Response to Department for Transport Consultation on Road Safety Compliance
Driver retraining and re-assessment
Annex A: Suffolk Coroner's verdict concerning a fatal accident in November 1996
Annex B: A statistician's comments on TRL Project Report PR58 "Speed, Speed Limits and Accidents" (1994)
Annex C: A statistician's brief comments on TRL Report TRL511 "The relationship between speed and accidents on rural single-carriageway roads" (2002)
Annex D: Letter from Jim Fitzpatrick MP to the ABD concerning Road Accidents: Prevent or punish?
The Association of British Drivers (ABD) welcomes the opportunity to comment on the proposals in the road safety compliance consultation. It is concerned, however, that the Government is still unwilling to accept that its road safety policies could possibly be flawed. The Government continues to claim that serious injuries have fallen significantly, despite concerns expressed by the Transport Committee and others about the reliability of official police statistics. The Government's insistence that casualty reduction targets are being met, despite strong evidence to the contrary, is a refusal to acknowledge its fallibility.
While there are some positive proposals in the consultation, there are others that appear to be motivated by a desire to issue as many penalties as easily as possible, with little regard for justice or the true safety implications of the offences concerned. This amounts to a further extension of the 'punishment culture' that has been promoted by the Government in all aspects of life, and is to be deplored.
On speed, the ABD cannot support the Government's proposals to tackle 'extreme speeders' when so many speed limits have been reduced to unreasonably low levels. The Government clearly does not understand the proper use of speed limits, or that those set too low create danger by increasing non-compliance, the spread of speeds and driver frustration. Speed limit setting must return to the 85th percentile principle and most local speed limits on rural roads should be removed.
Even properly set speed limits are only a guide to what may be a safe speed under a particular set of conditions, and drivers must retain responsibility for adjusting their speed as those conditions change. The emphasis on speed limit enforcement is sending the wrong message that safety is about religiously observing speed limits, so drivers no longer have to think for themselves.
Camera enforcement of speed limits is mainly penalising responsible drivers at locations where exceeding the limit is frequently safe, while being ineffective against those who have not registered their cars with the DVLA. The latter are also more likely to drive while uninsured or unlicensed and should be the main targets of enforcement. Vehicle-activated signs are a much cheaper and effective alternative to cameras for alerting drivers to hazards ahead, and do not penalise them unnecessarily.
Driving at an inappropriate speed for the conditions is a symptom of an underlying failure to recognise a hazard ahead or respond to it by slowing down. This can only be tackled by education, and retraining courses should be offered instead of penalties to those who demonstrate a lack of skill in this area. Such drivers can only be detected by trained police officers, so there needs to be a much greater presence of traffic patrols. It is vital that the police are able to exercise discretion in this area, according to the level of risk involved, and they must not be constrained by targets.
Since the Government now recognises the low risk associated with speeding on motorways and dual carriageways, it is high time the motorway speed limit was raised, to at least 80mph. A higher limit would also be appropriate on those dual carriageways built to modern standards and with full grade separation.
On drink driving, the ABD supports the advice that drivers should not drink at all if they are planning to drive. Nevertheless, it would not support a reduction in the legal blood-alcohol level because of the likely increase in morning-after convictions, when drivers might be over the limit but show little sign of impairment.
The ABD would not object in principle to withdrawal of the right to a blood or urine test, provided there were adequate safeguards against wrongful convictions on the basis of a breath test alone. This would probably mean a requirement for two independent breath tests to be conducted, at least 20 minutes apart.
The ABD recognises the difficulty of tackling drug driving, when there are so many substances involved and it may be hard to establish levels at which a driver becomes unfit to drive. Nevertheless, fair enforcement has to be based on impairment and the ABD would not support the creation of an absolute offence of driving with any trace of an illegal drug in the body, regardless of whether it affected a driver's ability. It is right to discourage the use of illegal drugs, but driving penalties should not be used as a punishment for such use if no impairment can be established.
The ABD supports the actions proposed to improve drug screening and impairment testing, and to streamline procedures for taking samples from drivers. These measures represent a better and fairer way of tackling the problem of drug driving than creating a new, absolute offence.
On careless driving, the proposal to make this a fixed-penalty offence is one of the most worrying in the consultation paper. Careless driving covers a wide range of behaviours, from one-off errors to prolonged driving of a poor standard. It is largely subjective, therefore, and not suitable for a fixed-penalty approach, which can only apply to absolute offences.
All drivers make mistakes occasionally, so the threat of legal sanctions cannot make them into perfect human beings. Where a case of careless driving is sufficiently serious to warrant legal action, it should always be heard in court so that the correct penalty or remedial training can be prescribed. The Government's desire to apply fixed penalties to careless driving shows it is more interested in punishing drivers than preventing accidents. It would simply result in large numbers of drivers being penalised with no reduction in casualties, in the same way that the ten-fold increase in speeding fines has had no measurable effect.
With the current target-driven culture, some police officers might routinely issue a fixed penalty for careless driving after an accident, regardless of cause, knowing that most drivers would not be able to afford to challenge the allegation in court. This would cause serious further damage to police-public relations. More police traffic patrols, not driven by targets, would be able to detect more careless driving behaviour before accidents occur. Drivers who make minor errors are more likely to respond positively to a word of advice than a fixed penalty.
On driver retraining, the ABD agrees that it is better to offer remedial training to drivers penalised under the New Drivers Act than disqualify them, with the risk that they will continue to drive unlicensed. The high accident involvement of new, young drivers is related to the belief that learning to drive only begins after passing the test. This belief can be overcome with the right training in mental skills, managing risk, and how to learn from experience.
The ABD is convinced that the right kind of driver training has enormous potential to reduce road casualties. Safe driving is about evaluating many different factors, the risks involved and how to respond to them. This complex process cannot be replaced by insisting on blind obedience to simplistic rules. Internalising a safety culture within a driver's mind is much more effective than trying to impose it externally.
The ABD was formed in 1992 to campaign for a better deal for Britain's motorists. In particular, its founder members were very concerned that the increasing use of technology to enforce driving laws would undermine the traditional 'Three Es' approach to road safety (Education, Engineering and Enforcement) that gave Britain the safest roads in the world. Those fears have been more than realised.
The ABD is a voluntary organisation funded by subscriptions and donations from its members and supporters. It receives no funds from public bodies or private-sector businesses, so is truly independent. The ABD is a member of the Parliamentary Advisory Council for Transport Safety and the National Council of Voluntary Organisations.
The ABD's national committee of twenty members has agreed the response to this consultation. The response is in line with the policies outlined on the ABD's website. Individuals who join the ABD are assumed to support its policies in general, so it is not necessary (nor would it be practicable) to seek the views of the entire membership in preparing responses to consultations of this type.
