mean probability of the model correctly predicting a rule-following crossing in the marked intersection was .9079. In other words, across the 20 locations, the model correctly predicted an average of 90.79 percent of the crossings. The median prediction accuracy was 91.28 percent with a range of 80.55 percent at Location 8 to 95.22 percent at Location 18. This model was more successful in predicting rule-following crossings in some locations than in others. However, overall, the model was able to successfully predict rule-following crossings using the features of the roadway environment.
To better accommodate the different overall numbers of pedestrian crossings at each of the crossing locations, a series of analyses were performed using proportions of crossings. These proportions were used to examine both crossing type and environmental factors on crossings.
Overall, the mean percentage of the crossings that took place in the marked intersection was 83.88. (This value was derived by calculating the percentage of crossings that took place in the marked intersection at each location and then these values were taken together to calculate the mean.) The range was 50.88 percent (at Location 3) to 96.98 percent (at Location 7) with a median value of 88.5 percent.
Location 3 is an outlier in this group at 2.70 standard deviations below the mean. Further, when its value is removed, the mean percentage of pedestrians crossing in the marked intersection jumps to 85.61. Yet, when examining the percentage of pedestrians that crossed in unmarked non-intersections, Location 3 does not stand out. In fact, pedestrians were approximately equally likely to cross at the unmarked intersection as the marked intersection. This finding suggests that pedestrians perceive the unmarked intersection to be a safe and acceptable place to cross the roadway. This is likely the result of a few unique characteristics of Location 3. Between the two marked intersections, there is a T-intersection that is light controlled for vehicular traffic (but not for pedestrian traffic). The light phases allow pedestrians travelling to and from the side street to begin crossing, wait on the median, and then complete the crossing when the vehicle light phase changes or an acceptable gap is presented. Given that 41 percent of the crossings at the unmarked intersection involved waiting on the median, it appears that pedestrians were not trapped on the median as previous research has reported. Rather, at this location, pedestrians plan their crossing in phases—crossing a segment, waiting on the median, and then completing the crossing. This is a tactic that presumably increases perceived control. Beyond the ability to divide the crossing into two portions, there are environmental factors that both encourage crossing at the unmarked intersection and discourage traveling to the marked intersection.
This path is likely desirable to pedestrians for several reasons. First, the path to/from the neighboring Metro stop (about 1 block west) and the neighborhood area on the side street is more direct when crossing at the unmarked intersection. Because Rhode Island Avenue Northwest is a diagonal street, pedestrians are required to travel further south and then back north if they elect to cross at the marked intersection. In other words, more travel distance is required. Furthermore, the marked crossing is multipart with variant timing, which means that pedestrians may need to wait on a median/refuge island to complete a crossing, independent of which intersection is used for crossing in this area. To pedestrians, these two factors combined may outweigh the potential benefits of crossing during a protected light phase—especially given the relatively rare occurrence of a vehicle–pedestrian collision. As a result, pedestrians are likely to select to cross at the unmarked intersection as a result of both convenience and a likely perception of more control. Beyond this, the unmarked intersection has several factors that afford a crossing. It is obvious that this is at an intersection; locations where most marked crossings take place. Further, this lies at a junction where it is natural to want to travel. There is Metro station only one block west of the marked intersection, on the north side. As a result, crossing at the unmarked intersection when traveling to/from the Metro station, along Marion Street, is likely the most direct and efficient route. Finally, to the pedestrian, the median in the unmarked intersection looks as though it is a sidewalk (see figure 8). Pedestrians can clearly see a concrete area on the end of the median that is approximately the width of a standard sidewalk (e.g., a firm, raised surface that serves as a barrier from roadway vehicles). As such, this area affords the same things to pedestrians as a standard sidewalk. It is likely that pedestrians interpret this area as they would any other sidewalk area.
Based on these findings, many recommendations can be made for the design of new roadways or the implementation of roadway environment modifications if crossing location (and timing) is a consideration. First, it appears that pedestrians will treat intersections (whether marked with a crosswalk or not) as an acceptable place to cross the roadway. Given sufficient pedestrian volume, protected pedestrian crossings should be taken into account when designing intersections. This is especially important if the intersection is in route to/from a high-density trip originator.
A mean percentage of 13.89 of crossings occurred in unmarked non-intersection areas. This value is especially low considering the percentage of pedestrian fatalities that occur at non-intersection locations (approximately 64 percent). The discrepancy between these two percentages could be the result of several factors. The first possibility is that pedestrians in the Washington, DC, metropolitan area are fundamentally different than other pedestrians. However, this seems unlikely. Another possibility is that because pedestrians infrequently cross outside intersections, drivers do not anticipate pedestrians in these areas. Consequently, drivers are not able to react quickly enough to avoid collision. Furthermore, traffic speeds are most often higher at non-intersections than intersections, thus the outcomes of collisions are more severe.
