Fatality rate trends.
We say speed cameras and the policies that support them are now costing over 1,000 lives every year.

Here are our calculations and conclusions.


We've been looking at the fatality "rate" on UK roads. The "rate" is the number killed per billion vehicle kilometres.

In each graph the red line is actual official data, plotted completely without adjustment.

Basic data from the official publication "Transport Statistics Great Britain". (click here) See tables 9.10 and 9.7

The yellow and blue lines are calculated trend lines. See text. 

This graph with its log scale is the clearest overview. There's long been a fairly regular year on year reduction in the road fatality rate. This change has a constant annual factor. (i.e. it changes by approximately the same percentage every year.) This sort of curve is correctly termed "exponential".

Broadly we have long been used to seeing the fatality rate reduce at about 5% every single year. The log scale is useful because it shows such changes as a straight line. This helps to clearly reveal changes in the fatality rate change.

Modern fatality rate changes have not been entirely constant. We think it's clearest to divide the line into three different sections. Firstly there's the period 1950 to 1978.  Then the period 1978 to 1993, and finally the period 1993 to date.

Average slopes of different sections are:

Year band Average annual factor or
1950 to 1978 95.58%  4.42%
1978 to 1993 93.19% (yellow trend line) 6.81%
1993 to 2001 97.05% 2.95%
1950 to 1993 94.74% (blue trend line) 5.26%
1999 to 2001 99.58% (i.e. no significant trend) 0.42%

It's quite clear that there was a loss of the previous trend that occurred in about 1993. More worrying still is that since 1993 things have continued to worsen and the latest years are the poorest in modern times. We expect an increase in the fatality rate to be announced for the year 2002.

This is the same as graph 1, but with a conventional scale. It's included for those folk who may not feel confident with the clearer log scale graphs.
This graph is the same as graph 1, but we've "zoomed in" on a more recent period. It's clear to see the red actual data line diverging from the trend lines.
This graph is the same as graph 2, but we've "zoomed in" on a more recent period.
Making choices

You can't do work like these graphs without making choices. We've based our choices on reasoning which is explained in this section.

Why do the curves touch at 1993/4?

Look at figure 3. It's plain to see that the actual line (red) diverges from the recent (15 or 16 year) trend (yellow) sharply in 1993/4. It's a pivotal point where the old trend stops and a different new trend starts. We are specifically interested in comparing the old trend and the new. We have therefore correctly and deliberately ensured that the trend lines pass through this point. You can see this same sudden change in views of other data too. See graphs 1.2, 2.5, 2.6 and 2.7 (here

Which is the true slope?

The yellow curve is the best match for the period 1978 to 1993. The blue curve is the best match for the overall period 1950 to 1993. We do not know of any reason why the modern trend (yellow) could not have continued from 1993 to date.

Why show log scale graphs?

The curves of constant factor show as straight lines on log scale graphs and this aids the eye in spotting variations in trend. We've included exact equivalent standard scale graphs for comparison purposes. There's no attempt to hide information or mislead by using log scale graphs.

Drawing conclusions

Something happened in about 1993 to badly damage the long term downward trend in British road fatalities. The changes are very significant. In fact if we'd followed the yellow trend line 3,657 people who have died would still have been alive at the end of 2001. The extra lives lost in 2002 (figures not yet available) will be well over a thousand.

Compared to our trend lines the following unnecessary deaths have already occurred on UK roads between 1993 and the end of 2001: We've added some drink drive estimates for comparison purposes.

Table 1. Deaths estimates and comparison data

Blue deaths estimate
Yellow deaths estimate
Annual speed camera convictions (England and Wales)
Drink Drive fatalities for comparison
Think campaign spend on "speed kills"
480 est

Think Campaign info from (here) Drink drive info from (here)

new We were astonished to note such a close relationship in the pattern of growth of speed camera convictions and the calculated extra deaths. We didn't even need to scale the figures to produce this graph, except to divide convictions by 1,000. When the (blue and yellow) trend lines were being considered we'd not even looked at the specific numbers for convictions by speed camera. It was only after the convictions were added to table 1 above that the closeness of the trends was spotted. So there you have it. We have a provisional figure of one extra death for every 1,000 speed camera convictions.

What changes occurred on British roads starting in about 1993 and continuing to the present date?

