Bruce Walton’s seminar about road safety statistics (25 Sept 12.00 – 1.00pm) attracted a range of questions covering various topics from Road Safety GB Academy members.
The seminar, on the Road Safety GB Academy website, focused on getting the best out of road safety statistics.
Bruce Walton has a background in IT and data consultancy and heads up the online technologies work for Road Safety Analysis, which includes important road safety projects such as MAST Online and Road Casualties Online.
Questions put to Bruce during the seminar covered topics including: the quality of STATS19 data; the importance of ‘regression to mean’; how confidence levels are calculated; and whether fatalities are the most accurate casualty statistics.
Academy members can see the questions and answers by logging onto the ‘online seminar’ section on the Academy website.
The Road Safety GB Academy is the professional development arm of Road Safety GB. It was formed in April 2013 as a result of the merger between Road Safety GB and IRSO (Institute of Road Safety Officers). The Academy runs a monthly programme of online seminars which are available free of charge to members.
Complimentary membership of The Academy is offered to road safety practitioners employed by local authority members of Road Safety GB – up to a maximum of 10 places per authority. Applications are also welcomed from practitioners working in the private and voluntary sector, subject to meeting the membership criteria. The annual cost of membership for these practitioners is £35.
For more details or to apply for membership, visit The Academy website or contact Emma Norton, head of membership.
Fascinating I’m sure, but this is a road safety forum and if someone is trying to discredit one aspect of the 3 ‘E’s of road safety, for what would seem to be personal reasons, it’s worth pointing out in the interests of balanced discussion.
Hugh Jones, Cheshire
0
In comment 3 by Bruce on RTM, he fails to address timing of the RTM reduction, which happens instantly. I have found time and again that Partnership and other claims of camera effectiveness (e.g. the recent RAC Foundation method, currently being re-written following complaints from me and others) take credit for reductions which had already happened before the cameras were installed. Accordingly, following Dave Finney’s lead, I re-worked my analysis to use monthly data as well as annual. The evidence of that utterly consistent steep and immediate fall right across the country is compelling.
Also, any camera effect should lead to a quite sudden dip in numbers in the weeks or few months following installation. Analysis continues but as yet I have found no evidence of any such dip. Equally, accident reductions should be greatest where speed reductions are greatest – but even the RACF method admits there is no relationship.
Idris Francis Fight Back With Facts Petersfield
0
Hugh – I started my “right to silence” case because as the son of a solicitor I understand legal principles including the importance of “innocent until proven guilty” and that authorities should prove their case beyond reasonable doubt, not threaten greater costs and penalties for refusing to confess. Courts elsewhere including America and Hong Kong agree with me, as did Liberty who took over my case and every independent lawyer I knew.
Now the position is worse in that innocent defendants using the para. 4, S172 defence “I don’t know” (provided by Parliament on the basis that no one should be penalised for failing to do what they are unable to do) risk 6 penalty points and higher costs if they lose – and too often accept 3 for offences they did not commit rather risk 6 and perhaps a ban.
I look forward to hearing from anyone here who has never broken a speed limit.
Idris Francis Fight Back With Facts Petersfield
0
Modestly put Idris!
I think most readers do acknowledge the time and effort you’ve put into researching this aspect of road safety and accident reduction, but at the same time, they might remember that you yourself were prosecuted for driving at 47mph in a 30 zone by one of these speed cameras and suspect you’ve been trying to discredit them ever since. Still, as far as we know, you haven’t re-offended and you’ve subsequently educated yourself on the subject to the extent that you are now in a position to inform the authorities it would seem so, there you are – speed cameras can have a positive effect!
Hugh Jones, Cheshire
0
I agree with John Morrison’s comment on the limitations of relying on KSI figures to determine road safety. In addition, it is accepted widely that a major reason for child obesity increasing is the reduction in children walking to school because parents avoid the risk of letting them walk down a dangerous road. Obviously this reduces the KSI numbers over time but increases illness and mortality caused by obesity.
Malcolm Whitmore Loughborough
0
Thousands of hours studying road casualty statistics including where and when 6m injury accidents happened over 25 years mean that I know more about RTM than anyone – including Bruce and indeed the DfT and Acpo, to both of whom Eric Bridgstock and I recently presented evidence including Dave Finney’s findings (see this site) that Thames Valley mobile cameras achieved less than nothing. Nor, as we can now demonstrate, have any others.