Rule-following and rule-breaking crossings were explored. A mean percentage of 70.89 pedestrians crossed entirely during the walk phase of the marked intersection. That is, most pedestrians do complete rule-following crossings. Location 13 was an outlier with only 28.41 percent of the crossings completed being rule following. This value is not surprising considering the travel pace required to cross during the walk phase. The walk signal is only illuminated for 10 s and the roadway is approximately 50 ft wide. This requires pedestrians to travel at a rate of 5 ft/s, which is outside the range of comfortable travel for many people. This rate is greater than the MUTCD recommended rate of 3.5 ft/s. Furthermore, the MUTCD states: “Where pedestrians who walk slower than 3.5 feet per second, or pedestrians who use wheelchairs, routinely use the crosswalk, a walking speed of less than 3.5 feet per second should be considered in determining the pedestrian clearance time.” If one is concerned with rule-breaking crossings, the length of the walk phase should be long enough to allow most pedestrians to cross at a reasonable pace and at minimum, meet the MUTCD standards for time provided to cross the roadway.
Analyses also revealed relationships between environmental factors and where pedestrians crossed the roadway. First, a relationship between the width of the crossing and rule-following was found. The wider the crossing, the more likely pedestrians are to make a rule-following crossing. Or, alternatively, pedestrians are more likely to make a rule-breaking crossing when the road is narrow. In addition, the presence of physical barriers along the sidewalk influenced where pedestrians crossed the roadway. Specifically, the more objects that prevent easy access from the sidewalk to the roadway, the less likely pedestrians were to cross at unmarked non-intersection locations. Given these findings, for areas where there might be a problem with dangerous non-intersection crossings, a simple intervention of increasing the barriers along the sidewalk could be made. Placing obstacles like flower planter beds, benches, or other decorative barriers will not eliminate the ability to cross the roadway. However, these barriers will reduce the affordance to cross outside the marked intersection (and maybe even increase the visual esthetic of the area).
Somewhat interestingly, traffic directionality was significantly negatively correlated with crossings made in the marked intersection entirely during the don’t walk phase. In other words, pedestrians were more likely to cross entirely during the don’t walk light phase signal on one-way streets than on two-way streets. This finding makes sense given that pedestrians are better able to predict traffic that is approaching from one direction rather than two. While changing roadway directionality may not be an appropriate roadway safety intervention, it should be considered when designing/modifying existing road segments. For example, if it is known that pedestrians are more likely to cross a one-way street during the don’t walk phase, the speed could be modified to reduce the impact of potential conflicts or pedestrians could be provided with an activation button that triggers the walk phase (increasing perceived control of the crossing).
Of the total crossings at each location, there was a mean percentage of .98 that involved pedestrian yielding. Although pedestrian yielding is discussed, it should not be taken as an indication that there was potential for a collision. Rather, it was simply an action taken by pedestrians to allow vehicles to pass. In the marked intersections, the mean percentage of pedestrians that yielded to vehicles was .41. Similarly, the mean percentage of pedestrians yielding to vehicles during rule-following crossings was .43. Location 17 was an outlier in both cases, with 6.52 percent of the crossings in the marked intersection involving a pedestrian yielding to a vehicle. All of these yielding behaviors were observed while pedestrians crossed entirely during the walk phase, resulting in yielding in 7.27 percent of the rule-following crossings. The exact circumstances of these yielding behaviors were not recorded. However, observational notations indicate that these actions are the result of pedestrians yielding to vehicles turning on to and from the busier main street (North Washington). In other words, pedestrians allowed vehicles to pass, which increased vehicular traffic flow. While patterns such as these show common roadway courtesy and may increase the overall level of user satisfaction, they may lead to unforeseen consequences. For example, if a group of pedestrians yields to a vehicle, the vehicle may proceed to turn to cross the path of the pedestrians. This is not an issue if all pedestrians yield. However, an incident may occur if another pedestrian does not yield to the driver, who is assuming that all pedestrians are yielding.
Overall, Location 10 was an outlier with 13.25 percent rule-breaking crossings involving yielding to a vehicle. Additional notations indicate that some of these yielding behaviors may have been the result of a person parking on one side of the street and seeking a trip destination on the opposite side of the street. That is, pedestrians began crossing somewhere between the two marked intersections and ultimately yielded to a vehicle before completing the crossing. If parking away from a marked intersection does indeed lead people to cross at unmarked non-intersection crossings to reach destinations on the opposite side of the road, it may be difficult to implement interventions to promote crosswalk use. However, items such as barriers that create separation between bi-directional traffic and reduce crossing affordances may be effective (although they have many drawbacks, including cost and feasibility).