The remainder of this document will address the questions listed in Annex B to the consultation paper. Before dealing with those questions, there is a claim made within the consultation paper that cannot be allowed to go uncorrected. In paragraph 2 of the Executive Summary and paragraph 1.3 of the Introduction, it is stated that the Government is on course to meet its target reduction in killed and seriously injured casualties by 2010. This statement is based on the claim that, by 2007, serious injuries had fallen by 36 per cent compared with the 1994-98 baseline.
As the Department for Transport is aware, the House of Commons Transport Committee, in its report entitled Ending the Scandal of Complacency: Road Safety Beyond 2010, was very concerned about the reliability of official figures for serious injuries derived from STATS19 forms. There is substantial evidence that not only do those official figures underestimate the numbers of serious injuries, but that the problem has been worsening over time. Hospital data suggests there has been little or no reduction in serious injuries compared with the baseline, so the claim that the Government is on course to meet its target reduction is not justified.
The Transport Committee's report was published in October 2008, so its findings were known to the Government before the consultation paper was published in November. It is disingenuous, therefore, for the Government to continue making a claim that is demonstrably false. Its attempt in January 2009 to describe hospital data as inconsistent and unreliable is a risible effort to justify its reliance on flawed figures to support its casualty reduction claims.
In its evidence to the Transport Committee, the ABD explained why casualties have reduced more slowly since the mid 1990s due to the adverse effects of the road safety policies pursued by the Government. The continuing refusal to acknowledge that casualties are not reducing as claimed is clearly an attempt to avoid an embarrassing admission that the Government's policies have failed. This is unacceptable and the claim that casualty reduction targets are being met must be withdrawn.
In answering the questions raised in the consultation paper, the ABD will explain why the Government's road safety policies are not working.
- Q.1 Do you agree that extreme speeders should receive a 6-point fixed penalty?
- 1.1 Before answering this question (and questions 2 to 5) it is necessary to explain how speed limits should be used and, just as importantly, how they should not be used. The ABD is strongly of the view that the Government's emphasis on speed and, in particular, its obsession with speed limit enforcement, is having a negative impact on road safety.
- 1.2 A speed limit should confirm a reasonably competent driver's assessment of the safe speed for the road on which he or she is travelling - it should not conflict with that assessment. Research since the 1940s has shown that the best way of setting speed limits, to obtain maximum compliance and a minimum spread of speeds, is by adopting the 85th percentile principle. That is, a speed limit should be set as close as possible to the speed that 85 per cent of drivers would not exceed anyway in the absence of a speed limit.
- 1.3 The 85th percentile principle was used as the basis of guidance on speed limit setting in the UK for many years, but was abandoned by the Government in its most recent guidance issued in 2006. This followed the growing practice of many local authorities to ignore the previous guidance and reduce speed limits below the 85th percentile. The result is that speed limits in many areas are now below a level that is seen as reasonable by most drivers, resulting in a high level of non-compliance. The few drivers who do observe these low speed limits cause bunching of traffic and frustration in following drivers, leading to potentially dangerous overtaking.
- 1.4 The detrimental effects of unreasonably low speed limits are not confined to the roads on which they are set, but bring speed limits into disrepute elsewhere as well. The knock-on effects can thus detract from safe driving practices over a wide area, leading to an overall increase rather than a decrease in accidents. This is exemplified in the case of Suffolk, which introduced 450 new 30mph speed limits at the end of 1995, many of them on roads that were previously subject to the national speed limit. The graph shows the way in which a steady downward trend in casualties to 1995 was dramatically reversed.
Suffolk casualties, all severities combined, 1989-2001 (DfT figures)
- 1.5 The detrimental effect on driver behaviour of unreasonable speed limits was set out in a verdict of the Suffolk Coroner on a fatal accident that occurred in November 1996. That verdict is reproduced in Annex A.
- 1.6 Another trend in recent years has been to apply local speed limits of 40mph or 50mph to significant lengths of rural roads, where drivers would expect the national speed limit to apply. Many of these speed limits have been introduced in the belief that lower speed limits lead to lower average speeds, which in turn lead to fewer accidents, regardless of whether they are speed-related. This belief has been fostered by reports commissioned by the Government from the Transport Research Laboratory (TRL) that purport to show a relationship between average speeds and accidents rates.
- 1.7 Two of these reports, PR58 ("Speed, speed limits and accidents") and TRL511 ("The relationship between speed and accidents on rural single-carriageway roads") have been analysed by a highly qualified statistician and his findings are set out in Annexes B and C respectively. It will be seen that he raises serious issues with the analyses and presentation of the findings of these reports. In particular, the choice of data analysis method in TRL511 may have allowed the researchers too much freedom to group the data in ways that enabled them to produce the desired conclusions.
- 1.8 The ABD has little doubt that successive governments commissioned these reports to provide retrospective evidence in support of policy decisions already taken. PR58 was published in 1994, so it would have been commissioned shortly after the first speed cameras were installed. TRL511 was published in 2002, shortly after the cost-recovery system, whereby camera partnerships were allowed to recover enforcement costs from speeding fines, was rolled out nationally.
- 1.9 The use of the flawed relationships in these reports between average speeds and accident rates could, in theory, be used to justify driving speed limits down indefinitely. As this process continues, more safe drivers will be criminalized as non-compliance increases, and casualty reduction rates will continue at levels below those that could otherwise be achieved. There needs to be a return to the 85th percentile principle as the basis of speed limit setting, and the removal of most local speed limits outside built-up areas.
- 1.10 Even where speed limits are set in accordance with scientific principles, they can never give more than an approximate guide to the maximum safe speed. Apart from purpose-built modern motorways and dual carriageways, most roads vary continually in width, alignment, forward visibility, frequency of junctions and accesses, and the presence of cyclists, pedestrians and horse riders. The maximum safe speed is changing continuously, and is also affected by weather conditions and time of day or night. At times it may be dangerous to drive up to the posted speed limit, at other times it may be safe to exceed it.
- 1.11 An essential driving skill, therefore, is to recognise and respond to changing conditions and vary speed accordingly. The growing emphasis on speed limit enforcement is eroding and denigrating that skill by suggesting that safe driving is about sticking rigidly to speed limits. This is leading to a worsening of driving standards and is having an adverse effect on road safety.
- 1.12 Paragraph 1.7 of the consultation paper states that there were 727 deaths in 2007 in which 'speeding' was recorded as a contributory factor. But, as acknowledged in paragraph 2.3, more than half of these were attributed to driving too fast for the conditions. It must be obvious that enforcement of speed limits cannot affect accidents taking place within the speed limit, and even in the remaining cases there will generally be more than one contributory factor, so eliminating all speed above the speed limit would not prevent all those accidents.
