Examining the impact of spatial inequality in socio-demographic and commute patterns on traffic crash rates: Insights from interpretable machine learning and spatial statistical models

Pengfei Cui, Mohamed Abdel-Aty, Chenzhu Wang, Xiaobao Yang and Dongdong Song

Transport Policy, 2025, vol. 167, issue C, 222-245

Abstract: Traffic crash rates are often closely related to the region's socio-demographic, commuting behavior, motivated by the risks associated with increased region density and excessive congestion. However, crash rates of different severities may vary considerably due to socio-spatial disparities and the mitigation behaviors adopted across regions. Thus, this study elucidates the intricate effects of socio-demographic dynamics and commuting behavior on overall and fatal traffic crash rates across Florida's counties, with particular emphasis on the underlying factors of spatial inequality. Employing an interpretable machine learning model, specifically eXtreme Gradient Boosting (XGBoost), we demonstrate its superiority in detecting spatial heterogeneity and the complex effects of various factors compared to traditional spatial statistical models, e.g. Spatial Lag Model (SLM) and Multiscale Geographically Weighted Regression model (MGWR). A comprehensive simulation experiment was designed to validate the dependability of modeled fittings, which confirms XGBoost as a reliable alternative to conventional spatial statistical models, particularly when dealing with datasets that including complex effects including spatial lag, spatial heterogeneity, non-linear effects and potential interaction effects. Furthermore, the totally empirical findings for Florida reveal the spatial variations in overall and fatal crash rates, correlating significantly with socio-demographic and commute pattern variables. An endogeneity test has been conducted initially for empirical datasets, accompanied by strategies to eliminate the biasing effect of endogenous variables on subsequent modeling process. Finally, key variables include population demographics, commute duration, education levels, unemployment rates, and intersection density produce heterogeneous effects on overall and fatal traffic crash rates. Notably, the study dispels the conventional belief that higher overall crash rates directly correlate with higher fatal crash rates in the same regions, underscoring the importance of distinct analysis. Policy measures should focus on improving accessibility to road infrastructure and healthcare services, with tailored approaches for sparse and densely populated areas. These findings underscore the need to address spatial inequalities in transportation infrastructure and policy measures to reduce traffic crash rates and improve road safety across all regions.

Keywords: Traffic crash rates; Socio-demographic; Commute patterns; Spatial inequality; Endogeneity; Simulated experiment; Interpretable ML (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
http://www.sciencedirect.com/science/article/pii/S0967070X25001349
Full text for ScienceDirect subscribers only

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:eee:trapol:v:167:y:2025:i:c:p:222-245

Ordering information: This journal article can be ordered from
http://www.elsevier.com/wps/find/supportfaq.cws_home/regional
https://shop.elsevie ... _01_ooc_1&version=01

DOI: 10.1016/j.tranpol.2025.03.033

Access Statistics for this article

Transport Policy is currently edited by Y. Hayashi

More articles in Transport Policy from Elsevier
Bibliographic data for series maintained by Catherine Liu ().