Statistical Causal Analysis of Freight-Train Derailments in the United States

Abstract

Freight railroads contribute to the national economy by moving over 40% of intercity ton-miles of freight. Meanwhile, train accidents can damage infrastructure and rolling stock, disrupt operations, and possibly cause casualties and harm the environment. Understanding major accident causes is the first step in developing and prioritizing effective accident prevention strategies. The literature has predominantly focused on nationwide train accident cause analysis, without accounting for possible variation in accident cause distributions by railroad and season. This research develops a log-linear statistical model that can estimate the number of freight-train derailments accounting for railroad, accident cause, season, and traffic volume. The analysis shows that broken rails and track geometry defects are the two leading freight-train derailment causes on four major U.S. freight railroads. Fall and winter appear to have a higher likelihood of a broken-rail-caused derailment than spring and summer, given the same railroad and traffic level. By contrast, track-geometry-defect-caused derailments occur more frequently in spring and summer than in fall and winter, given all else being equal. The statistical modeling techniques in this paper can be adapted to other types of train accidents or accident causes, ultimately leading to the prioritization of train safety improvement resources on various spatial and temporal scales.