Numerous evaluation studies have reported large accident reductions when road accident blackspots are treated. A critical examination of these studies reveals that many of them do not account for the effects of well known confounding factors, like the regression-to-the-mean effect that is likely to occur at road accident blackspots. This paper shows that the more confounding factors evaluation studies account for, the smaller becomes the accident reduction attributed to blackspot treatment. Studies that account for both regression-to-the-mean and a possible accident migration to neighbouring untreated sites do not show any net accident reduction at all. This tendency conforms to the so called Iron Law of evaluation studies, which states that the more confounding factors an evaluation study accounts for, the less likely it is to show beneficial effects of the programme evaluated. Possible explanations of accident migration are discussed in the paper.