The Effectiveness of Average Speed Cameras in Great Britain

  • Published: RAC Foundation & Road Safety Analysis, 2016
  • Authors: Owen, G. Ursachi, R. Allsop,
  • Date Added: 19 May 2017
  • Last Update: 19 May 2017
  • Format: pdf


The research objective of the study was to establish levels of occurrence of collisions before and after ASC installation (with consideration given to site-selection period, pre-installation and post-installation periods).


The research introduced an independent methodology for reviewing site boundaries and the collisions that have taken place within them since 1990. Using the official Department for Transport collision records, it has been possible to create, on a month-by-month basis, the collision history for each site. These outputs have been used to review the effectiveness of ASCs in reducing collisions at the combined sites, applying a statistical model adopted by Professor Richard Allsop, in a form adapted from that used in the study of spot speed camera data.

Key Findings:

  • A 36.4% (95% confidence interval: 25-46%) reduction in the mean rate of FSCs was estimated in the post-installation period. The change in PICs was lower, with a 16% (95% confidence interval: 9-22%) reduction; both results classified as highly statistically significant according to the model. These results allow in part for any RTM through the removal of SSP data from the pre-installation period. They also take into account the ‘trend’ data from the comparison sites. The other effect estimated in the model is the level of collisions in the SSP relative to the level in the rest of the pre-installation period. The results here show an increase in FSCs of 24.9%, and 16.7% for PICs. This supports the view that the SSP typically exhibits higher-than-normal collision numbers; again, both results were highly significant when tested in the model. It should be borne in mind that the SSP effect has already been accounted for in the installation effect analysis.

  • For FSCs the ASC installation effects at low- and high-speed sites were estimated reductions of 42.2% and 32.3% respectively, both being highly significant. The difference in the two results in itself was not significant, and could well have arisen from random variation. The PIC installation effect at low-speed sites was strong, with a 25% reduction at a high level of significance. The results for high-speed sites was lower at 7.9%, but this was statistically significant only at the 20% level and thus may have arisen through random variation.

  • For the low-speed sites both the FSC and PIC results were statistically insignificant. The estimated increase of 9% (for FSCs) and 5% (for PICs) compared to the rest of the pre-installation months could therefore have happened through chance. The results at high-speed sites were significant, and display increases of 30.2% for FSCs and 21.8% for PICs in the SSP compared to other pre-installation periods.

  • For sites installed for non-collision-reduction reasons, the estimated FSC reduction of 20% was not statistically significant because of the wide difference between reductions at the two sites, although the 24.2% PIC reduction was highly significant when tested in the model. However, comparison of the 95% confidence intervals for these two estimated reductions with those for the corresponding reductions for the sites installed for collision-reduction reasons provides no evidence that the reductions in collisions at these two sites differ from the reductions at the other 25 ASC sites that were selected based on a high collision record.


  • The results show that ASC systems are effective in reducing collisions, especially those of a high severity. Even after allowing for the effects of trend and regression to the mean, highly significant reductions are noted. There is no evidence for the existence of any optimum speed limit that leads to the installations achieving greater collision reduction – they appear to be as suitable for deployment in higher speed limits as in lower ones.Speed-limit enforcement, Average speed cameras, Regression to the mean, Site-selection period.