Development of Analysis Methods Using Recent Data

Principal Investigator

Gary Davis, Professor, Civil, Environmental and Geo-Engineering

Co-Investigator

  • John Hourdos, Director, MN Traffic Observatory, Civil, Environmental and Geo-Engineering

Project Summary

In order to more readily and accurately measure crash risk, this research explored the relationship between crash risk and various non-collision traffic incidents (such as near-collisions, conflicts, and roadside encroachments). These incidents, collectively referred to as "surrogate collision measures," are more easily and frequently observed than actual collisions and have been used in the past to assess crash risk, despite the lack of a commonly accepted analytical framework for such an analysis. This project aimed to lay a solid analytic foundation for using conflicts and near-collisions as surrogate measures. Two major components formed the foundation for this effort: first, an explicit description of a class of observable traffic events that do not result in crashes; and second, an implicit assumption that the events in this class are similar to counterfactual events that do result in crashes. In other words, in order to determine whether or not an observed event is a conflict or near-crash, it is necessary to compare that event to similar, counterfactual, events where a crash does occur. If the changes needed to convert the proposed near-crash event to a crash event are small enough, this is a legitimate near-crash. The fact that this counterfactual comparison is left unspecified (and in most cases, unrecognized) is a primary source of confusion and lack of progress on the issue of surrogate measures. Making this counterfactual simulation explicit was expected to lead to the identification of strong, plausible surrogate measures.

Sponsor

  • Strategic Highway Research Program

Project Details

  • Start date: 02/2007
  • Project Status: Completed
  • Research Area: Transportation Safety and Traffic Flow
  • Topics: Safety, Traffic Modeling and Data