Kangaroo–vehicle collisions are frequent on Australian highways. Despite high economic costs, detrimental effects on animal welfare, and potential impacts on population viability, little research has been done to investigate the impact of road mortality on kangaroo populations, where and why accidents occur, and how the collisions can be mitigated. We therefore collected data on species (Macropus rufus, M. giganteus, M. fuliginosus, M. robustus), sex and age of kangaroos killed on a 21.2-km bitumenised section of outback highway over 6 months in far western New South Wales, Australia. The spatial and temporal distribution of road-killed kangaroos was investigated in relation to the cover and quality of road-side vegetation, road characteristics, the density of kangaroos along the road, climatic variables and traffic volume. A total of 125 kangaroos were found killed on the road at a rate of 0.03 deaths km^–1 day^–1. Grey kangaroos of two species (M. giganteus, M. fuliginosus) were under-represented in the road-kill sample in comparison with their proportion in the source population estimated during the day. No bias towards either sex was found. The age structure of road-killed kangaroos was similar to age structures typical of source kangaroo populations. Road-kills mainly occurred in open plains country. In road sections with curves or stock races, road-kill frequencies were higher than expected. Greater cover and greenness of roadside vegetation at the verge probably attracted kangaroos to the road and variation in this vegetation affected the spatial distribution of road-kills. The temporal distribution of road-kills was positively correlated with the volume of night-time traffic. The probability of a kangaroo–vehicle collision increased exponentially with traffic volume. Results are discussed in relation to the potential for mitigation of kangaroo–vehicle collisions.