by Idris Francis
When I realised that the official long-term accident and casualty data includes the grid references of every reported injury accident, it became apparent that while separating out all the many causal factors is and will remain impossible, separating out supposed camera effect from the all the others is not just possible but a matter of simple arithmetic.
In effect, double-blind trials, rejected as impractical for a variety of valid reasons have nevertheless taken place by default over many years because the vast majority of injury accident sites that met the KSI threshold never had cameras installed.
I obtained all the data on 4.2m accidents and related casualties from 1991 to 2007 and extracted from it the only ones I needed - Year, Police Area, Location (to within 1km square), K, SI and Slight injuries. I entered all of these parameters into a series of new identical databases, each record relating to a different 1km square area (equivalent to the supposed area of camera benefit) in Britain reported to have suffered any injury accident at all in those 17 years. Each record, having been established on the basis of the first injury accident there, then had details of all K, SI and Slight data added for each subsequent injury in that area.
My computer then searched these new databases and identified every example of site and 3 year period which met the usual 4 KSI in 3 years threshold for camera installation. Each example was entered into a further new database, with year, police area "before" and ‘after’ totals for K, SI and SI.
There turned out to be 131,303 such examples, far, far more than have ever had cameras installed — not only because cameras have never exceeded 6,000 in number but because numbers were few before 2000 and then grew to around 6,000 around 20007.
Average ‘before’ and ‘after’ figures and % changes were then calculated and entered into an Excel spread sheet. Figures are shown for each police area as well as the overall national figures. In addition national totals excluding London (a special case due to its size and traffic congestion) are shown.
Similar Excel pages were generated for 6, 8 and 10 KSI in 3 year thresholds, and again for the first and then second halves of the overall periods, and then the same again for accidents rather than casualties - making 24 Excel pages in all. The single most significant is the 1992 to 2004 (‘before’ periods) for 4 KSI in 3 years.
The patterns are consistent throughout, showing that the results are not unduly dependent on choice of era or KSI threshold. Essentially the figures show that — especially for the more serious injuries K and SI and hence KSI — falls in numbers in the "after" period (in the vast majority of cases without cameras involved) are similar to those routinely, cynically and incompetently claimed as speed camera benefit. The national average figures, excluding London, covering 115,773 examples, are falls of 44% (K), 35% (SI), 36% (KSI), 4% (SLT) and 9% (All).
In general, the % falls are highest — up to 81% — in more rural and remote areas, and lowest in metropolitan areas with high traffic density and slower speeds. This makes sense, as the more remote the area and the lower the traffic volume the lower the chance of another accident in the same place. Equally, though I have not done the figures, safer roads such as dual carriageways and motorways must show higher % falls.
This is the best evidence yet to confirm that claims for camera benefit conjure that illusion largely from accident and casualty reductions which happen anyway due to RTTM and other factors.
Idris Francis 13 August 2010