Investigation on the driver-victim pairs in pedestrian and bicyclist crashes by latent class clustering and random forests

Zhu, Chunwu; Brown, Charles; Dadashova, Bahar; Ye, Xinyue; Sohrabi, Soheil; Potts, Ingrid B. · 2023 · openalex_search

DOI: 10.1016/j.aap.2023.106964

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the disproportionate traffic crash fatalities and injuries experienced by marginalized and underserved populations, particularly Black, Indigenous, and People of Color (BIPOC), in the United States. Traditional safety research often treats drivers and vulnerable road users (VRUs) like pedestrians and bicyclists as independent entities, ignoring potential socioeconomic correlations between the two parties involved in a single crash. Motivated by the "close-to-home effect," which suggests drivers and victims often share similar neighborhoods and socioeconomic characteristics, this research investigates the interaction between driver-victim pairs. The study aims to identify crash patterns based on the income and ethnicity of both parties, assess their geographic distribution, and determine the contributing factors shaping these patterns to inform equitable transportation policy. The researchers analyzed four years (2017–2020) of pedestrian and bicyclist crash data from Harris County, Texas, sourced from the Texas Department of Transportation’s Crash Records Information System. After filtering for cases with complete demographic data, the final dataset included 2,822 pedestrian crashes and 1,123 bicyclist crashes. Since income data was not directly available in crash records, the study estimated driver and victim income levels using median household income from their respective residential census tracts, derived from the 2019 American Community Survey. Driver census tracts were matched via ZIP codes, while victim tracts were assumed to be the crash location. The analysis incorporated roadway infrastructure characteristics, traffic exposure metrics (Annual Average Daily Traffic and estimated pedestrian/bicyclist counts), and crash-specific variables. The methodology employed Latent Class Clustering (LCA) to classify crashes into homogeneous groups based on driver and victim socioeconomic traits. Subsequently, Random Forest algorithms and Partial Dependence Plots were used to model and interpret the factors influencing these clusters. The clustering results revealed a pattern of social segregation, indicating that drivers and victims with similar socioeconomic characteristics tend to be involved in the same crashes. The most influential factors contributing to these patterns included pedestrian/bicyclist exposure, driver and victim age, the year of the vehicle in use, Annual Average Daily Traffic (AADT), speed limits, roadbed width, and lane width. Specifically, crashes involving drivers and victims with lower incomes and non-white ethnicity were more likely to occur in locations characterized by higher pedestrian/bicyclist exposure, higher speed limits, and wider roads. These findings highlight that environmental and infrastructural disparities, rather than just individual behavior, play a significant role in the disproportionate crash risks faced by underserved communities. The significance of this research lies in its challenge to traditional safety models that treat crash parties as independent. By demonstrating the correlation between driver and victim socioeconomic profiles, the study underscores the importance of considering spatial and social similarities in crash analysis. The findings provide actionable insights for transportation planners and policymakers to develop targeted interventions that address environmental justice. Improving roadway infrastructure in areas with high exposure and unsafe design features, particularly in low-income and minority communities, is essential for ensuring equitable and sustainable safety for all road users. This approach supports a more nuanced understanding of traffic safety disparities, moving beyond individual blame to address systemic infrastructural inequities.

Key finding

Drivers and victims in pedestrian and bicyclist crashes tend to share similar socioeconomic characteristics, with crashes involving lower-income and non-white individuals occurring more frequently in locations with higher traffic exposure, higher speed limits, and wider roads.

Methodology

dataset

Sample size: 3945

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via scout_discovery on 2026-05-08.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified_with_issues.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• demographic disparities
• vru crash typology
• cyclist safety
• sex gender
• incidence prevalence
• motorcycle crash typology

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: crash risk outcomes