THE EFFECT OF SPEED CAMERAS ON FATALITIES ALL ACROSS GREAT BRITAIN
Most speed camera reports only consider the effect of cameras on collisions at the camera locations, but the effects of cameras may extend over much wider areas.
In February 2003, the governments initial pilot evaluation1 (prepared by PA Consulting and UCL) reported that speed cameras reduced the numbers of fatal and serious injuries over entire counties by 4% but, in October 2003, an ABD report2 found that speed cameras had damaged road safety all across Britain, calculating that this had cost 5,500 lives. Although very different conclusions were reached, PA Consulting, UCL and ABD all reported that speed cameras did affect road safety over very wide areas. In December 2003, a PACTS and SSI report3 confirmed ABDs analysis that fatality reductions did dramatically decline as the numbers of speed cameras increased but disputed claims by ABD, Autocar and "others" that it was the speed cameras that had caused this. PACTS and SSI claimed that 5 other reasons may have been to blame.
In December 2005, the governments final speed camera evaluation4 (also prepared by PA Consulting and UCL) contained
no analysis of the area wide effect of speed cameras.
This report investigates the above claims by PA Consulting, UCL, ABD, PACTS, SSI, Autocar and "others":
1) Did the trend in reductions of fatalities across Great Britain change as the number of speed cameras increased?
2) How many lives have been lost as a result of this change in trend?
3) Could the 5 reasons proposed by PACTS and SSI account for this number of lives lost?
4) Could speed cameras be responsible for causing this number of deaths on British roads?
1) Yes. There was a clear and definite change of trend as numbers of speed cameras increased.
2) Over 4,600 lives were lost by 2002 (the 1st 10 years of speed cameras), rising to over 12,000 by the end of 2009.
3) No. None of the 5 reasons that PACTS and SSI propose come anywhere near explaining why these lives were lost.
4) Possibly. There are many negative side effects of speed cameras which may extend over the entire road network.
The official DfT figures clearly show that the most important road safety indicator (fatalities per billion miles) had been reducing remarkably consistently ever since the start of the 1960s, with each decade having larger reductions than the previous decade, all the way up to the mid 1990s. Then something went seriously wrong. Reductions that had averaged over 6% pa for over 15 years suddenly collapsed to under 3% pa for the next 10 years. In other words, the 1st 10 years of speed cameras gave us the worst safety improvements on British roads since the 1950s, but were the cameras to blame?
Speed camera policy being the main cause of the 12,000 lives lost is supported by 4 reasons:
1) The timing. The change in trend started as soon as the speed camera programme started.
2) The divergence from trend. As the number of speed cameras increased, the divergence from trend increased.
3) The magnitude of the change. The speed camera programme is the largest, most expensive road safety policy ever implemented and influences driver behavour, road safety publicity and Police policy over the entire road network. It has the potential to make a huge difference to the number of fatalities on our roads.
4) The lack of any other reasonable explanation.
The 1st step in solving a problem, is recognising the problem exists. All speed camera information from all official sources (such as Department for Transport, councils, Police etc) should be prefaced with a warning that speed cameras may have serious negative side effects that could result in more fatal collisions over wide areas.
A scientific trial should be run where all automated* speed limit enforcement should be made illegal every other year for 6 years. The 3 years without any speed camera enforcement can then be compared to the 3 alternate years with speed cameras enforcing. Of course, Police will always be able to enforce speed limits at all times where they stop suspects at the time of their offences, but Police enforcement should be applied evenly throughout the trial.
This scientific trial should be independently organised and independently supervised.
The trial would require national publicity campaigns to ensure motorists are aware of whether or not they can be prosecuted by speed cameras so that the effects of cameras only exist when the cameras are enforcing.
Speed camera supporters and objectors should each be asked to produce their predictions for the number of fatalities that will occur on the nations roads in each following year so that claims of success or failure cannot be constructed after the event to suit political or financial purposes.
* "automated" refers to all speed limit enforcement where suspects are not stopped at the time of the offence
This report uses the official figures from the DfT.
This investigation is easy to do. All that is required is an understanding of cumulative percentages (mathematics taught in schools) and the official data. All the data for the main analysis is available on a single spreadsheet on the DfTs website5 and the spreadsheet for this analysis6 is available to download so everything can be verified.
