Context-Aware Evaluation of Signal Timing Interventions on Red-Light Running Using High-Resolution Traffic Controller Data

Abstract

Red-light running (RLR) is a persistent traffic safety issue that contributes to thousands of crashes annually. Improperly configured yellow and red-clearance intervals are key contributors to RLR violations, yet existing timing strategies often fail to adapt to real-world variability in driver behavior and traffic conditions. This study develops a novel, data-driven framework that integrates high-resolution Automated Traffic Signal Performance Measures (ATSPM) data with Causal Forest modeling to evaluate the impact of yellow and red-clearance interval adjustments on unintentional RLR events. Using event-level data from eight intersections in Hillsborough County, Florida, we estimate both Average Treatment Effects (ATEs) and Conditional Average Treatment Effects (CATEs) across varying operational conditions. Key findings show that yellow interval extensions are significantly more effective than red-clearance interval changes, with a one-second increase in yellow time reducing RLR violation probability by up to 30 percentile points in high-risk scenarios. A counterfactual simulation further demonstrates that targeting only 37–49% of the highest-risk cycles can capture 100% of the achievable RLR reductions, enabling cost-effective, data-informed deployment of timing interventions. The proposed framework enables adaptive, hour-by-hour, and movement-specific timing recommendations, offering a scalable and practical solution for intersection safety improvement. By supporting data-informed, resource-conscious decision-making, this study advances proactive traffic signal management and RLR mitigation strategies.