Wide area regression to the mean in the 8 speed camera trial counties
Overheard at a mythical scamera partnership HQ:

"There was a multi pile-up on the M99 three years ago; 50 people were killed or seriously injured. We put up a speed camera at the site and there hasn't been a multi pile-up there since. 

Aren't speed cameras wonderful? That camera alone has saved 50 people from death and serious injury every year!"

Important Notice

We have been advised that "changes in reporting practice" in Thames Valley region in 1999 are responsible for the large part of the unusually high serious and KSI figures for 1999. It is notable that serious injury figures for Thames Valley from 1997 run as follows:

959,  906,  1,654,  1,472,  1,462,  1,379
The leap to 1,654 is in 1999 and represents about 700 KSI. We will investigate.

In the meantime the information on this page (although correct and from official figures) may be misleading where KSI figures are used. The fatality figures and graph are unaffected.

Read with this anomaly in mind.


We have been examining the figures for the 8 counties in the speed camera hypothecation trial compared with the rest of the UK. It's fairly well known that the 8 had bad results in the year of selection, which of course would have tended to produce a benefit illusion in the first year of the trial. But we've found a large effect - going back to 1997 as a start point, the 8 had worse results than the rest of the UK and very bad results indeed in 1999.

It's not at all surprising that reasonable results were returned in 2000 based only on national trends and regression to the mean.

Regression to the mean happens when something is selected for treatment because it's doing untypically badly.

Figure 1 shows that fatal accidents in the trial 8 counties have been worse than UK trends based on 1997 as a sample year. Selecting the counties for treatment at a high point is a classic way to ensure a regression to the mean benefit illusion. Even with speed cameras the 8 counties have not managed to return to the 95% of 1997 levels that the rest of the UK has achieved.

Figure 2 show what a complete disaster speed camera policy is. The UK (excluding the 8) reduced "killed or seriously injured" accidents to just 83% of 1997 levels by 2002. The eight only managed to reduce "KSI" to 92% of 1997 levels. This substantial difference in performance clearly shows that the hypothecation scheme delivered no road safety benefit whatsoever in the 8 trial counties.

In Figure 3 we have calculated the number of "KSI" in the 8 counties who would not have been "KSI" if the 8 counties in the scheme had followed the trends in the rest of the UK. When the 8 counties were chosen for the scheme, they were already 1,092 KSI worse than the general UK trend. Without speed cameras and without speed camera hypothecation we might have expected the 8 to return to the general UK trend.  But no such return took place. The fall in 2000 (the first year of the scheme) was less than the rise in 1999! Even 2000 and 2001 together didn't equal the rise in 1999. And in 2002 the figures started to rise again relative to the UK trend.

Safe Speed concludes:

Obviously speed cameras don't do us any good. Even installed in a basket of counties with a recent upwards blip in accidents the cameras have be unable to show an overall benefit.

Of course, you can download the spreadsheet and check our calculations (click here) And you can download the 6 official "Main Results" reports containing our source data. (click here)

The biggest error of all in speed camera claims is "regression to the mean". What we are examining on this page is "wide area" regression to the mean. False claims of camera benefits often depend on site level regression to the mean benefit illusions. You can read more about regression to the mean (here). And you can read about other tricky techniques for making cameras sound as if they are successful (here)

Let's make speed cameras as unacceptable as drink driving

We have a strict editorial policy regarding factual content. If any fact anywhere on this web site can be shown to be incorrect we promise to remove it or correct it as soon as possible.
Copyright © SafeSpeed 2003
Created 4/09/2003. Last update 10/09/2003