Thames Valley Speed Cameras

An independent report        



A brief overview




What makes this speed camera report different to official speed camera reports?


Official reports consistently find large reductions in the number of people killed or seriously injured (KSI casualties) at sites after speed cameras were deployed, compared to a baseline period before. These reductions, though, were not caused by the speed cameras. The largest UK government report estimated that KSI casualties would have reduced substantially anyway, without any speed cameras, largely as a result of regression to the mean (RTM). No official report, though, has managed to either measure RTM effects (rather than just estimate them) or fully exclude RTM effects from the final results. Therefore official reports have not established what effect the actual speed cameras are having on road safety.


The report presented here is different for 2 reasons. Firstly, it contains measurements of the RTM effect at speed camera sites and, secondly, the effects of RTM have been fully excluded from the final results. Neither of these have been achieved before therefore this report is a "world first" and should be the most accurate report on the effects of speed cameras to date.


So what is RTM (regression to the mean)?


RTM occurs because speed cameras are often introduced at sites following "unusually" high collision rates that would reduce back to the "usual" (or mean) rate anyway, with or without speed cameras. This report uses two new methods that make use of graphs but the report starts by using the method that is standard practice in most official reports and finds:


"There was a 42% reduction in the number of people killed or seriously injured at sites in Thames Valley after mobile speed cameras were deployed (the first 3 years of operations, compared to a 3 year baseline period before)."


What happens, though, when the collisions that resulted in those casualties are displayed on a graph?

Figure 8.1 shows that there were significantly fewer KSI collisions in the first 3 years of mobile speed camera operations, than there were in the 3 year baseline period before (from 11 to 16).


This is where the 42% reduction in KSI casualties comes from but figure 8.1 also shows that the entire reduction had already occurred a full year before speed enforcement actually started.


Fairly obviously, speed cameras cannot travel back in time and prevent collisions when they weren't there, so what caused the reduction a year before they started operating?


Furthermore, the rate of KSI collisions in the year after the SSP was very similar to the rate in the 5 years before the SSP. Why were the rates so similar either side of the SSP?


Perhaps most striking of all, what could cause such a sudden increase in collisions at the start of the SSP, and would also cause such a similar and sudden reduction at the end?

The answers might be found in how the sites were selected:


When selecting new speed camera sites, the authorities (such as councils and partnerships) might plot the most recent KSI collisions (those during their chosen SSP) on a map of the area and look for clusters. Sites with clusters of KSI collisions might then be considered to be "accident black spots" and speed cameras might be more likely to be deployed at these locations.


But a high number of KSI collisions during the SSP does not necessarily mean that a site is an "accident black spot". The site could, due to random variations, just have had an "unusually" high number of KSI collisions.


Figure 8.1 shows what happened in practice. KSI collisions were occurring at their "normal" rate in the 5 years before the SSP and then these sites were selected after a high number of KSI collisions had occurred during the SSP. After the SSP ended, the KSI collision rate reduced back to "normal" and it then took a year before speed cameras were deployed.


These effects come about due random variations interacting with the site selection process and they are referred to as RTM.


Regression to the mean (RTM)


The sudden reduction in KSI collisions at the end of the SSP occurred without any speed cameras being deployed and without any other specific action being taken by the authorities. It may therefore be tempting to ask: "why were these sites relatively safe for the first 5 years, then suddenly become so dangerous from periods 11 to 15, and then suddenly return to being so much safer again after that?"


The answer is that they didn't. It's not the risk of a crash changing, it's simply an effect caused by the site-selection process.


The sudden changes in KSI collision rates at the start and end of the SSP do not reflect a sudden change in the probability of a KSI collision occurring at these sites. They are simply the effects of random variations in KSI collision rates, combined with a selection process that favoured selection of those sites where the rates were high during the SSP.


How the site-selection process can lead to RTM is demonstrated using random numbers here: regression to the mean.


So what effect did the mobile speed cameras actually have?


There were 2 periods of time prior to speed cameras that were free from the effects of RTM and when collisions were occurring at around their mean (or normal) rate: the 5 year period before the SSP and the 1 year period after the SSP. Had mobile speed cameras not been deployed, collisions should have continued at around the mid-point of these 2 rates.


If the mobile speed cameras had made these roads safer, then the mean KSI collision rate after they started operating should have reduced but, in practice, the opposite occurred. Mean KSI collision rates increased following the deployment of mobile speed cameras.


In order to determine the effects of speed cameras, it would be necessary to run scientific trials called RCTs (randomised controlled trials) but RCTs have never been run. In the absence of RCTs, the effects of speed cameras must be separated from those of RTM but the author cannot find any official report that has managed to do this.


This is believed to be the world's first speed camera report in which the effects of RTM have been fully excluded from the final results. It should therefore be the most accurate report on the effects of speed cameras to date.


This page is just a brief overview looking at the main issue of RTM. The actual report goes into greater detail including compensating for general influences and separate results for fatal, KSI and all severities for both collisions and casualties.