Since the last decade, road accident fatalities have been increasing rapidly in Malaysia. As a response, various efforts have been made to reduce the burden of this crisis to the society. However, little research has been conducted to address the effects of road features on the crash frequency, particularly in Malaysia federal road network where it accounts for over 40 percent of all fatalities throughout the country. This paper aims to focus on run-off-the-roadway (ROR) accidents on a 543-km sample of Malaysia federal roads. The data used in this study include records of ROR accidents from 2007 to 2010, as well as data on road geometric, traffic, and environmental attributes. A negative binomial regression model was applied to relate the accident frequency to a set of roadway characteristics. The results indicated that the ROR accidents exhibited overdispersion, and hence, the negative binomial model is an appropriate approach to address the issue. The findings also showed that variables of average daily traffic, heavy vehicle traffic, and unpaved shoulder width significantly influence the incidence of ROR accidents. The modeling process presented in this paper appear to be useful for applications such as identifying road blackspots and suggesting appropriate countermeasure to road-safety authorities for remedial treatment.
Run-off-roadway accident; Crash prediction model; Count-data models; Elasticity analysis