A detailed examination of speed camera statistics

There is, of course, no controlled experiment available that would conclusively demonstrate the effectiveness of speed cameras. Such an experiment would require the same road, with the same road users, the same vehicles, the same weather conditions etc, at the same time, both with and without speed cameras. All we have are ‘before’ and ‘after’ figures for casualties.

First, let’s look at the casualty ups and downs on a road prior to the installation of speed cameras. Above are the statistics for Kidsgrove Bank near Stoke, in Staffordshire.

Had we placed speed cameras or garden gnomes along the road in 2003, we could claim that either had reduced casualties from 9 in 2002, to 5 in 2003, and to 4 in 2004 – an impressive 44 per cent reduction close to the starting point of 3 casualties in 1999. This is an example of the long established statistical phenomenon known as ‘Regression to the Mean’ (RTTM) or ‘Bias Selection.’ Clearly, any casualty reduction claim should include a calculation for RTTM. The DfT are aware of RTTM, which is explained in the Road Safety Good Practice Guide:

Regression to the mean

“5.119 This effect, sometimes called bias by selection, complicates evaluations at sites with high accident numbers - blackspot sites, in that these sites have often been chosen following a year with particularly high numbers occurring. In practice their accidents will tend to reduce in the next year even if no treatment is applied. Even if three-year accident totals are considered at the worst accident sites in an area, it is likely that the accident frequencies were at the high end of the naturally occurring random fluctuations, and in subsequent years these sites will experience lower numbers. This is known as regression-to-the-mean.

5.120 In practice it is believed that the regression-to-the-mean effect can over-state the effect of a treatment by 5 to 30 per cent, chiefly dependent on the length of accident period chosen.
Possibly the most straightforward way of allowing for both the regression-to-mean effect and changes in the environment would be to use control sites chosen in exactly the same way as the treated sites, and identified as having similar problems, but left untreated. In practice, as stated earlier, it is both difficult to find matched control sites and, if investigated, to justify not treating them.”

Not until the Four Year Evaluation Report for The National Safety Camera Programme was an attempt made to calculate the effect of RTM. It was published at the end of the report in Appendix H:

The data used to estimate RTTM is listed below:

The information obtained for these sites was as follows:

• Name
• Camera type
• Road class (A-road or other)
• Speed limit
• Date 3-year baseline period ends
• Date after period starts
• Duration of after period
• Baseline traffic flow (million vehicles per year)
• Length of section over which collisions are monitored
• Number of minor junctions within monitored section
• Number of PICs and FSCs in baseline period
• Number of PICs and FSCs in after period

“In the case of fixed cameras, only 4 partnerships were able to provide data and the majority of these cameras (65%) are in one area (Staffordshire).”

“Thus RTTM accounts for about three fifths of the observed reduction in FSCs (Fatal and Serious Collisions) with the effects of the cameras and trend each accounting for a fifth.”

In other words, a claim of a 40 per cent reduction in KSI (Killed or Seriously Injured casualties) by speed cameras is reduced to 8 per cent.

The report concludes:

“RTTM effects were estimated to account for an average fall relative to the observed baseline collisions of 7% in all PICs and of 35% in FSCs. RTTM effects represented one quarter of the observed fall in PICs and three fifths of the observed fall in FSCs.”

Other factors that could reduce the claims for the effectiveness of speed cameras:

Under-reporting and misclassification of serious injury accidents

Road Casualties Great Britain: 2006 - Annual Report


“Very few, if any, fatal accidents do not become known to the police. However, research conducted on behalf of the Department in the 1990s has shown that a significant proportion of non-fatal injury accidents are not reported to the police. In addition, some casualties reported to the police are not recorded and the severity of injury tends to be underestimated. The Department is undertaking further research to investigate whether the levels of reporting have changed.”

Changes in safety on England’s roads: analysis of hospital statistics (published 23 June 2006) BMJ:
http://www.bmj.com/cgi/rapidpdf/bmj.38883.593831.4Fv1.pdf

Objective: To compare trends in the numbers of people with serious traffic injuries according to police statistics and hospital episode statistics (HES).

Design: Descriptive study based on two independent population based data sources.

Setting: Police statistics and hospital episode statistics in England.

Main outcome: measures Rates of injury and death and their change over time reported in each data source, for 1996 to 2004.

Results: According to police statistics, rates of people killed or seriously injured on the roads fell consistently from 85.9 per 100 000 in 1996 to 59.4 per 100 000 in 2004. Over the same time, however, hospital admission rates for traffic injuries were almost unchanged at 90.0 in 1996 and 91.1 in 2004. Both datasets showed a significant reduction in rates of injury in children aged ≤ 15, but the reduction in hospital admission rates was substantially less than the reduction shown in the police statistics. The definition of serious injury in police statistics includes every hospital admission; in each year, none the less, the number of admissions exceeded the number of injuries reported in the police system.

