Safety Effects of Widening Shoulders on Rural Multilane Roads: Developing Crash Modification Functions with Multivariate Adaptive Regression Splines

Abstract

The presented study used crash modification (CM) factors and CM functions to assess the safety effectiveness of widening shoulders. The study first evaluated CM factors for widening shoulders on rural multilane roadways with the empirical Bayes (EB) observational before–after method to check the overall safety effects. Second, the CM factors were calculated for each aggregated break point according to roadway characteristics such as the original shoulder widths of treated sites in the before period and the actual increased widths. Finally, CM functions were developed with multiple linear regression and multivariate adaptive regression splines (MARS) models to determine the variation of CM factors. The results indicated that widening the shoulders resulted in a 12% crash reduction for total crashes (all severities), 18% for injury crashes, and 21% for severe crashes on rural multilane roadways for all types of crashes. For run-off-roadway crashes, the overall crash reductions of the treatment were 25% for total crashes, 28% for injury crashes, and 31% for severe crashes. Moreover, the results showed that the safety effects varied across sites, depending on roadway characteristics. In particular, the original shoulder widths of treated sites in the before period and the actual increased width were significant parameters that affected the variation of CM factors. Moreover, it was found that the MARS models showed more reliable estimates than normal regression if the variation of CM factors with specific parameters had nonlinear relationships and interaction effects. The proposed CM functions provide general insights into roadway design and selection of sites for increasing shoulder width for reducing crashes.