This paper analyzes the effect on road safety of 95 roundabouts that were built in Flanders between 1994 and 1999. The study shows that the effect on the number and severity of road accidents adjusted for the trend and regression to the mean is significant, but varies considerably in accordance with the speed limit regime on the intersection. Roundabouts are most effective on intersections of a main road with a high speed limit (90 km/h) and an adjacent road with a lower speed limit (50 or 70 km/h). The empirical analysis reveals a reduction of 34% (varying between 15 and 59%) for the total number of injury accidents, 30% (14-45%) for light injury accidents and 38% (27-72%) for serious injury accidents. This study also takes a closer look at the impact of different post-implementation periods using accident data of one, three or six years after the construction of a roundabout on the calculated effectiveness results and warns for a severe underestimation when a one-year period is used. An effective traffic safety policy based on scientific results thus requires some patience from the policy-makers.Journal of Safety Research 36/3, p. 289-296.