Understanding roundabout safety through the application of advanced econometric techniques

Abstract

Abstract Intersections present a significant safety concern, as such in an effort to reduce the more serious injuries occurring at or near intersections, many jurisdictions have turned to implementing roundabouts. Despite the advantages that roundabouts provide, crashes still occur, and less severe crashes are on the rise. The study presented in this paper investigates a crash-based analysis to better understand the factors that may influence less severe crashes to those of more severe crashes given various roundabout configurations and crash types. Using Oregon’s crash database from 2011 to 2015 a series of log likelihood ratio tests were conducted to validate that four separate random parameters binary probit models by configuration type were warranted. The outcome of each tested configuration (full, three & four leg, four leg, and three leg models) shows a major difference in both the combination and variables included in each model and the magnitude of impact of those variables. These differences illustrate that various roundabout configurations (full, three & four leg, four leg, and three leg models) do in fact have different factors highlighting the need to examined crashes at roundabouts by configuration type. Variables related to driver error, weather, alcohol use, barrier conditions, vehicle movement, location of crash, and restraint use were found as key differences between the various tested configurations.