ESTIMATING SAFETY EFFECTS OF CROSS-SECTION DESIGN FOR VARIOUS HIGHWAY TYPES USING NEGATIVE BINOMIAL REGRESSION

Improvements in cross-section design are expected to reduce crash rates. Previous studies on the subject have concentrated on two-way, two-lane highways and given less attention to other types of highways. In addition, several of those studies used conventional regression analyses that are not suitable to estimate a discrete non-negative variable like crash frequency. This study uses negative binomial regression analyses to estimate the effects of cross-section design elements on total, fatality, and injury crash rates for various types of rural and urban highways at different traffic levels. The results show that, depending on the highway type investigated, increasing lane width, median width, inside shoulder width, and/or outside shoulder width are effective in reducing crashes. The results also indicate that on four-lane urban highways, the raised median is safer than the two-way left-turn lane median and that the use of an open-graded friction course in lieu of a dense-graded friction course does not have any effect on crash rates.