Understanding the effect of traffic congestion on accidents using big data

Abstract

Understanding the temporal and spatial dynamics of traffic accidents are a key determi-nant in their mitigation. This article leverages big data and a Poisson model with fixed effects to understand the causality of traffic congestion on road accidents in ten cities in Latin America: Bo-gota, Buenos Aires, Lima, Mexico City, Montevideo, Rio de Janeiro, San Salvador, Santiago, Santo Domingo, and Sao Paulo. Analyzing over 10 billion observations in 2019, results show a positive non-linear causality of congestion on the number of accidents. Overall, the results suggest that a 10% reduction in traffic delay would reduce accidents by 3.4%, equivalent to over 72 thousand traffic accidents. Sao Paulo and Mexico City would be particularly benefited, with reductions of 5.4% and 4.7%, respectively. The results of this paper aim to support policymakers in emerging economies in implementing measures to reduce congestion and, with it, the related direct and indirect costs borne by societies.