Conclusions: "The overall fall seen in police statistics for non-fatal road traffic injuries probably represents a fall in completeness of reporting of these injuries."

Thus, the KSI figure is misleading as it takes advantage of the under-reporting of serious injuries and masks the less than impressive reductions in fatalities.

Length of Camera Sites and location of casualties relative to the locations of speed cameras, before and after camera installation.

According to DfT guidelines, a fixed camera site can be 400 metres to 1.5 kilometres in length, and a part-time mobile site can be 400 m to 5 Km in length. Also, casualties from adjacent roads to the road where the cameras are sited are sometimes used to justify cameras. The point of this is that cameras can only hope to be effective where they can be seen, and if justifying accidents actually occurred near to the cameras. Siting cameras away from where accidents actually took place and then claiming a reduction would constitute statistical fraud. Furthermore, drivers can’t be expected to slow down for a camera, say 1km distant. In short, the location of accidents before and after camera installation is an important factor in assessing the effectiveness of speed cameras.

Contributory factors in accidents from Police STATS 19 forms

Obviously, for speed cameras to be effective at least some of the justifying accidents must have ‘exceeding the speed limit’ as a contributory factor. The DfT looked at the contributory factors from 147,509 fatal and injury accidents for 2006. ‘Exceeding the speed’ limit contributed (as distinct from 'cause') to 5 per cent of all accidents, and 12 per cent of fatal accidents. The other, unrelated speed factor of ‘excessive speed for the road conditions, UNDER the posted limit’ is an education problem that can’t be tackled by speed cameras. This is a bigger problem than exceeding the speed limit, according the the DfT's categorisation of speed, being a contributory factor in 12 per cent of accidents and 17 per cent of fatal accidents. The DfT 'spin' the figures by adding the two entirely different factors 'exceeding the speed limit' and 'excessive speed for the road conditions' together in order to claim speed 'speed' is a factor in 17% of all accidents (18% for the 2007 casualty figures).

Of course, the STATS 19 forms filled in a the roadside by police officers attending road accident involving injury or death can be subjective, and therefore represent the opinion of officers rather than being the result of an extensive accident investigation. 'Excessive speed for the conditions' isn't an offence in law, unless driver behaviour can be shown to be 'without due care' or 'dangerous.' Speed is only described as excessive if it actually contributes to an accident. Despite the potential flaws in the box ticking system of the STATS 19 forms, these are the only official figures that we have. They don't show exceeding the speed limit as the largest 'contributory factor,' let alone 'cause.' The average number of contributory factors per accident is 2.4. Very rarely is there only one. How does driving above a number on a stick cause accidents, or driving at or below that number prevent accidents? Speed has to be appropriate for the prevailing road conditions, which is often below the speed limit. Many speed limits have been lowered to well below what would objectively be considered to be a sensible maximum in ideal conditions, which is surely the function of a speed limit. Appropriate speed limits achieve the greatest compliance and require the least or no enforcement.

From DfT Contributory factors in accidents 2007:

* Failed to look properly was the most frequently reported contributory factor and was reported in 35 per cent of all accidents. Four of the five most frequently reported contributory factors involved driver or rider error or reaction. For fatal accidents the most frequently reported contributory factor was loss of control, which was involved in 33 per cent of fatal accidents.

* Younger and older drivers are more likely to have a contributory factor recorded than drivers aged 25-69. Younger drivers, particularly males, are more likely to have factors related to speed and behaviour, whereas older drivers are more likely to have factors related to vision and judgement.

http://www.dft.gov.uk/pgr/statistics/datatablespublications/accidents/casualtiesgbar/roadcasualtiesgreatbritain20071


Safety Engineering Measures at Camera Sites

Safety engineering measures accompanying camera installation should be taken into account when assessing camera effectiveness.

Some potential negative effects of speed cameras on road safety

Nationally, in 1998 deaths from drink driving reached an all time low of 460. Between 1996 and 2003 traffic police officer numbers fell by about 30 per cent (written answer to a parliamentary question) , due to officers being replaced by speed cameras. There may well be a connection between increased drink-drive deaths due to less traffic police carrying out less breath tests. By 2004 the number of drink drive deaths had reached 580. In 2006 the figure was around 540. In 2007, the number was back down to 460, presumably contributed to by police patrols being more vigilant where drink-driving is concerned.

Single function speed cameras can’t detect the myriad of driving offences tackled by the traffic police officers that speed cameras replaced.

Other negative effects include the installation of cameras instead of other more appropriate measures, such as engineering or pelican crossings. Distraction from the driving task is another potential problem.

