This paper presents a new method to rank the severity of an impact with a roadside hazard that is based on observable crash data. This method has been incorporated into the third update of the Roadside Safety Analysis Program. The equivalent fatal crash cost ratio (EFCCR) is a dimensionless value that represents the severity of a crash on a scale of zero to unity, where zero represents no chance of injury and unity represents absolute certainty of a fatal injury. The method uses a census of police-reported data, which ideally covers a range of speed limits. The purpose is to determine the severity distribution of crashes in cases in which no events preceded the crashes with the hazard under evaluation and which did not result in a penetration or a rollover. The number of unreported crashes is estimated and added into the severity distribution with the assumption that the number represents crashes that led to property damage only. The average expected crash cost is then calculated and normalized to a reference speed of 65 mph so that it is directly comparable to EFCCR values calculated for other types of hazards. Unlike the earlier subjective severity index method, the new EFCCR method has its basis in observed crash data and uses a systematic approach to calculate crash severities that can be used in benefit–cost and other safety analyses.