Crash modification factors for Sonepat using empirical Bayes and negative binomial modelling

This study develops locally calibrated crash modification factors (CMFs) for median width, shoulder width, and minor road junctions in urban and peri-urban corridors of Sonepat, Haryana. Empirical Bayes methods were applied to cross-sectional crash and roadway data to account for regression-to-the-mean bias, while negative binomial regression was used to analyse six years (2017–2023) of verified crash records and roadway characteristics. Wider medians (3.05–12.20 m) were associated with up to 17% lower crash frequencies, whereas increasing shoulder width from 1.22 m to 2.44 m was associated with an approximately 12% reduction in crashes. Stop-controlled and signalised junctions were associated with crash reductions of approximately 21% and 25%, respectively. A Pugh matrix was used to rank alternative roadway configurations by integrating safety performance with implementation feasibility. The configuration comprising a 9.15 m median, a 2.44 m shoulder, and a signalised junction emerged as the preferred alternative. The proposed framework provides a replicable approach for developing context-sensitive CMFs and supports evidence-based roadway design, safety audits, and road safety policy.