A substantial number of camera sites show either no reduction or an increase in casualties.

The national speed camera programme claims an average 2.2 mph speed reduction at camera sites, compared to Staffordshire’s claim of just a 1mph reduction.

Also, from the DfT Road safety Good Practice Guide:

Other factors to consider

Accident migration

5.124 The existence of accident migration is a fairly controversial issue but has been reported to be a real effect �Mountain et al, 1992; Boyle et al, 1984; Persaud, 1987. It is simply that an increase in accidents tends to be observed at sites adjoining a successfully treated site, giving an apparent transfer or migration of accidents. It is unclear precisely why this effect occurs but is hypothesised that drivers are compensating for the improved safety at treated sites by being less cautious elsewhere.

5.125 Obviously to detect such an occurrence, you need to compare the accident frequencies before and after implementation of a scheme and those for the surrounding area with a suitable control group.

5.126 However, research and practical evidence eg Brindle, 1986; Websterand Mackie, 1996 have demonstrated that local area traffic restraint schemes do not create a significant increase in accidents on surrounding roads. Mountain 1998. has more recently concluded that a more likely explanation for any observed increase is a reverse regression-to-mean effect arising due to bias in the selection of the neighbouring sites.


Factors that reduce death and injury

EuroNCAP claim a 12 per cent reduction in KSI for every safety star awarded for vehicle crash performance. Thus a 5 star vehicle could potentially reduce the chances of being killed or seriously injured by 60 per cent. Better paramedic care and improved treatment for head injuries make a substantial contribution to reducing KSI. Safety engineering an road improvements save lives.

A note about mobile cameras

Mobile Lastec cameras are operated on a variety of roads on a part-time basis, typically half a day. They are often operated between existing fixed cameras, and in transitional zones between speed limits in defiance of the ACPO guidelines for a 200 metre ‘buffer zone’ following a speed limit change, where drivers should not be prosecuted. When a speeding driver sees a camera van, it is probably already too late due to the range at which speed is measured is far greater than that of the more visually obvious fixed cameras. The fact that mobile cameras are only present on a part-time basis and are less obvious means that they are likely less effective than fixed cameras at slowing drivers down, and the claims for casualty reduction are even less likely to be real

Speed cameras INCREASE injury accidents: Safety Performance of Traffic Management at Major Motorway Road Works (Transport Research Laboratories Report TRL 595, 8/5/2005

The Department for Transport (DfT) funded, then suppressed, a study that shows a 55% increase in injury accidents when speed cameras are used on motorway work zones and a 31% increase when used on motorways without construction projects. According to the Transport Research Laboratory (TRL), the "non-works [personal injury accident] rate is significantly higher for the sites with speed cameras than the rate for sites without."

Analysis of this data, on 43 of the report, Effect on Personal Injury Accidents:

Enforcement Type, Construction Zone, No Construction

Conventional speed cameras, 55% increase, 31% increase
Speed-averaging cameras (SPECS), 4.5% increase, 6.7% increase
Police patrols, 27% reduction, 10% reduction

The TRL study compared accident reports covering 29 motorway construction zone projects over 730km of road from November 2001 to July 2003 with an equivalent period without the construction zones, controlling for changes in traffic volume. The taxpayer funded report was never made public and was revealed by a FOIA request. The report demonstrates that police patrols reduce injury accidents, whereas Gatso type cameras in particular increase injury accidents.

A pdf of the report can be found here:

http://www.thenewspaper.com/rlc/docs/04-trl595.pdf

Vehicle Activated Signs versus Speed Cameras

In 2003 the Transport Research Laboratory (TRL) carried out a study on Vehicle Activated Signs under contract to the Road Safety Division of the Department for Transport. Speed cameras were claimed to reduce accidents by 14%, but the study suggested that electronic speed limit reminders reduced accidents by 58% and electronic bend & junction warning signs reduced accidents by 26%. The TRL study was the largest of its kind ever conducted, examining the effects of four different electronic warning signs at over 60 locations, and measuring the impact on accident rates over a three year period.

The report concluded that electronic warning signs achieve substantial accident reductions, can operate at thresholds well below speed cameras, remain effective for years and have much lower operating costs than cameras.

