Estimating Pedestrian Volume for Intersections Using Nationwide-Based Social-Demographic Data

Abstract

Pedestrian volume is among the key predictors when evaluating pedestrian safety. Despite its importance in safety assessment, pedestrian volume data are rarely collected. Thus, researchers estimate pedestrian volume using statistical models. However, such estimations involve small-scale local databases that hinder the transferability of the models. The availability of nationwide socio-demographic databases, such as the Smart Location Database (SLD) and Social Vulnerability Index (SVI) in the U.S., provides an opportunity for large-scale pedestrian volume estimation. This study investigates whether these existing nationwide databases can provide a relatively easy source for estimating pedestrian volume at the intersection scale. This study applied the negative binomial model to estimate pedestrian volume using variables collected from the SLD and SVI database and a few added site characteristics data. The developed model used on-site pedestrian volume data collected in California, U.S. The study found that five SLD variables—zero auto ownership, total employment, employment within 0.5 mi of a fixed-guideway transit stop, tract area, and total road network density—are key predictors of pedestrian volume. Further, three SVI variables—single-parent households with children under 18 years, persons aged 65 and older, and unemployed civilian population—are statistically significant and affect pedestrian volume. Further, the developed model produced a relatively higher performance, as revealed by the R-squared value, and showed a potential for transferability to other states. The prediction equation provided in this paper can be used to provide a reasonable estimate in other states.