1. ¹Institute of Criminology, University of Cambridge, UK
2. ²Department of Criminal Justice, University of Massachusetts Lowell, Massachusetts, USA

Correspondence: David P Farrington, Institute of Criminology, University of Cambridge, Sidgwick Avenue, Cambridge CB3 9DT, UK. E-mail: dpf1@cam.ac.uk

Abstract

If an area with a relatively high crime rate is chosen to receive a crime prevention programme, its crime rate is likely to decrease after the programme because of fluctuations in crimes over time caused by fluctuations in influencing factors. This is called "regression to the mean". If this high-crime experimental area is compared with a control area with a lower crime rate, crimes are likely to decrease more in the experimental area than in the control area because of these fluctuations, possibly creating the illusion that the programme had an effect when in fact it did not. This article estimates the importance of "regression to the mean" using recorded crime rates in police Basic Command Units (BCUs) in England and Wales in 2002–03 and 2003–04. It is concluded that, in reasonable comparisons between areas with high and moderately high crime rates, this effect may cause a 4% decrease in crimes. Hence, it could not account for much of the 19% decrease in crimes following improved street lighting.