Our aim was to ascertain the extent of under-reporting of seriously injured motor vehicle traffic crash victims, as recorded by police in New Zealand, and to what extent this coverage was biased by crash, injury, demographic, and geographic factors. Hospital data and police records were linked using probabilistic methods. During 1995, less than two-thirds of all hospitalised vehicle occupant traffic crash victims were recorded by the police. Reporting rates varied significantly by age, injury severity, length of stay in hospital, month of crash, number of vehicles involved, whether or not a collision occurred, and geographic region, but not by gender, ethnicity or day of the week of the crash. Those using these police files for prioritization, resource allocation and evaluation purposes need to be aware of the extent and nature of these biases contained within these databases.