- 1.13 Inappropriate speed (regardless of the speed limit) is certainly a factor in a significant number of accidents. Apart from a tiny minority of deliberately reckless drivers, however, the inappropriate speed is itself a symptom of an underlying failure either to recognise a potentially hazardous situation ahead or to respond to it correctly by reducing speed. Greater enforcement of speed limits will not tackle this fundamental problem, which can only be addressed by education - and by promoting a culture of drivers accepting responsibility for their choice of speed. Current policies are having the opposite effect.
- 1.14 The consultation paper repeats many of the spurious claims used to justify greater speed limit enforcement. The claimed reductions in fatal and serious injuries at speed camera sites make no mention of the fact that regression to the mean accounts for a significant part of those reductions (where a camera is installed after a random upward blip in accidents, which then drop again anyway). As noted earlier, official figures for serious injuries (which outnumber fatalities by 10 to 1) are highly suspect. The claims for accident reductions on roads monitored by average-speed cameras take no account of engineering improvements that may have been introduced at the same time, or of traffic diversion, as some drivers change route to avoid the stressful experience of having to watch their speed constantly for miles at a time.
- 1.15 Many speed camera 'sites' are several kilometres in length and have varying characteristics. Concentrations of accidents are invariably at locations with the greatest hazard density, where speeds are lowest. Cameras are frequently installed, however, not where accidents have occurred but at the locations with lower hazard density, where speeds are above the speed limit as a result. Indeed, the Department for Transport's guidance to camera partnerships recommends that cameras should be installed where the 85th percentile speed is significantly above the speed limit - indicating that the speed limit is too low at that point. The camera is not being used to prevent accidents, therefore, but to catch as many drivers as possible, most of whom will be driving safely.
- 1.16 A major objection to the use of speed cameras that penalise drivers after the event is that, if a driver was not aware of being caught at the time, he or she may have little recollection of the circumstances when a notice of intended prosecution arrives in the post two weeks later. This makes it difficult to assess whether the allegation of speed is well founded, or possibly who was driving at the time. The judicial process is thus stacked against the accused.
- 1.17 Another objection is that speed cameras are ineffective in the case of vehicles for which the DVLA does not have up-to-date keeper details. Drivers who do not notify the DVLA of their acquisition of a vehicle are also more likely to drive uninsured, without an MoT certificate, or even without a licence. They are thus likely to be less safe than the majority of drivers and should be the main targets of enforcement, but instead are the most likely to avoid detection.
- 1.18 The whole emphasis of current Government policies relating to speed is aimed at being seen to 'do something' about reducing casualties at the lowest possible cost. Speed limits have been elevated to a level of importance they do not deserve simply because speed is easy to measure, enabling drivers to be prosecuted in large numbers. The dramatic slowdown in the rate at which fatalities have fallen nationally since the mid 1990s shows these policies are counter productive. Instead of admitting this, dubious statistics and flawed studies are repeatedly used to deny reality.
- 1.19 The ABD believes, therefore, that a fundamental change is required in Government policies towards speed limits and their enforcement. This should start by recognising the purposes of speed limits, which are:
- To guide inexperienced drivers away from grossly exceeding safe speeds.
- To warn drivers of expected hazard density.
- To provide a basis for enabling the police to prosecute those who drive at recklessly high speeds for the conditions.
- 1.20 Speed limits are not a substitute for drivers' judgement and must not be used to micromanage drivers' speed throughout their journeys. For the same reason, the ABD opposes Intelligent Speed Adaptation, which would encourage drivers to think they are no longer responsible for their own speed.
- 1.21 The limitations of a system of fixed speed limits needs to be recognised by encouraging the police to apply a commonsense approach to enforcement, depending on individual circumstances and the degree of danger involved. This can never be the case with automated enforcement by cameras, which operate at a fixed threshold around the clock. The proliferation of speed cameras has led to a reduction in police traffic patrols, and those remaining are subject to a target-driven culture that has eroded their discretion to decide between warning and prosecuting a driver. This is having a damaging effect on the relationship between police and public.
- 1.22 The situation needs to be reversed by reinstating police traffic patrols, scrapping targets for prosecutions, and removing most if not all speed cameras. Where cameras have been sited on the approaches to potentially hazardous locations, such as junctions, these could be replaced by vehicle-activated signs (VAS), which flash a warning to drivers if they are approaching above a preset speed. VAS have several advantages over speed cameras:
- They can be set to trigger at speeds below the speed limit, to warn of individual hazards such as sharp bends or blind junctions.
- It is almost impossible for a driver not to notice them, unlike a camera, so they are more effective at alerting drivers to a potential hazard ahead.
- They can be used to warn a driver if they are exceeding a speed limit, and can be triggered at a lower threshold than the camera enforcement level.
- They do not needlessly penalise large numbers of safe drivers.
- They are much cheaper than cameras to install and maintain.
- 1.23 Returning to Question 1, the ABD believes it is premature to be considering higher penalties for 'extreme speeders' before there has been a fundamental review of speed limits and their purpose. Many roads are now subject to speed limits that are 20mph - or even more - below the level that a competent driver would consider reasonable. Large numbers of drivers could be classified as extreme speeders, therefore, when a few years ago the same speeds would have been legal and regarded as safe. This is perverse and unjust.
- 1.24 Only once speed limits have been restored to reasonable levels should there be any move to consider applying higher penalties to those who exceed them by a large margin. In reality, a return to sensible speed limits would most likely lead to a reduction in the highest speeds, since speed limits would once again command widespread respect.
- Q.2 Do you think that 20/30mph limited roads should have a lower threshold for a 6-point penalty?
- 2.1 The ABD considers that 30mph is generally the appropriate speed limit in built-up areas. Even within these areas, however, actual safe speeds may be higher or lower than 30mph, and vary with changing conditions, so discretion needs to be used in enforcement. There are also many cases of suburban main roads, which used to have 40mph speed limits, being reduced to 30mph. Some local authorities have used liberal definitions of 'village' to impose 30mph limits on rural roads that were previously subject to the national speed limit. The ABD cannot, therefore, support a lower threshold for increased penalties until these anomalies have been corrected.
- 2.2 A 20mph speed limit may be appropriate in some residential roads where, for example, on-street parking severely restricts a driver's view of the presence of vulnerable road users. Such limits should only be imposed, however, where speeds are already at or close to that level. The ABD does not support the blanket use of 20mph limits in built-up areas on roads where 30mph is perfectly safe. It is unacceptable to reduce speed limits below reasonable levels and then condemn those who break them.
- 3. Do you think that 70mph limited roads should have a higher threshold for a 6-point penalty?
- 3.1 The ABD has long campaigned for the motorway speed limit to be raised, to at least 80mph. This would bring the UK into line with many EU countries, where a 130kph (81mph) limit applies. Table 2.1 in the consultation paper shows that the mean speed of cars on motorways is above 70mph and the 85th percentile is above 80mph, so an increase in the speed limit is long overdue.