Figure 1: Fatalities reached a peak in the mid 1960s, but then there was a clear downward trend through the 1970s and 1980s. The fall in fatalities at the start of the 1990s was substantial, larger than any previous fall. All of this was before speed cameras but from 1994, just as any effect of speed cameras might start to be seen, the fatalities leveled off, hardly falling at all for the next 9 years. Therefore PACTS, SSI, ABD, Autocar and "others" are correct, the rate of reduction of fatalities did decline after the introduction of speed cameras.
But traffic levels have been rising and this can be included in the analysis.
Figure 2: The F/bvm (fatalities per billion vehicle miles), possibly the most meaningful indicator of road safety, is calculated simply by dividing the numbers of fatalities each year (from figure 1) by the number of billion vehicle miles each year (also from figure 1).
F/bvm has been reducing remarkably consistently for many decades. There was an initial reduction in the 1950s followed by a fairly static period. Fatalities were rising, but roughly in proportion to the rise in traffic. As soon as the 1960s started there were over 3 decades of reductions, with each decade having a higher average annual reduction than the previous decade until 1994 when the trend suddenly changed. Although road safety was still improving, the rate of improvement markedly slowed down.
Figure 3: The change in F/bvm is easier to see by examining the period starting 15 years before 1993, extrapolating the trend after this point and comparing the trend to the F/bvm that occurred.
Note: The entire reduction of F/bvm in 1983 can be attributed to the seat belt law7 but, with consistent seat belt wearing rates from this point onwards, the falls in subsequent years were entirely due to other reasons (Figures 7+8).
Figure 4: The difference between the actual F/bvm after 1983 and the trend from figure 3 is used to calculate the number of fatalities that would have occurred had the trend not changed, as shown in the red trend line. Instead of continuing to fall as they had for the previous 15 years, fatalities remained almost constant for over a decade.
The downward trend eventually resumed in 2007, when the recession started, but the number of lives lost due to the change of trend (the addition of the numbers above the red trend line) was over 4,700 by the end of 2002, and over 12,000 by the end of 2009 (an average of 780 lives lost pa).
What happened locally where speed cameras were 1st used such as in Thames Valley?
Figure 5: Thames Valley, consisting of the 3 counties of Oxfordshire, Buckinghamshire and Berkshire, saw falls in fatalities for 5 consecutive years up to 1993. Then the trend completely reversed with fatalities rising as more and more Gatso speed cameras were installed, reaching a total of 359 cameras in 212 sites. After 1993, it was 10 years before fatalities dropped below the number when cameras started but the downward trend only eventually resumed in 2007, as it did nationally.
The number of lives lost in Thames Valley due to the change of trend (the addition of the numbers above the red trend line) was over 700 by the end of 2008 (an average of 49 lives lost pa).
Note: For the trend line, figure 5 assumes that traffic levels in Thames Valley changed in proportion to the national change.
Why did road safety collapse when speed cameras started?
PACTS and SSI claim3 that the change of trend wasn't caused by speed cameras, but may have been caused by these 5 reasons:
1) "continued increases in traffic"
2) "sharp increases in motorcycle casualties"
3) "a levelling-off of drink drive fatality numbers"
4) "a decline in seatbelt-wearing"
5) "increased use of mobile phones while driving"
Could these 5 reasons explain the loss of 12,000 lives?
1) Increases in traffic are included in the figures for F/bvm (fatalities per billion miles) therefore do not account for the change of trend.
2) Figure 6. Motorcycle fatalities fell to their lowest in 1993, levelled-off for 3 years and then rose to a peak in 2003. Clearly this is a part of the change of trend, but what caused this?
Motorcyclists may well be more sensitive to changes in road safety strategy than other road users and are the only "vulnerable road users" targeted by speed cameras. If speed cameras have damaged road safety, then this may influence motorcycle fatalities in particular.
3) Drink drive fatality numbers "levelling-off" is exactly the problem as stated right at the start for all fatalities on the roads therefore it is part of the change of trend, but it doesn't tell us why?
As the focus of road safety switched to vigorous enforcement of speed limits, or if speed cameras meant fewer traffic police, drunks may feel that they are less likely to get caught. It is possible that speed camera policy may have indirectly contributed to more drink driving.
4) Figure 7. Seat-belt wearing declined slightly from 1995 to 2002 costing an estimated 102 lives7, but the change of trend had already cost over 4,700 lives6 by this time. Furthermore, seat-belt wearing increased again after 2003 while road safety deteriorated even further. Therefore the slight "decline in seatbelt-wearing" was not a significant factor in the change of trend.
