Skip to main content
You can also sort pages by filters.
Table of Contents
Download the Full Book

Practitioners and researchers need high-quality data to both identify problems and to evaluate countermeasure effectiveness. Three primary types of data are needed for a more complete picture of pedestrian safety:

  • Safety/outcome data that describes crash events or reports surrogate measures.
  • Exposure data that describes the amount of activity in a place or by a group.
  • Contextual data that describes the environment in which travel occurs and can provide insight into potential risk factors associated with crashes.

This document’s focus on safety data should not minimize the importance of exposure data or contextual data. Exposure is a crucial aspect of analyzing crash risks because, all other factors being equal, greater exposure will increase the chance of a crash. High crash figures may simply reflect high pedestrian activity. Alternatively, a corridor or intersection with low crash figures may be a result of actual or perceived danger that dissuades people from using the facility.

Crashes are currently viewed as the most objective and reliable measurements of road safety. However, there are challenges with crash data, such as human error in reporting, unreported crashes, and the length of time it often takes for crashes to be entered into a database (Carter et al., 2017). Only traffic-related crashes involving motorists are reported to authorities and most States have a minimum threshold of injury or property damage for the police department to file crash reports.

Police-reported crash data, when recorded completely, often has details on pre-crash maneuvers, crash dynamics, and crash locations. These details allow researchers to analyze the causes and correlations of crashes more effectively and to help agencies identify targets for treatment. Since 2014 NCSA has used the crash typing framework Pedestrian and Bicycle Crash Analysis Tool (PBCAT)[1] to describe the events and maneuvers that preceded fatal bicyclist and pedestrian crashes (i.e., the crashes included in FARS). In 2021 FHWA released an updated version of PBCAT, PBCAT3, that improves user functionality and offers crash typing logic that supports coding consistency and objectivity. This new version complements the data that States currently collect. It is important to consider on-site field review of behaviors and site-specific characteristics before determining which engineering or behavioral countermeasures are appropriate (Zegeer, Sandt, & Scully, 2008). Another pedestrian and bicycle crash typing schema is the Location-Movement Classification Method, which provides detail about pedestrian or bicyclist’s position relative to the roadway or intersection and theirs and the vehicle’s movements preceding a crash. This crash typing helps identify what measures could be applied in specific locations to reduce crashes (Schneider & Stefanich, 2016).

While crash types are informative, crash type studies do not capture all the conditions that can lead to injury or death for people walking. Predominant crash types may differ by geographies, injury severity, area type, lighting context, infrastructure facilities, vehicle fleet characteristics, and populations using the facilities. Enforcement and safety norms may also affect crash types (Saleem et al., 2018). Prevalent crash types have been found to change over time, even when study methods and areas were consistent (Preusser et al., 2002).

Another consideration when analyzing crash data are that pedestrian crashes (as well as bicycle crashes) tend to be underreported. Underreporting of traffic-related crashes on road rights-of-way likely decreases as the crash severity increases because police are likely to be called to injury and fatal crashes, and the pedestrian is more likely to be transported or seek treatment at a healthcare facility. Many States may not require reporting nor collect off-road or private-road crash records. Non-roadway crashes may, however, constitute a significant portion of pedestrian-related crashes with motorists. In several studies, parking lot and driveway-related crashes represented up to 15% to 25% or more of all reported pedestrian crashes (Stutts & Hunter, 1999a; Thomas & Levitt, 2014). Many people in the United States and worldwide suffer from falls, including falls incurred while avoiding collisions with motorists (Stutts & Hunter, 1999a, 1999b; Sciortino et al., 2005; Methorst et al., 2017). Many more roadway and non-roadway crashes go unreported. Research is needed to better understand the extent and causes of non-roadway pedestrian crashes and effective countermeasures. NHTSA’s Non-Traffic Surveillance (NTS; NCSA, 2023a) monitors and reports on not-in-traffic-related motor vehicle deaths.

Hospital and EMS data can be an important form of safety data, as not all crashes involve police response. These data are usually more accurate than police reported crash data, especially for determining crash severity outcomes. They also may include more information about the nature of an injury and crash than police reports, but rarely include detailed location data. Health-related datasets are often deidentified, which makes it challenging to link them with other datasets (i.e., police-reported crash data). Sometimes linkage is possible by working with individual States or after negotiating data agreements. One source of data is through the National Emergency Medical Services Information System, the national system used to collect, store, and share EMS data from the U.S. States and Territories. It is a collaborative system to improve prehospital patient care through the standardization, aggregation, and utilization of point of care EMS data at a local, State, and national level. Some data on NEMSIS are publicly available and other data can be obtained through requests.

Surrogate measures of safety can be used to identify crash potential, even when a crash has not occurred or been reported. The most common surrogates are conflicts/near misses measured as a sudden change in speed or trajectory/path and observed behaviors such as driver yielding, driver speed, pedestrian use of crosswalk, and more. User perceptions or rankings by roadway users or expert groups can be used in setting performance measures or safety ratings for roadways. The choice of surrogate safety indicators should be context dependent (Johnsson et al., 2018).

Crash and injury data is often the only available form of safety data, but in a small area (or short duration of time) there may not be enough data for proper analysis. Observing interactions and near misses on local streets may be an effective means of understanding where interventions are needed (Cloutier et al., 2017). Research has linked walking (and bicycling) behavior to perceptions of safety, and if certain locations feel unsafe, there may be no pedestrian traffic. Thus, measuring suppressed trips is also important for gaining a more complete understanding of safety problems (Ferenchak & Marshall, 2019).

Considerations for Improving Data

Improving data on pedestrian transportation is a critical need. A research roadmap developed for the AASHTO Council on Active Transportation calls “improving data on pedestrian and bicyclist fatalities” a high priority (Dill et al., 2021). While crash data is the main source of safety data, a comprehensive nonmotorized safety analysis often means being able to access a wide array of data from sources and disciplines. Key to achieving better understanding of safety is improving police reported crash data, improving exposure data, and increasing the frequency of travel surveys.

A consensus report by the Safe States Alliance provides an overview of pedestrian injury surveillance data that could supplement State level crash data or bolster analyses of safety (Injury Surveillance Workgroup 8, 2017). Fatality and injury data and (primarily) proxy measures of exposure from a variety of sources can be tailored to local needs and used in analyses to understand crashes in greater detail and context.


[1] PBCAT: Pedestrian and Bicycle Crash Analysis Tool, version 3. (n.d.).