- 3.2 On dual carriageways, while the percentages of cars exceeding 70mph and 80mph are slightly lower than on motorways, they are still high, showing that the limit is widely regarded as too low. There is a greater variation in design standards for dual carriageways, largely dependent on when they were built. It is not recommended, therefore, that there should be a blanket increase in the speed limit on dual carriageways to 80mph, but the higher limit should apply on roads designed to modern standards with full grade-separation. Other cases should be reviewed on their merits, with a lower speed limit retained on sections with poorer geometry or frequent junctions and accesses.
- 3.3 The ABD welcomes the recognition in the consultation paper that motorways and dual carriageways are our safest roads, and that higher speeds pose relatively little risk. Rather than consider whether there should be a higher threshold for applying increased penalty points for speeding, these admissions should lead the Government to support the ABD's view that the speed limit should be raised.
- 3.4 As with any other speed limit, enforcement needs to be flexible to reflect the varying degree of risk that a particular speed may bring, depending on the weather and traffic conditions that obtain at the time. There should be no rigid thresholds for determining the degree of penalty to apply.
- 3.5 The maximum fine for speeding on a motorway is £2,500, while that for speeding on other roads is £1,000. Since the Government now recognises the relatively minor risk that exceeding the motorway speed limit brings, it is time this anomaly was removed and the maximum fine for speeding on a motorway reduced to the same level as elsewhere.
- Q.4 Do you agree that we should not graduate speeding fines?
- 4.1 The ABD agrees that there is no merit in graduated fines for speeding. As explained in paragraph 3.5 above, the anomaly of a higher maximum fine for speeding on a motorway should be removed.
- Q.5 Do you agree that we should not offer 2-point fixed penalties for marginal breaches of the speed limit?
- 5.1 For all the reasons already given, marginal breaches of a speed limit where no danger is caused should not attract penalties at all. Speed limits should be restored to reasonable levels and enforced with common sense by trained police traffic officers, who must not be constrained by a target-driven culture. Under those conditions, the police would have the discretion to give advice or a warning to drivers instead of issuing a penalty. In many cases, the police would probably not even stop a driver who was a few miles per hour over a speed limit and not creating danger.
- 5.2 Legal sanctions should be reserved for those drivers whose speed, given all the circumstances at the time, posed a genuine risk to themselves or other road users.
- Q.6 Do you have any comments on the use of targeted checkpoint testing for drink drivers?
- 6.1 The ABD believes the police have adequate powers to stop anyone they suspect of drink driving. It does not object to the police setting up road blocks to check all vehicles and drivers for a range of possible offences. If a police officer then suspects a driver of having been drinking, it is perfectly acceptable to require a breath test. However, the campaign of stops conducted by North Wales Police, as described in the consultation paper, appears to have been specifically targeted at drink driving, so could be construed as random testing, which is illegal. If the Government wishes to legalise random breath testing it should make it a specific proposal so that a full and open debate can take place.
- Q.7 Do you think we should withdraw the statutory right to a blood or urine test as an alternative to a breath test?
- 7.1 If new breath testing equipment can be guaranteed to provide a sufficient level of accuracy, the ABD would not object in principle to withdrawing the right to a blood or urine test. There need to be adequate safeguards, however, against unjust convictions on the basis of a single breath test, especially in the case of blood-alcohol levels marginally above the limit. If a breath test is taken shortly after a driver has finished drinking, the breath-alcohol concentration may not be an accurate indicator of the blood-alcohol level. It is suggested, therefore, that two breath tests should always be required, the second one to be conducted not less than 20 minutes after the first, and both must show a reading in excess of the limit for a driver to be charged.
- Q.8 Please comment on the three options in respect of the proposal to take away cover for High Risk Offenders to drive after submitting a re-application for a licence, while medical procedures are being carried out:
- we move now to implement the change provided for in the Road Safety Act 2006 on the basis that we are satisfied that existing procedures allow ample time for medical examinations before a disqualification expires; or
- we develop further powers either to require an HRO to submit a medical report with their re-application for a licence or to give them that option, to be implemented probably after we have removed the cover to drive: or
- we defer implementing the change provided for in the Road Safety Act until we also have powers either to require HROs to submit a medical report with their re-application for a licence or give them that option.
- 8.1 It is clearly important that high risk offenders who are alcohol-dependent are prevented from returning to the roads, so it is reasonable that HROs should be required to prove their fitness to drive before a ban expires. Provided procedures are in place to give HROs ample time to submit this evidence, the ABD sees no reason to oppose implementation of the change provided for in the Road Safety Act 2006. The ABD has no strong views on this, however, so would not object to one of the other two options.
- Q.9 Do you agree that the costs of implementing and enforcing a judicial alcohol ignition interlock scheme would be disproportionate?
- 9.1 The ABD would not support a scheme where the costs are likely to far outweigh the benefits. Alternative schemes should be developed if they can be shown to be cost-effective.
- Q.10 What priority do you think should be given to a change in the prescribed alcohol limit for driving?
- 10.1 The ABD believes that the current limit of 80 mg of alcohol per 100 ml of blood is the best compromise between deterring irresponsible behaviour and unnecessarily penalising drivers who present a very low risk. The ABD does not condone drinking and driving, and supports the advice that ideally drivers should not drink at all if they plan to drive. The ABD opposes a reduction in the legal limit, however, partly because it would cause more responsible drivers to fall into the morning-after trap. The current limit also has widespread support, with little sympathy for those caught in excess of it. This could change if the limit were reduced.
- Q.11 What evidence are you able to offer - and what further evidence do you consider should be obtained - to support a fully considered decision whether or not to change the limit?
- 11.1 Table 3.2 of the consultation paper shows that just 1 per cent of drivers and riders killed in road accidents had a blood-alcohol concentration (BAC) in the range 50-80 mg. The figure was 2 per cent in the range 80-100 mg. The fact that a driver had some alcohol in their system does not necessarily mean that this was the cause of their accident, since they could have been the innocent victim of another road user's error or bad driving. There is almost always more than one contributory factor in any accident. There needs to be detailed analysis of the causes of accidents where a driver has a BAC in the 50-80mg range, to establish whether the effects of alcohol were a substantive cause or merely coincidence.
- 11.2 Another approach would be to carry out research to establish what proportion of drivers overall, who are not involved in accidents, have a BAC in the range 50-80 mg. If this turned out to be similar to the 1 per cent involved in fatal accidents, it would suggest that a BAC in this range does not significantly increase accident risk.
