British Rail (BR), the former unified main line railway operator in Great Britain, was divided into about 100 separate organisations and privatised from April 1994. There was concern in the run-up to privatisation that the fragmentation of the system and the entry of new operators might compromise safety. This paper investigates what has happened to safety by analysing data on almost all fatal railway accidents, together with the most important non-fatal train accidents, from 1967 to 2003, with additional brief analyses back to 1946. BR had achieved downward trends in the mean numbers of accidents per train-kilometre for all the main classes of accident in the 27 years up to 1993, and the paper takes the extrapolation of these favourable trends as the yardstick by which to judge the safety performance of the privatised railway. The paper finds that the privatised railway had fewer accidents than this yardstick for all classes of accident. Only one indicator is adverse: the number of fatalities in train collisions and derailments is higher than expected, because of the severity of the accident at Ladbroke Grove in 1999. The principal conclusion is that there is no evidence that privatisation caused railway safety to deteriorate.