As one might expect, there were significantly more instances of pedestrians yielding to vehicles in unmarked non-intersection areas than in marked intersections. Similarly, there were significantly more pedestrian yielding behaviors in rule-breaking crossings than rule-following crossings. These findings suggest that when pedestrians do make rule-breaking crossings, they are prepared to yield to vehicles.
Pedestrian yielding is also significantly correlated with several environmental factors. There was a significant correlation between the length of the walk signal light phase and the percentage of pedestrians that yielded to vehicles in the marked intersection. This is perhaps not surprising. If one thinks about a moderate to heavy volume crossing, pedestrians often group together as they wait for the signal to change. After the initial bulk of pedestrian traffic has crossed the intersection, queued vehicles begin to make turns onto perpendicular streets. As pedestrians, who had not been waiting to cross, reach the intersection, they may yield to allow the vehicles to complete crossing. This serves both to keep vehicular traffic flowing and to provide pedestrians with some control of their safety in the crossing.
Traffic volume, as estimated by AADT, was also significantly correlated with pedestrians yielding during rule-breaking crossings. This is not surprising given that as vehicular traffic increases, there are more opportunities for pedestrians to yield to vehicles. Given these data, it is difficult to say that areas with higher AADT are necessarily more dangerous for pedestrians. Rather, pedestrians are simply more likely to yield to vehicles, which may actually indicate a safer area for pedestrians.
Overall, an average of 8.93 pedestrian crossings involved a vehicle yielding to a pedestrian. Location 7 was an outlier, with 38.69 percent of the crossings involving vehicle yielding. This trend remained consistent with both crossings in the marked crossing (39.80 percent) and rule-following crossings (50.71 percent). This area had a substantial number of vehicles turning right and passing through the intersection of interest. As a result of the turning and through traffic having a green light (but not a protected right turn) and the pedestrian signal in the walk phase, vehicles often waited to complete their turn (i.e., yielded) while pedestrians crossed the roadway. These yielding behaviors do not necessarily indicate a safety concern. However, turning traffic in high pedestrian volume areas should be evaluated in intersection design.
Overall, vehicles yielded to 2.55 percent of pedestrians making rule-breaking crossings. Both Location 17 (11.67 percent) and Location 12 (9.23 percent) were outliers. Location 12 is unusual because there were very few rule-breaking crossings overall. As a result, the six observed vehicle yielding behaviors resulted in a high percentage of the total rule-breaking crossings. Location 17 however, is unusual in a different way. This block is in an area known as Old Town Alexandria. In this particular neighborhood, the street is relatively narrow, has parking on both sides of the street, has slow-moving traffic (often looking for parking), and is surrounded by small shops and businesses. This is an area where people move at a leisurely pace. These factors combine to create a unique atmosphere where drivers often stop and “wave” pedestrians to cross the road. Although these specific actions were not measured, they were observed and noted on many crossings. This vehicle yielding in this area may suggest that cultural differences in neighborhood may influence the need (or lack of a need) for different safety interventions.
Overall, significantly greater percentages of vehicles yielded to pedestrians crossing in the marked crossing than in the unmarked non-intersection. The same pattern is true with rule-following and rule-breaking crossings.
A relationship between the number of trip originators and destinations in an area and vehicles yielding to pedestrians making rule-breaking crossings was found. That is, the more trip originators and destinations present, the higher the percentage of pedestrian rule-breaking crossings that involved vehicle yielding. These findings suggest pedestrians may be more motivated to make potentially dangerous rule-breaking crossings to reach a trip destination. It also points out the necessity to take special consideration of pedestrian crossing facilities in areas with a high density of commercial businesses.
Not surprisingly a significantly greater percentage of vehicles than pedestrians yielded in rule-following crossings. The same is true for crossings taking place in the marked intersection overall. These findings suggest that drivers are respectful of shared road use with pedestrians in the marked intersections. However, it is not known whether this relationship remains true for non-intersection marked crossings (i.e., not light-controlled marked midblock crossings). This specific type of crossing was not explored in the present study. However, vehicles and pedestrians were equally likely to yield in unmarked non-intersection areas. The same remains true in rule-breaking crossings overall.
It is perhaps a bit surprising that there were no significant differences in vehicle and pedestrian yielding outside the marked intersections and in rule-breaking crossings. It might be expected that pedestrians would take a more proactive safety approach when making a rule-breaking crossing. The lack a difference in yielding may also provide insight into the disproportionate percentage of crossings at non-intersection locations and high fatality rates in these areas. This may also indicate that vehicles will yield proactively to allow pedestrians to cross outside the normal rule following crossings and that pedestrians expect vehicles to yield. Both have implications for pedestrian safety and potential collisions.