5) Mobile phones were not in common usage while driving in the late 1990s as the size and cost of the phones was prohibitive back then. Also, the danger of using a mobile phone while driving has been vastly overstated. In the 5 years 2005 to 2009, Police investigations show that "Driver using mobile phone" was a factor in just 0.8% of fatal collisions8. Considering that driving on the phone has become so widespread now, yet is still a factor in fewer than 1% of fatal collisions, mobile phones cannot possibly have had any noticable impact over 15 years ago when road safety changed so dramatically.
When did speed cameras start?
Apparently there were a few experimental speed cameras in 1992 but the actual speed camera programme started in 1993. For example, although Thames Valley claimed to have speed cameras in 1992, the 1st one actually installed was on 02/07/1993 (No603, in site TV4501). As the programme started later in the year, and because it took quite some time to start prosecuting large numbers of motorists, speed cameras could not have had any noticable impact on the figures for 1993. Any effect the speed cameras had would be seen in the figures for 1994 onwards (the 1st full year of operations).
Furthermore, the sooner that medical services can attend the scene of a serious collision, the more lives that can be saved. Mobile phones in vehicles may well have assisted in reducing the number deaths on our roads over the last decade by enabling the medical services to be contacted much more quickly after a collision (often immediately). Therefore the "increased use of mobile phones while driving" played no part in the loss of trend.
Conclusion, none of the 5 reasons that PACTS and SSI propose can explain the loss of trend.
Why did road safety improve all the way up to the start of speed cameras?
The main reasons may have been:
1) Improving vehicle design
2) Improving medical knowledge, equipment and speed of emergency response
3) Safer roads by engineering and design
4) Improving driver behaviour and attitude assisted by road safety information and traffic police in marked cars
The 1980s was the decade that had the best road safety improvements and makes up the bulk of the 15 years before the start of speed cameras (1978-1993). The entire reduction of F/bvm in 1983 can be attributed to the seat belt law of that year7 but, with consistent seat belt wearing rates from this point onwards (Figure 7), the reductions in subsequent years were entirely due to other reasons (Figures 2+3).
The 1980s was also the decade of the fastest cars ever available to the public. Started by the Golf GTi in 1977, the 1980s saw an explosion of hot hatches with many car manufacturers launching their GTi, XR, GTE etc. With these cars came very fast speeds along with the necessary improvements in tyres, brakes and suspension etc. Faster cars tend to be more enjoyable to drive and, with enjoyment, comes concentration. So there may be some positive benefits to fast cars (better active safety and driver concentration) along with some negative side-effects of longer stopping distances from faster speeds. The combination of all these influences contributed to the best safety improvements of any decade we have ever had, yet all without ABS, airbags or speed cameras etc.
What should have happened since the mid 1990s up to now?
Vehicle safety design has really taken off since the mid 1990s. Vehicles have much better suspension, tyres and brakes along with a host of new safety features such as ABS, ESC, seat-belt pre-tensioners, front airbags and later side air-bags and now dual inflating air-bags, far superior crumple zones, stiffer passenger compartments, side-impact protection, higher speed crash tests, side impact tests, etc. For example, just for ESC on it's own: "...studies now have estimated that ESC reduces fatal single-vehicle crashes by between 30-50% among cars and 50-70% among SUVs".
And medical knowledge and equipment has continued to improve as well. The speed of response after injury and the procedures to save lives have improved dramatically. Many areas have fast-response medics who are often deployed away from hospitals ready to respond faster than conventional ambulances deployed from a base. And there are now air ambulence and Police helicopters for even faster recovery to hospital. And the fire service now have better techniques to extract vehicle occupants without incurring further injury, including the jaws of life. The proliferation of mobile phones has also enabled emergency services to be alerted to accidents almost immediately, and GPS technology can pinpoint accident locations and guide emergency services straight to the scene.
The roads continue to be engineered to be safer, eg much of the road furniture (lamp posts, signs etc) now used are designed to crumple upon impact, rather than remain solid, so that injuries are less severe and deaths are prevented.
All this suggests that 1994 to 2003 inc (the first 10 years of speed cameras) should have given us the best road safety improvements we have ever had, yet we actually had the worst since the 1950s!
But was it the policy of using speed cameras that caused the change of trend?
