Executive summary

Pedestrian–vehicle crashes are both common and deadly. In 2010, 13 percent of all fatal crashes involved pedestrians.⁽¹⁾ Of these, 68.1 percent occurred outside intersections. As a result of the large proportion of pedestrian fatalities that occur at non-intersection locations, it is important to investigate the causal factors of these collisions. Despite the large proportion of crashes, little research has investigated the reasons pedestrians cross roadways at unmarked locations.

As a result, the present study sought to better understand the environmental influences on both where and when pedestrians elect to cross the road. The circumstances surrounding when and where more than 70,000 crossings took place were recorded and analyzed. A model to predict crossing behaviors was created. These data have the potential to guide roadway design. Furthermore, this approach may aid in the selection and location of pedestrian crossing interventions (e.g., new pedestrian activation crossing beacons), ultimately increasing pedestrian safety in shared use environments.

Pedestrian roadway crossings were coded at 20 different locations in the Washington, DC, metropolitan area. Each location was one block in length and was flanked by two marked crosswalks at intersections. Crossings were recorded within one marked, light-controlled crosswalk and the roadway between it and the next marked crossing (but not within the far crossing). Pedestrian crossings were coded for several different factors:

A.    Location. Within the marked crosswalk, or not.

B.     Traffic status. Walk or don’t walk sign illuminated.

C.     Yielding. Pedestrians yielding to vehicles or vehicles yielding to pedestrians in the roadway.

D.    Evasive Actions. Any evasive movement made by a vehicle or pedestrian to avoid collision (e.g., running or abrupt braking).

Stable components of each location were also recorded:

1.      Distance between the marked crosswalks.

2.      Average annual daily traffic volume (AADT).

3.      Street directionality (one- or two-way).

4.      Physical barriers in or along the roadway that might prevent pedestrians from easily crossing between the roadway and sidewalk.

5.      Presence and location of bus stops.

6.      Number of potential pedestrian trip originators/destinations.

7.      Availability of street parking.

8.      Presence of a center turn lane.

9.      Presence of a right turn only turning lane

10.  Length of the walk light phase.

11.  Length of the don’t walk light phase.

12.  Width of the roadway/pedestrian crossing.

13.  Presence and type of median (e.g., raised concrete or painted asphalt).

14.  Presence of a T-intersection between the two marked crosswalks.

15.  Traffic control device of the second crosswalk (i.e., traffic signal, stop sign, or none).

16.  Pace at which pedestrians are required to travel to complete a crossing entirely during the walk light phase.

Data were used to create a model to predict where pedestrians are likely to cross the road (marked intersection crosswalk or non-intersection). The accuracy of the model ranged from 80 to 95 percent based on location. The model correctly predicted a mean of 90 percent of crossings. Overall, the model was successful in predicting whether pedestrians would cross in marked crosswalks at intersections or outside a marked crossing.

A mean of 13.89 percent of pedestrian crossings took place at unmarked non-intersection locations. Given the disproportionate percentage of fatalities that take place outside marked intersections, this number may be a bit surprising. However, these data suggest that some locations are more prone to have more unmarked non-intersection crossings than others. This was indeed the case here. Non-intersection crossings ranged from 3.02 to 36.55 percent.

The location with 36.55 percent of the crossings that took place at a non-intersection was different from many of the other locations in very specific ways. A wide, grassy median separates traffic directionality. This median allows pedestrians to cross one road segment, wait on the median for a gap in traffic, and complete the second portion of the crossing. In addition, the juxtaposition of a Metro (subway) train station and a surrounding neighborhood is such that the most direct route (in terms of absolute distance) between the two areas involves crossing outside the marked intersection. Given that some might consider traveling through the marked crosswalks to be out of the way, many may increase their perceived control of the crossing by using the median and cross midblock.

Environmental factors were also examined in terms of their influence on crossing behaviors. For example, a significant relationship between the width of the crossing and the percentage of pedestrians who crossed entirely during the walk signal light phase at each location was found. In other words, the longer the distance that pedestrians were required to travel to cross the road, the more likely they were to cross entirely during the walk phase of the light cycle. Interestingly, a significant relationship between crossing entirely during the don’t walk signal phase and traffic directionality was found—pedestrians were more likely to cross during the don’t walk phase on one-way streets than on two-way streets.

Not surprisingly, when physical barriers that might prevent pedestrians from easily crossing between the roadway and sidewalk were present, pedestrians were less likely to cross the roadway at unmarked non-intersection areas. Thus it appears that even small barriers, such as flower planters, reduce the perceived affordances to cross the roadway.

Overall, only .98 percent of crossings involved pedestrians yielding to vehicles. Not surprisingly, a significantly greater percentage of crossings in non-intersections involved pedestrian yielding than in marked crosswalks.

Overall, 8.93 percent of crossings involved a vehicle yielding to a pedestrian. A significantly greater percentage of crossings in the marked intersection involved vehicle yielding than crossings in the unmarked non-intersection areas. This discrepancy is largely attributable to turning vehicles yielding to pedestrians crossing in the marked crossings during the walk phase.

Within the marked intersections, a significantly greater percentage of crossings involved vehicle yielding than pedestrian yielding. However, outside the marked pedestrian crossing, pedestrians and vehicles were equally likely to yield to avoid collision.

It is recommended that in new, redesigned, or problematic environments an evaluation of the environmental features should be made to determine where pedestrian crossings are likely. The developed model was successful in predicting an average of 90 percent of the pedestrian crossings. Areas that have a high predicted likelihood of unmarked non-intersection crossings could be proactively targeted to modify the crossing affordances of the environment—leading pedestrians to cross at marked intersections. Presumably this would reduce the number of pedestrians crossing midblock. A combined effort of pedestrian education and shared road use planning would hopefully reduce pedestrian injuries and fatalities and ultimately increase roadway safety.