TRL Report 548 entitled: 'Vehicle-activated signs - a large scale evaluation' can be downloaded here:

http://www.trl.co.uk/store/report_detail.asp?srid=2699&pid=211

TRL 548: "There is an established, positive relationship between vehicle speeds and road accidents. On rural roads, driving too fast for the conditions is more likely to be a factor in accidents than exceeding the speed limit. Encouraging drivers to drive at suitable speeds for the conditions is particularly important, since driver error is the major contributory factor in 95 per cent of accidents. A range of rural road safety engineering measures, in particular vehicle-activated signing, has been developed to encourage drivers to approach hazards such as bends and junctions at a safe speed, and to encourage them to comply with the speed limit, eg through villages. The signs display a message relating to road conditions such as these to just those drivers exceeding a set threshold speed. A study of the effectiveness of over 60 installations on rural roads in Norfolk, Kent, West Sussex and Wiltshire has been conducted by TRL for the Department for Transport (DfT). The trial aimed to assess the effect of the signs on speed and injury accidents, and drivers' understanding of the signs. The results will be used to develop best practice for sign installation. The signs appear to be very effective in reducing speeds, particularly those of the faster drivers who contribute disproportionately to the accident risk, without the need for enforcement such as safety cameras. In this study, a substantial accident reduction has been demonstrated."

Of course, many of the criticisms of speed camera statistics can also be levelled at the casualty reduction claims for vehicle activated signs (VAS). Nevertheless, there is no apparent reason for choosing speed cameras over VAS other than the fact that cameras can raise money and prosecute drivers, whereas VAS can't.

Overall Conclusions:

Prior to the introduction of speed cameras on UK roads in the mid-1990s, there had been decades of an impressive trend of falling fatalities, despite rising vehicle numbers and road use. If the ‘speed kills’ mantra and the focus on speed as the main cause of accidents were the correct focus, we would have hoped to continue the trend in the speed camera era. Sadly, the trend has levelled off despite speed cameras.

So, have speed cameras made Staffordshire’s roads safer above the benefits we would expect from better engineered cars and roads, plus improved medical care? Given that cameras are supposed to have been placed in accident black spots, it is not unreasonable to expect to see a significant reduction in the overall number of fatalities on Staffordshire’s roads. No reduction in the fatalities trend has taken place:

http://abdmercia.blogspot.com/2008/09/staffordshire-road-fatalities-1992-2006.html

As previously shown, figures for serious injury accidents have long been suspect and therefore fatalities are the only reliable statistics that we have. Furthermore, speed cameras can only fine traceable drivers, the potentially most dangerous, illegal drivers escape detection and prosecution. Every £40,000 spent on a speed cameras would fund an extra police officer. There has been a decline of 20% in the number of traffic police in England and Wales over the last decade. There are now 1,507 fewer patrolling the roads. Bad news for road safety, good news for dangerous and illegal drivers. My message is clear and well illustrated in this article - the claimed casualty reductions by speed cameras are a 'benefit illusion' derived from creative and demonstrably flawed statistics. The removal of all speed cameras (except at sites where it can be conclusively demonstrated that there is definite link between exceeding a sensible speed limit and accident causation, and there is no other or engineering measure that would rectify the situation) would be a positive move for road safety. Vehicle Activated Signs (TRL Report 548) are a cheaper alternative to speed cameras, although they don't have the potential to raise revenue.

The government finally ended the funding method that encouraged speed camera proliferation in April 2007 - camera partnerships are no longer allowed to keep the money from camera fines, which now goes direct to the treasury. Instead, a 'road safety grant' is given to each council in order to encourage appropriate road safety measures, thus re-affirming the often ignored guideline that cameras are only to be used as a 'last resort.' Councils now lumbered with the huge expense of maintaining fixed speed cameras are likely to consider removing cameras in favour of more cost effective road safety measures. Swindon are the first council to scrap fixed speed cameras, other councils may well follow suit, potentially heralding a new era in road safety where the remedy actually fits the problem. If inspiration is required, we can look to Britain's most improved road for safety, the A453, where a claimed 88% reduction in KSI has been achieved without speed cameras. Of course, had the measures taken on the A453 been accompanied by the installation of fixed speed cameras, then they would have unjustifiably taken much of the credit.

Driving is about risk management. The less hazards there are along a length of road, the faster vehicles tend to travel. This was borne out by TRL 511, where, to their horror, the authors found the rural roads with the fastest speeds were the safest - the opposite of what they were looking for - so they had to use creative statistics to reverse the findings, breaking a TRL 'rule' in the process i.e. they studied roads with very different traffic flows together, which is against the advice of Walmsley and Summersgill, 1998.

Read more about the report TRL 511, which is a fatally flawed attempt to claim that 'a 1mph reduction in average speed reduces casualties by about 5%,' here:

http://www.abd.org.uk/trl511.htm

So, as risk managers, drivers drive more slowly where prevailing road conditions increase the potential number of hazards. Road safety has to work on the basis that drivers don't want to kill or injure themselves or other road users, or be banned from driving, or fined, or go to prison. Nor should drivers expect to be fined by speed cameras for driving safely.