The mean percentage of crossings that included an evasive pedestrian action was 3.06. Both Location 1 and Location 8 were outliers. Location 8 requires pedestrians to travel at a pace of 3.7 ft/s, faster than the recommended rate of 3.5 ft/s. Given that there is an elementary school approximately two blocks away from this intersection and that there are other small suburban type establishments (including a church, neighborhood market, and library) in the general area, it is likely that there is substantial pedestrian traffic near this intersection that regularly travels at a rate of less than 3.5 ft/s.
While pedestrians were not specifically queried about their crossings, it is possible that pedestrians may feel rushed while crossing at the marked intersection. This rapid pace required to cross at the marked intersection during the walk phase may lead pedestrians to feel hurried and uncomfortable crossing at this location. Furthermore, if pedestrians are not able to complete the crossing during the walk phase, they may be forced to take an evasive action to complete the crossing. This intersection is also just outside a traffic circle. This prevents pedestrians from being able to view traffic from a distance adequate to determine whether it will continue traveling within the traffic circle or exit toward the intersection. Furthermore, this reduces pedestrians’ abilities to confidently determine whether the vehicle will cross their potential path during a crossing. At this location, pedestrians may begin a crossing but not be able to clearly view all vehicular traffic. As a result, pedestrians might be forced to take evasive actions to complete the crossing safely.
Location 1 was an outlier both in terms of evasive actions in marked intersections and unmarked non-intersections. At Location 1, there were many instances of “courtesy” acceleration by pedestrians. In other words, it appeared that pedestrians would run, or accelerate, through a crossing to allow turning vehicles to complete their turn during the signal phase. This turning traffic may have also been the cause of the evasive actions made outside the marked intersection. Pedestrians may have started crossing the roadway while no traffic was visible. However, after the pedestrian initiated the crossing, a vehicle could have turned onto the main roadway and forced the pedestrians to take an evasive action to avoid a potential collision. These circumstances were not specifically coded in the data; however, several notes indicated that this did in fact occur. It may not be possible to target a specific intervention for these potential sources of conflict. However, vehicle turning phases (or lack thereof) and pedestrian density should be considered in roadway design.
As one might expect, pedestrians had a significantly greater percentage of crossings in unmarked non-intersection areas that involved an evasive pedestrian action than crossings in the marked intersection. This same pattern is true for rule-following and rule-breaking crossings. In other words, when pedestrians make a rule-breaking crossing, they are more likely to need to take an evasive action to avoid collision.
A significant negative relationship between the percentage of evasive actions taken in unmarked non-intersections and the length of the walk light phase was found. That is, the shorter the walk phase, the more likely that crossings in the unmarked non-intersection will involve an evasive pedestrian action. Thus it appears that light phasing may be an important factor to consider in terms of potential vehicle/pedestrian conflicts. However, light phase timing is also associated with traffic density. As a result, simply increasing the length of the walk signal is not likely to eliminate all pedestrian evasive actions.
Overall, the mean percentage of pedestrian crossings that involved an evasive vehicle action was .10. Location 12 was considered an outlier in the percentage of crossings involving an evasive action, both overall and unmarked non-intersection areas. However, only two vehicle evasive actions were recorded at this location. Given the small number, these crossings are not discussed further.
Location 17 is also an outlier in terms of the percentage of evasive vehicle actions made during rule-following crossings and crossings in the marked intersection overall. However, Location 17 only had one recorded evasive vehicle action. As a result, it is not discussed further here.
No significant difference between the percentage of crossings in the marked intersection and the unmarked non-intersection that included a vehicle evasive action was found. No difference in the percentages between rule-following and rule-breaking crossings was found either.
A significant correlation between the length of the walk phase and the percentage of evasive vehicle actions during marked intersection pedestrian crossings was found. In other words, the longer the walk phase, the more likely vehicles were to take an evasive action while pedestrians crossed in the marked intersection. Given the relationship between pedestrian yielding in the marked intersection and the length of the walk phase, this may not be surprising. As discussed in a previous example, it is simple to imagine how pedestrians might group together to wait for a signal to change at a moderate to heavy volume crossing area. After the initial bulk of pedestrian traffic has crossed the intersection, queued vehicles begin to make turns onto perpendicular streets, passing through the marked pedestrian crossing. However, pedestrians who reach the intersection later may come as a surprise to vehicles. As a result, drivers may be forced to make an evasive maneuver to avoid collision.
For crossings made both in the marked intersection and the unmarked non-intersection, pedestrians had a significantly greater percentage of evasive actions than did vehicles. The same pattern is true for rule-following and rule-breaking crossings. This difference can be interpreted in many different ways. In general, it appears as though pedestrians proactively take action to avoid potential collisions. That is, independently of whether pedestrians are making a rule-following or a rule-breaking crossing, they are more likely to take an evasive action to avoid collision than are drivers. Challenges may arise, however, when pedestrians are not able to take an appropriate evasive action, are not aware that one is needed, or when a driver anticipates a pedestrian will take an evasive action and he or she does not.