- 11.3 Since Table 3.2 shows rapidly growing percentages of drivers killed as BAC levels exceed 100 mg, there can be little doubt that alcohol is more likely to be the main contributory factor as BAC levels rise. Even without the further analysis of accident causes recommended above, however, Table 3.2 suggests there would be little impact on accidents if the limit were reduced to 50 mg. The main problem is clearly caused by those who exceed the current limit by a substantial margin, so this is where enforcement should be targeted.
- 11.4 A lower limit would make more drivers vulnerable to the morning-after trap, when they may have avoided driving while drinking the night before but still be over the limit the following day. There is evidence that, for a given BAC, a driver is more impaired when the level is rising than when it is falling. The following comes from Loosening the Grip: A handbook of alcohol information by Kinney and Leaton (ISBN 0-8016-2769-9, page 42 in the 1991 edition):
The neurophysiological basis for intoxication is not fully understood, but the intensity of the effect is directly related to the concentration of alcohol in the blood, and hence the brain. The degree of intoxication is also dependent on whether the blood alcohol level is rising, falling, or constant. It is known that the central nervous system and behavioural effects of a given blood alcohol concentration (BAC) are greater when it is rising. This is called the Mellanby effect. It is as if there were a small "practice effect" or short-term adaptation by the nervous system to alcohol's presence. Thus for a given blood alcohol level, there is more impairment if the blood level is rising than is found with the same BAC when the level of alcohol in the blood is falling.
- 11.5 Consequently, a driver might feel, and be, safe to drive the next morning while still having an illegal level of alcohol in their blood. Drivers prosecuted for having a BAC in the 50-80 mg range in these circumstances might feel they have been treated unjustly, and this could weaken support for the drink-driving laws. While no fixed limit for BAC can ever exactly match the level at which significant impairment begins in different individuals, the current limit has the virtue of being based on a large-scale study in the 1960s and is generally accepted as reasonable. This is important and the current limit should not be reduced unless strong evidence emerges that a lower limit would prevent a significant number of accidents. No such evidence appears to exist at present.
- Q.12 Do you think that a new offence of driving with an illegal drug in the body is required to make the regulation of drug driving more effective?
- 12.1 The ABD accepts that drug driving is a more difficult problem to tackle than drink driving, since many substances, both legal and illegal, may be involved and taken singly or in combination. This makes it hard to establish criteria based on concentrations in the blood for deciding when a driver's ability is significantly impaired. Nevertheless, any tightening of the existing laws on drug driving must continue to be based on evidence of impairment.
- 12.2 It would not be acceptable, in the interests of securing easy convictions, to introduce a law that made it an offence to drive with any trace whatsoever of an illegal drug, if there was no evidence of actual or likely impairment at the measured blood level. As the consultation paper recognises, traces of some drugs, notably cannabis, can be detected long after any impairment effects have disappeared. If a driver, suspected of drug driving, is tested and found to have traces of an illegal drug, but not at a level likely to impair their driving ability, they should be prosecuted under existing drugs legislation. While the use of illegal drugs should be discouraged, this worthy aim must not be pursued by punishing drivers who are not causing actual danger.
- Q.13 Do you think that such a new offence should apply to illegal drugs only, and not those that have been taken legally or prescribed?
- 13.1 On the basis that impairment to drive safely should determine whether a driver has committed an offence, there should be no differentiation between legal and illegal drugs. As stated above, the ABD would not support a new law that made it an absolute offence to have any trace of a particular drug in the body if there was no evidence of actual impairment at the measured concentration.
- Q.14 How do you think we should identify the drugs that would be the subject of the proposed offence? How should we incorporate new drugs under the proposed offence?
- 14.1 The ABD has no particular expertise to advise on the identification of drugs that might affect a driver. The side effects of legal drugs should be known, so those that impair the ability to drive can be readily determined. The ABD is concerned at the use of the words 'proposed offence', suggesting that the Government has already made up its mind in principle to go ahead.
- Q.15 Do you have any other comments about the proposed new offence?
- 15.1 This proposal needs a radical rethink. It is a classic case of the Government trying to address a complex problem by introducing a simplistic absolute offence that may lead to unjustly penalising people causing little or no danger.
- Q.16 Do you have any other comments about our drug driving proposals?
- 16.1 The ABD supports all the actions proposed in paragraph 5.27 of the consultation paper. In particular, improved drug screening and impairment testing, and streamlining the procedures for taking samples from drivers arrested on suspicion of drug driving, are welcomed. These measures represent a better and fairer way of tackling the problem of drug driving than the sledgehammer approach of creating a new, absolute offence.
- Q.17 Do you agree that we should make careless driving a fixed penalty offence?
- 17.1 Careless driving, in all its forms, is certainly a major factor in many road accidents, and far more important than exceeding arbitrary speed limits. The term 'careless driving' covers a wide spectrum of behaviours, from momentary lapses of concentration or errors of judgement to prolonged and wilful failure to give adequate attention to the driving task.
- 17.2 All drivers make mistakes or misjudgements occasionally, but other factors nearly always have to combine for an accident to occur. Most minor examples of careless driving will not, therefore, have serious consequences and will generally go unnoticed and unrecorded. It is only when an accident occurs that the lapse or error may be recorded as a contributory factor.
- 17.3 Since human beings are fallible, it is impossible to use the threat of prosecution to turn them into perfect drivers who never make mistakes. Legal sanctions in these cases, whether by fixed penalty or summons, cannot be expected on their own to prevent future errors or misjudgements. Of course, drivers whose errors result in damage or injury to others have a civil liability to compensate those affected, but criminal charges serve no useful purpose beyond retribution.
- 17.4 Consequently, the ABD does not support the proposal to make careless driving a fixed penalty offence. Instances of careless driving that are sufficiently serious to warrant prosecution should always be heard in court so that the appropriate penalties or remedial action can be applied. Behaviour that should result in such action would include drivers who routinely allow themselves to be distracted, and those who show a consistent inability to detect or respond to hazards. The latter would include drivers who travel at inappropriate speed: as explained in the section on speed, driving too fast for the conditions is a symptom of an underlying failure of observation and hazard response.
- 17.5 If there were more police traffic patrols on the roads, there would be a greater chance of careless driving behaviour being detected and addressed before accidents result. In the case of minor lapses and misjudgements, a word of advice from a police officer would have more effect than a fixed penalty notice and would cause less resentment. More serious cases should go to court, but remedial training in areas such as hazard perception could be offered as an alternative to fines and penalty points.
- 17.6 The proposal to make careless driving a fixed penalty offence is another example of the Government's belief that complex problems can be solved by penalising large numbers of people for minor offences. It has not worked with speed limit enforcement, where a ten-fold increase in speeding tickets has had no measurable effect on road fatalities, which have been decreasing more slowly than before. It will not work with careless driving either, but will only lead to further alienation of the public from the police.
