Determining Crash Modification Factor Resulted from Shoulder Safety by Creating a Crash Prediction Model by Empirical Bayes

The shoulder width of the main roads has risen by 35% by revising the Highway Geometric Design. In this research, determination of the effect of simultaneously increasing shoulder width and its paving on decreasing crashes in rural two-way two-line highways was targeted. The studies were conducted in Ilam province highways and the experimental empirical Bayes (EB) method was used. Several highways sections with narrow shoulders paved or non-paved were selected as reference groups and crash data, traffic flow, and their geometry was collected and modeled. Three types of crash prediction model (SPF) were created for the total number of crashes, fatal crashes and shoulder related crashes in the cold seasons (autumn and winter) using the negative binomial regression model for the reference group. Some sections of which shoulders had become safety by widening and paving were also selected as an improved group. To estimate the crash modification factor (CMF), the expected crashes in all improved sections were summed up and then compared with the total number of observed crashes in the period after the countermeasure. The EB method was also repeated twice for the improved group of improving the narrow shoulder to the wide and paved shoulder. Once with the SPF for each of the three crash categories, considering the improved group may not be well-matched with the reference group, and once only with the SPF of the total crashes. The results showed that Widening and paving the shoulders affects considerably on the safety of the shoulders related crashes in the cold seasons with CMF of about 0.42 (58% reduction) and the fatal and injuring accidents with CMF of about 0.72 (28% reduction). It can also contribute to a 12 percent improvement of the total crashes of the cold seasons.