And we only have a single answer to offer: The government concentrated on "speed kills" as its primary road safety policy and began to litter the country with speed cameras.  This policy sends entirely the wrong messages to all sections of the roads community including:

  • Pedestrians and cyclists became much more afraid of traffic and began to desert the roads; "Speed Kills"
  • The Police managers decided that "PC Gatso" could do their job on the roads, and they could divert resources to "more important" matters. (click here)
  • Drivers received the false message that if they stuck to the speed limit they would be safe.
  • The driver's responsibility for setting speed was reduced.
  • Local authorities and police forces have recently been told that they can fund their road safety programs with fines from speed cameras.
See further theories and evidence throughout this web site. Especially consider the information (here) which describes how a subtle effect applied across a wide population might result in accidents where there were none before.

You might wonder if the reduction in fatality rate is something that's going on everywhere; perhaps we've reached the buffers, perhaps "all the easy gains have already been had"? So we've looked at other countries to see what's happening there. Astonishingly, in every country which we have sufficient data for we find that fatals are not falling in countries with a "speed kills" road safety policy. Non speed kills countries continue to enjoy pervious levels of improvement. (click here)

If anyone has any other sensible or supportable theories about the cause of the change in trend we will be delighted to publish them here on this page.

You can download the complete Excel 97 spreadsheet with graphs (click here). If anyone has a better way to develop the trend lines, please let us know.

Factors affecting changes in fatality rate reduction

We'd like to look at all the major factors that might affect the slope of the fatality rate curve in normal economic conditions. It's a system with a lot of inertia and we should not expect to see big changes. We've created the list below and placed it in what we consider to be order of importance.

Official policy

Within "official policy" we include messages to drivers, changes in law, changes in Policing, changes to driving tests, changes in advice to drivers, instructions to local authorities etc. These changes may in turn lead to altered driver behaviour. The average behaviour isn't going to change very fast. With the fatal accident rate all set to rise, we think bad policy must be making the roads more dangerous to offset the improvements in road and vehicle engineering and medical care which we are still getting. 

Vehicle safety improvements

Vehicle safety improvements have always been ongoing and driven by technical change and market forces. If anything the emphasis on vehicle safety is accelerating, with motor manufacturers spending more on the development of safety features. Previous vehicle safety innovations have included: Hydraulic brakes, disc brakes, radial ply tyres, heated rear windows, crash test dummies, crumple zones and safety cells, collapsible steering column, seat belts, laminated windscreens, head restraints, side impact protection, ABS, air bags, traction control, emergency brake assist, cornering brake control etc. We couldn't stop these improvements even if we tried. We think they are very likely to be worth around 2% per annum reduction in the fatality rate. 

new Medical improvements

Improvements in medical care help to preserve life at accident sites, and it is reasonable to assume that each year the average accident victim is more likely to be saved. In the absence of alternative evidence we should expect these improvements to be ongoing at a steady rate. We think medical improvements should be worth about 1% per annum reduction in the fatality rate. 

Road engineering improvements

Better roads and signs, bypasses, anti skid surfaces, black spot treatments, etc.

Technological change

In this category we'd include changes that might affect vehicle safety and an example that comes to mind is mobile phones. Mobile phones are a serious safety concern, but are extremely unlikely to account for the loss of fatal accident trend. RoSPA have only been able to attribute 19 deaths to mobile phones over a 12 year period.

Social change

We don't know of any ongoing social changes that are likely to affect the fatality rate, although it's clearly conceivable that a dangerous fashion could emerge. We've seen no increase in young men showing off in cars for example. It clearly takes place, and may well be dangerous, but then it always took place and it was always dangerous.

new There has been an increasing preference for "4x4" vehicles over the period in question. Clearly these bigger heavier vehicles can on occasion do more damage when they crash. On the other hand they tend to provide better protection for their occupants. The trend towards 4x4 emerged in the 1980s, and the growth patterns appears not to co-incide with the loss of trend in the fatality rate. In any event, with occupants better protected, it seems likely that any overall changes would be small and certainly no big contributor to the 3,000 extra deaths we should be looking for to the end of 2001.


Comments on the above are welcome. If there is a demand we will create a comments page. We will be delighted to publish all suitable emails including those whose content we disagree with. Email comment.

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.
You can't measure safe driving in miles per hour.