- 17.7 In 2007 the ABD arranged the republication of John Leeming's 1969 book on road safety, Road Accidents: Prevent or punish? As a former county surveyor, Leeming had studied the causes of road accidents and had concluded that society was more interested in punishing drivers than seeking to reduce the toll on the roads. The ABD sent a complimentary copy of the book to leading politicians with transport responsibilities, including Jim Fitzpatrick, MP. He responded by saying that he had asked officials to take account of Leeming's book in developing a road safety strategy beyond 2010. His letter is attached as Annex D. It is clear from the consultation paper that this assurance has been worthless and the Government still prefers punishment to prevention.
- Q.18 Do you agree that the fixed penalty for careless driving should be £60 and 3 penalty points?
- 18.1 Since the ABD does not agree that careless driving should be made a fixed penalty offence, it cannot support any level of fine or penalty points. That this question was asked at all suggests the Government has already made up its mind to go ahead, making a mockery of the consultation.
- Q.19 Do you have any further comments about our careless driving proposals?
- 19.1 Only to reiterate that careless driving is too complex a subject to be amenable to a fixed penalty approach. If this proposal goes ahead, it may lead to drivers avoiding involvement of the police after an accident, in case they are given a fixed penalty. This would lead to police accident figures becoming even less reliable than they are already. With the current target-driven culture, some police officers may routinely issue a fixed penalty for careless driving after an accident, regardless of cause, knowing that most drivers will not be able to afford to challenge the allegation in court. The damage to police-public relations could be disastrous.
Driver retraining and assessment
- Q.20 Do you think we should specify a retraining course for cases where a vocational licence has been revoked on the advice of the Traffic Commissioners?
- 20.1 If the Traffic Commissioners have revoked a vocational licence and believe that a retraining course would be appropriate if the driver wants to get their licence back, this seems a sensible suggestion.
- Q.21 Do you think that disqualified drivers who are subject to a re-test should be required to take remedial training first?
- 21.1 Since compulsory retests are required only for the most serious driving offences, it would be sensible to require remedial training first. That training should focus primarily on the beliefs and attitudes of the candidates and ensuring they realise that safe driving is an exercise in continuous risk management. A driver might perform well on a test but may continue to drive dangerously afterwards unless they have the right mental approach.
- 21.2 Where a banned driver was previously licensed to drive several categories of vehicle and is required to take a retest, there seems little point in requiring a retest for each class of vehicle. It is more important that remedial training, focused on attitudes and beliefs, takes place first.
- Q.22 Do you agree that we should develop a course for people who incur penalties while subject to the New Drivers Act, linked to a new assessment for the recovery of a revoked licence?
- 22.1 While the ABD would agree this is a sensible idea, the fact that some 16,000 licences are revoked annually under the Act indicates that there is something lacking with current methods of driver training. The problem is identified by Stephen Haley, author of Mind Driving, as the common and persistent belief that "You only learn to drive after you have passed the test." This belief is the root cause of the poor attitudes of many new drivers, especially the young.
- 22.2 The ABD agrees with Haley that many of the key skills that experience brings can be taught. These include mental skills, managing risk, and teaching new drivers how to learn from experience. Teaching these skills needs to be included in pre-test driver training and in any remedial course developed for drivers who have been penalised under the New Drivers Act.
- Q.23 Please comment on the three options for the proposal in question 22:
- it could be a mandatory step to recovering a revoked licence;
- it could be offered as an alternative to revocation - a driver accepting remedial training would be allowed not to incur points for the offence which would otherwise trigger a revocation (this option would require primary legislation);
- it could be available to other new drivers incurring points that were not of sufficient number to trigger revocation.
- 23.1 Given the concern that many young drivers whose licences are revoked under the Act continue to drive unlicensed and uninsured, the second and third options are preferred. It is far better to retrain than to disqualify and force the problems underground.
- Q.24 Do you think we should change the rules relating to designated countries in the New Drivers Act? If so, how?
- 24.1 The ABD does not have any strong views on this, but any unfairness that may result from the current arrangements should be mitigated if the options supported in question 23 were brought into effect.
- Q.25 Do you have any further comments on our proposals on driver retraining and re-assessment?
- 25.1 The ABD is convinced that the right kind of driver training has enormous potential to reduce road casualties, far more than by creating any number of new offences and penalties. Safe driving is about evaluating many different factors, the risks involved and how to respond to them. This complex process cannot be replaced by insisting on blind obedience to simplistic rules.
- 25.2 Teaching the key attitudes, beliefs and risk management skills for safety must become the core of driver training, both while learning to drive and for any subsequent remedial courses. Internalising a safety culture within a driver's mind is much more effective than trying to impose it externally.
- Annex A — Suffolk Coroner's verdict concerning a fatal accident in November 1996
This is the Coroner's Verdict re Frank Gray deceased. I accept Dr Biedrzycki's report as to the medical cause of death and it follows from that that I find that the injury causing death is: "1(a) Multiple Injuries".
I now come to the most important part of my Verdict and that is the legal cause of death, what is called on the Verdict form "Conclusion of the Coroner as to the Death". Quite plainly, I only have one reasonable Verdict open to me there and that is one of Accidental Death. Part 3 of my Verdict which I have deliberately skipped until now is the time place and circumstances at or in which the injury was sustained, I
find that that was between 6 and 6.10am on the 4th November 1996 on the A134 road at Bradfield Combust, circumstances: "due to a road traffic accident".
I have had reported to me three fatal accidents on this road and these three fatal accidents follow very shortly after certain speed limits have been imposed on this road at Alpheton, Bradfield Combust and Sicklesmere.
I think that there is a very high probability indeed that this tragic fatality has the speed limits as a contributory cause. I say no more than "contributory" and I made a similar finding on the first of the three Inquests that I have held. First of all, if we look at the speed limits themselves, there is a 30 mile an hour speed limit through Sicklesmere, a 40 mile an hour speed limit through Bradfield Combust and a 30 mile an hour speed limit through Alpheton.
I am going to deal with the speed limit at Bradfield Combust first. I think almost anybody would agree that it is thoroughly reasonable to restrict cars to 40 miles per hour as they negotiate the double bend which is, coming from Sudbury, just beyond the Bradfield Manger. Not only is there a double bend there but also there is a road junction. However, I think that the 40 mile an hour speed limit extends too far in either direction from there and I think that it would be more reasonable and that more drivers would keep to the limit if where the 40 mile an hour signs are at the moment, that is at the ends of the speed limits, there were 50 mile an hour signs and a bit later on the legal speed was reduced to 40 miles an hour.
In respect of the speed limits through Sicklesmere and Alpheton, I don't think there can be any doubt whatsoever that 30 miles an hour is ridiculously slow to compel drivers to go through those two villages.