Of the 4 main reasons that road safety improved prior to speed cameras, 3 of those continued and, if anything, accelerated after the mid 1990s through to now. The 4th reason, namely Improving driver behaviour and attitude assisted by road safety information and traffic police in marked cars, stands out as being the only main factor left that could explain the change. Although this is very difficult to quantify objectively, speed cameras are specifically intended to affect driver behaviour, the speed camera programme has changed the messages in road safety publicity (and public confidence in them) and may have substantially influenced the number and role of traffic Police officers.
It is clear that something went dramatically wrong in road safety from 1993 onwards yet, when I read the websites of the DfT, local councils, safety camera partnerships or Police, or speak with the officials involved, I don't find any recognition of this problem. They all seem to be "spinning" the statistics in order to claim great success.
The 1st step in solving a problem, is recognising the problem exists!
Note1: The two year pilot evaluation prepared by PA Consulting and UCL (University College London) is on the
p6 "In the six comparable pilot areas (the whole partnership areas not just at camera sites) the annual number of killed and serious injuries has fallen to 4% below the long-term trend. In this respect, the six areas have outperformed the rest of Great Britain. This means that across the six pilot areas there were about 530 fewer people killed or seriously injured. A little over half of this reduction occurred at camera sites".
PA Consulting and UCL thought the area wide effect would be so large that it could be measured from hospital statistics:
p30 "All areas joining the system should endeavour to measure the impact the system is actually having on the NHS, for example by monitoring the number of hospital bed-days required for people injured in road traffic accidents".
Note3: PACTS (Parliamentary Advisory Council for Transport Safety) report on PACTS website and on
Note4: The governments final speed camera evaluation is generally refered to as the "4 year evaluation". Download from DfT website or download here 4_year_evaluation.pdf. This report is reviewed here, government reports.
Note6: Spreadsheet for the main trend analysis, GB_road_safety.xls.
The decline in seat-belt wearing after the mid 1990s is investigated by examining 1983 (when wearing front seat-belts became manditory) and calculating the percentage of vehicle occupant lives saved for every 1% increase in the number of drivers that wore seat-belts. This is then applied to the "decline in seatbelt-wearing" that occured from 1994 to 2002 (when seat-belt wearing dipped slightly, figure 7) to estimate the resulting number of fatalities.
From the governments survey data, driver seat-belt wearing rates rose from an average of 38% in 1982 to 93.9% average in 1983 to 1985 due to the new seat-belt law. The law therefore resulted in a 55.9% rise in seat belt wearing. In the 3 years before 1983, F/bvm of vehicle occupants was relatively constant at an average of 15.0 and from 1983 to 1985 it was again relatively constant but at the lower average of 12.3 (figure 8). This drop led to a reduction of 500 fatalities pa average and, as the total reduction in fatalities was around 500 pa, the entire drop in 1983 can therefore be attributed directly to the seat-belt law.
This works out to 9 lives pa saved for every 1% that wear their seat belts, or 900 lives pa saved if 100% wear seat belts, or 500 lives pa saved due to the increase in seat belt wearing as a result of the law.
Note that, with consistent seat belt wearing rates after 1983 (figure 7), the falls in subsequent years were entirely due to other reasons (Figures 2+3).
Seat-belt wearing fell from 92.5% (1992-1994 average) to a low of 90% in 1997 so the "decline in seatbelt-wearing" cost around 100 lives over the 8 years 1995-2002, but the change of trend had already cost over 4,700 lives during this period. Furthermore, seat-belt wearing increased again after 2002 whilst road safety deteriorated even further.
All the data for national fatalities and billion vehicle miles is available on the DfT website and it is not difficult to do this for yourself. The only correction factor that had to be used was due to this statement under the DfT data:
"From 1993 the data has been estimated using the expansion factors and the new methodology for measuring road lengths, they are not directly comparable with the figures for 1992 and earlier."
This refers to the "All traffic (billion vehicle miles)" figures. It can be seen that the change from 1992 to 1993 is actually incredibly small and seems hardly worth the DfT mentioning. Even so I have therefore calculated and used a correction factor of 1.4% (gb_road_safety.xls) for all traffic miles figures from 1950 to 1992. In practice this makes no discernable difference to any of the figures or calculations. I have emailed the DfT several times with FOI requests asking what correction factor they calculate or recommend but, after many months (edit: now years), they have still not replied.