Speed limits which are unduly restrictive are harmful for many reasons but of course I'm only really concerned, and I've only got the right to mention, those respects in respect of which unnecessary speed limits are detrimental to safety. Unnecessary speed limits are detrimental to safety for various reasons: they reduce the opportunity to overtake, thereby making drivers try harder at other times, they cause traffic to bunch, they cause frayed tempers, they cause delay which makes drivers try harder at other times to make up time that they have lost. Another unfortunate effect that they have is that each unnecessary speed limit leads drivers to think that speed limits are imposed arbitrarily and therefore makes drivers less likely to observe speed limits when they ought to. Furthermore, speed limits can lead to road rage. I know of at least one incident concerning the Alpheton limit where exactly that happened.
None of the factors which I have mentioned are things which can be measured statistically and of course there is no way where a speed limit contributes to an accident can excuse a mistake by a driver, particularly in regard to matters concerning frayed tempers or road rage. A driver with a frayed temper is not going to drive anything like as well as a driver who is calm and relaxed. Similarly and obviously a driver suffering from road rage is a hazard. Drivers should of course concentrate on staying calm and relaxed and they are at fault if they do not do so but none of that alters the connection that there can be between an accident and an unnecessary speed limit and usually where there is such a connection, I suggest it's undetectable either
statistically or any how else.
I personally use this road often and have taken a special interest in what percentage of drivers observe the speed limit. I know that when I go through one of these speed limits, I am not talking about the one at Bradfield Combust but the speed limits at Sicklesmere and Alpheton at 30miles an hour, and I mean real 30 miles an hour not an indicated 30 miles an hour, because I am one of those few drivers who knows what the errors in his speedometer are.
Any time I do that I collect a queue behind me and it seems that those drivers who are keeping to the limit is mainly because they are in a queue. When a high percentage of drivers ignore a particular speed limit, and I am not talking now about these particular limits themselves, I am speaking generally, when a high percentage of drivers ignore a speed limit everyone says how terrible that is and how awful drivers are. None of them stop to think, or if they do stop to think they don't say so publicly, if say 80 percent of drivers ignore a speed limit, there might be something wrong with that speed limit itself.
In fact, if a high percentage of drivers ignore a speed limit it means that more than that high percentage disagree with the speed limit because of those drivers who keep to the limit it could be one of several other reasons. For instance, and the most important one perhaps, is fear of prosecution and points on their licence. It can be because the driver is in a queue as I've already mentioned where the leader is observing the speed limit or it could be, thank goodness, some people are just plain dutiful. So where we have a majority of drivers who ignore a speed limit it means that not only do those drivers disagree with the speed limit but a high proportion of those that keep to the speed limit disagree with it.
I believe that the A134 from Bury St Edmunds to Sudbury would become more reasonable and safer if the speed limits through Alpheton and Sicklesmere were increased to 40 miles hour, preferably with a 50 mile an hour lead in from where the existing signs are and that the Bradfield Combust speed limit should have its 40 mile an hour section made shorter, replacing the existing end signs of the limit with a 50 mph lead in.
- Annex B — A Statistician's Comments on the Transport Research Laboratory Project Report PR58 "Speed, Speed Limits and Accidents" (1994)
The above paper sets out to investigate three topics drawing upon a number of studies from a variety of countries. These include:
The problems associated with this paper fall into two areas. Those associated with methodological weaknesses in the original studies, and those that relate to the actual method and analysis presented in the review. It is not possible to comment on weaknesses in the original studies without investigating the relevant primary source material. So comments here will be confined to the method and analysis employed in this paper.
- A link between individual mean driving speeds and the likelihood of being involved in an accident.
- The effects of changes in posted speed limits and actual changes in mean driving speed.
- The relationship between both the above and actual accident frequency.
General Issues associated with the Paper's Methodology
Perhaps one of the most striking problems with this paper concerns the lack of inclusion criteria associated with the main dependent variable used in all analysis. Criteria are not established at the outset regarding the accident frequency measure employed in the analysis. That is to say although the authors discuss the validity of different measures of accident rates (page 5 and 6) they fail to establish which measure they have decided to use. This choice is mainly between the accident base rate (absolute frequency of accidents) and an accident frequency rate per distance driven. The authors dismiss a distance adjusted measure as potentially flawed given that an imperfect relationship may exist between this measure and traffic flow (page 6).
This is a very surprising approach to take at the outset of the report, but it does allow the inclusion of almost any study however seriously flawed the main dependent variable may be. Clearly, accident frequency by distance driven does at least make some adjustment for traffic flow (number of users on the road), whereas base rate raw accident frequency cannot adjust for this component. Put simply, using any study that simply measures and compares the frequency of accidents alone may essentially just be measuring the number of vehicles present in a given location. Changes in the number of vehicles using a given stretch of road will vary over time and location, and will be directly proportional to the expected accident rate. If fewer people use a stretch of road there will be fewer accidents, compared with elsewhere.
The problem here is that when speed limits have been changed traffic flow (i.e. road use) may vary, and this alone may account for changes in accident base rates. The authors fail to explore this issue in any detail, or provide any evidence that the accident frequency base rates employed in some of the studies which were included in their analysis do not reflect simple changes in road use. Indeed one of the other reasons for setting up study inclusion criteria at the outset of a review paper is to avoid selecting articles for inclusion whose results happen to support study predictions.
The foregoing points perhaps cast the most serious doubt over the paper's conclusions as a whole. That said, further examination of the analysis undertaken indicates other areas of weakness that are worth exploring.
The paper undertakes statistical analysis of different papers without attempting to obtain a standardised metric for effect, or give weightings for sample size by study. These and other aspects of meta analysis techniques are simply not addressed. This may partly have arisen because the paper does not appear to have been formally peer reviewed. This is standard practice before an article can appear in a scientific journal and provides a safeguard against spurious findings. The TRL publication procedure may be quite rigorous but a report from this organisation may not be subject to the same conventions found in peer-reviewed scientific journals.
A Reanalysis of the Paper's Main Claims
One of the most concise claims made relates to the content of Fig 3. This plots the changes in signposted speed limits against changes in actual mean speeds driven. The authors report a best fitting regression equation that indicates that for every unit change in speed limit imposed a corresponding 0.25 change in actual mean speed driven is expected. The problem with this is that the data used includes five studies which fall outside standard error confidence limits for inclusion in this analysis (as data represented a summary mean from each study standard error limits should be imposed to excluded outliers). The graph (1) below shows their original analysis without confidence limits.
Notice that those data points that fall beyond the inner curved lines ought not to have been included in the regression. They represent studies whose results are so significantly different from the rest that they ought not be included in the analysis.
Graph 1. Change in Mean Speed by Change in Posted Speed Limit
When these studies are excluded and the data reanalysed it is clear from graph 2 the much quoted link between speed limit changes and actual mean driving speed becomes much more modest. Now for every unit change in speed posted, the actual observed speed change is near to 0.120. This is basically half the figure cited in the paper (although Rsrd=0.7145, P>0.02 is still positive and significant) and suggests that the link between posted speed limit changes and actual compliance may be even weaker than the authors claim.
Graph 2. Change in Mean Speed by Change in Posted Speed After Outliers Removed
Turning to the second major claim of the paper, which relates to the association between the actual speeds driven and accident rates (however measured). This is a crucial link which the authors recognise cannot be demonstrated as a causal link in their analysis. This introduces an interesting problem for all researchers, and a variety of statistical techniques can be used to improve claims regarding the causal antecedents to a dependent measure (like accident rates by distance). These include structural equation modelling, staged regression models, and time series analysis, but the authors do not use any of these techniques. This is not unreasonable given the data that they work with, and essentially causal inferences are a matter of experimental design not statistical analysis, but nevertheless, perhaps too little is made of the fact that the associations presented throughout are not evidence of a causal chain.
Where regressions are presented which relate to the key link between actual road speed and accident rates problems appear with the presentation. For example, on page 16 the results of regressions undertaken by Fieldwick and Brown (who are themselves citing earlier work!) do not accord with the textual claims made, and in the case of the second regression equation on that page are missing the beta weight, and have involved the use of a % variable which violates assumptions of normality in regression modelling. Basically the link between speed and accidents present on page 16 is not clear. Indeed in the one multiple regression (Garber and Graham 1990) in which many predictors were entered increasing speed limits from 55 to 65 mph only accounted for 15% (of I assume variance in the overall model which uses fatalities as a dependent measure).
The problem here is that changes in traffic flow over the duration of the study are not entered as a covariate. So again even the modest 15% figure quoted may have been due to the general increase in road usage that coincided with the increases in speed limits (this would also explain the accident spill-over effect into roads with unchanged speed limits during the same period).
Turning to the data presented in figures 5a and5b. Although a shortage of time prevents the reanalysis of this data, a couple of points are worth noting. Figure 5b highlights the much cited 1mph reduction in speed results in a 5% reduction in accidents. The logistic function provides little additional power over the linear model, and this despite constraining an asymptotic effect at the upper and lower bounds. The point to note however is that the simple 1mph/5% claim is not related to the logistic function. Because of this although the logistic function can be criticised without even reanalysing the data, I shall deal with the linear regression that underpins the 1mph/5% claim.
Notice that the graph plots only to -50 on the y axis and -l0 on the change in mean speed x axis. These cut-off points give a hint as to the problem with this simple statistic. Suppose that a change in mean speed of 20mph in either a positive or negative direction could be engineered. This would mean that for a fall in speed of 20mph the linear model would predict that no accidents should occur at all (i.e. -100% on the y axis). This is blatantly absurd, as are some of the implications of the logistic model.
The basic problem is one of data presentation and analysis. In order to reach a sensible conclusion regarding the potential relationship between speed and accident rates, a linear model should relate to the actual speed from which change will occur. For example, the effect of a 10mph change in speed is likely to be different depending upon whether it is a change from 30 to 20…or 40 to 30...or 50 to 40 etc.
The problems that arise with the models presented and so often quoted arise because I believe they are an inappropriate summary that disregards the absolute speed in question in favour of speed change alone. This may provide a simplistic but misleading sound bite, but it gives little clear indication of the likely effects of speed limit changes from current actual speed limits. Not only that, but the relationship between posted speed limits and mean driving speeds seems (as above) to be in the ratio of around 1 to 0.12. This means that to obtain any of the benefits implied by the speed-accident relationship suggested in fig 5b, an 8-fold reduction in actual posted speed limits would be required for every lmph of driving speed reduction and the purported 5% decrease in accidents.
This is probably a much weaker finding in relation to speed limit effectiveness than is often either claimed or might be believed based on this paper, and even then this relationship would only hold across a certain range of absolute speeds but it is not possible to say what these might be exactly, given the type of analysis which was presented in the paper. Finally, even this conclusion only holds for as long as the base rate accident measure used throughout the paper can be seen as valid rather than simply an indicator of levels of road usage.
- Annex C — A Statistician's Brief Comments on Transport Research Laboratory Report TRL511 "The relationship between speed and accidents on rural single-carriageway roads"
I would draw attention to the following striking points, which may be of interest to others in the ABD who are familiar with factor analytic stats. The study uses Principal Components Analysis (PCA) followed by Discriminant analyses to group road types into five groups for subsequent within-group procedures. The problem here is that there are a number of anomalies concerning the PCA.
The factor solution does not appear to have been rotated; one would expect this when labelling factors, and the lack of detail in this respect is a matter of concern.
Was a two-factor solution initially selected? If not how was the initial factor selection determined?
I believe the relationship between speed and accidents in factor 1 is misrepresented in appendix B page 22. In fact as it stands table B1 actually shows a strong negative correlation exists between mean speed and accident rates adjusted for traffic flow, irrespective of the estimated loading between the accident variable and factor 1 summary scores.
A reliability alpha is not computed for the composite factor 1 scale which would have at least supported a claim that the factor solution was reasonable.
Normality and colinearity issues regarding the speed variable are simply not mentioned in respect of either the GLM or the PCA.
All the main results of this study depend on the PCA procedure. These results are no stronger than the PCA upon which grouping depended. Yet this is a technique notoriously prone to subjective biases and judgements when selecting and labelling factors. In other words using PCA gives one the opportunity to create almost any set of five groups to suit the moment. What is required here is access to the data related to Appendix Bl "Variables Used". This would allow you to undertake your own PCA to establish whether their factors are a reasonable summary of components in the correlation matrix (and it is possible that they are, we just don't know without running a few PCAs on this data).
One just wonders why a stepwise regression with road geometry entered as covariates was not selected as an alternative and more straightforward method than the procedure used. This would have removed the error variance associated with the DV and hence the need for PCA altogether (given the rationale that PCA aided homogeneity within groups).
One is also struck by the fact that so many statistical hoops have had to be jumped through in order to obtain a suitable finding. What you might ask is if even a modest effect size for the speed accident relationship cannot be found without removing all extraneous variance from their models, what does the actual result mean in practice and on the road?
In essence PCA may have allowed the researchers too much free range over road type grouping characteristics, almost like having complete freedom to choose which of many recent football matches that your team has played can go forward to support your belief that they are a good team!
If we can get access to their PCA data then I can have a look to see if they picked the "matches" in a balanced way. I am happy that any portion of this email is used in support of a request to the TRL. And do let me know if I can assist in any request you may decide to make in any other way.
(A request was made to TRL to release the PCA data but it was refused.)
- Annex D
Letter from Jim Fitzpatrick MP, 15 January 2008