We now have available thousands of graphs showing that (a) no group of sites that qualify for cameras in a given 3 year period qualifies at any other time (b) the RTM return to normal levels invariably happen in the first month – not just first year – after site selection (c) invariably before cameras are installed and (d) that claims to the contrary have from the beginning at best been incompetent and at worst fraudulent.
Watch this space.
Idris Francis Fight Back With Facts Petersfield
0
Shouldn’t statistics tell us what we don’t know? Florence Nightingale used them to show what injured soldiers actually died of (infection) rather than what people thought they died of (wounds) and that knowledge eventually saved tens of thousands of lives.
It seems to me that all this erudite discourse on the validity of the latest set of statistics is a bit like those philosopers that used to argue about the number of angels that could dance on the head of a pin. It is much better to list what you actually know (and can prove) and then see if the statistics can tell you the things that you don’t know.
With the huge amount of money being hosed at cyclists, I would like to know whether they are being hit from behind, the side, or from the front in the accidents they are involved in. This would be a seriously good thing to know because once you know it you can work out ways of fixing the actual problem rather than wasting huge sums on trying to fix the problem you think you have.
Duncan MacKillop, Stratford on Avon
0
Again, a well-written answer. On the subject of the reporting of injury accidents, it’s hard to imagine a scenario where someone feels the need to call 999 for an ambulance on the basis of an injury following an accident, but not at the same time notifying the Police. I would have thought most people would. Does anyone know if the ambulance service notify the Police as a matter of course when they attend, or have attended a RTC where no Police are present?
Hugh Jones, Cheshire
0
Hugh. Bruce was asked this question, whihc touches on the issue you raise:
If the government admits that the reported casualties are one fifth of the central estimate of road casualties, what is actually happening out there and does this confirm that we use fatal as they are the most accurate?
Bruce answered as follows:
Nobody knows the total picture of “what is actually happening out there”: “accurate” information of all non-fatal casualties on Britain’s roads is impossible to obtain as no system exists which would enable the collection of that information. Those other sources of casualty figures mentioned by RRCGB are collected by different agencies for different reasons, are all subject to various kinds of reporting bias and are not directly comparable to each other. All statistics are inevitably restricted by the scope of the system which gathers their data, and must always be interpreted within that context.
Even in the absence of perfect information, there is still a pressing need for road safety strategies and priorities to be evidence led. Therefore, we must work with what we’ve got to the best possible effect. The best way to do this is not to rely solely on information about fatalities, for these reasons among others:
1. All crashes represent a system failure of sufficient gravity to cause injury, and of enough significance to those involved that someone decided to report it to the police. Not to use this information at all would be to ignore thousands of well documented circumstances which actually precipitated road user risk – which is difficult to justify.
2. Fatal crashes are so small in number that it would be dangerous to jump to conclusions from them in isolation. For example, Westminster averages only 11 fatal crashes a year (of all types), so no significant inference could be drawn from a single year with as many as 14 such incidents or as few as 7. This is even truer of casualty figures, which often provide a distorted picture because of single incidents which happen to result in multiple casualties.
3. The data collection methodology used for police reported injury crashes has been in use consistently all over the country for a long period of time, and is often based on the professional opinion of an officer who visited the crash scene and spoke to involved parties. Using this information provides a more robust approach for meaningful trend and comparative analyses than relying on other data sources, which are collected on more inconsistent bases and are derived from sources with less insight into the circumstances of the incident. In short, it’s likely to be more representative of “what’s actually happening” than any other sources.
Nick Rawlings, editor, Road Safety News
0
Nick/Bruce: A very well explained and informative answer if I may say so.
Whilst writing, was there any discussion about whether perhaps road safety should be measured only in terms of actual accidents, regardless of the consequences i.e from damage-only all the way up to fatalities? Or is it still thought that simply totting-up the number of casualties is enough of an indicator? This can be misleading after all….six casualties could mean one accident or six separate ones.
Hugh Jones, Cheshire
0
Dave:
This is the question put to Bruce (by me!):
In discussion threads on the Road Safety GB newsfeed we often hear the phrase ‘regression to mean’ – usually in relation to speed management. Can you explain what this means, and how important it is when looking at casualty/collision data?
And this is Bruce’s answer:
Regression to the mean describes the tendency for information measured over time to ‘average out’, meaning that extremely high or low measurements in one time period will tend to move closer to the mean result the next time they are measured. For instance, if a particular road exhibited an unusually high collision rate in one period, then more often than not it will exhibit a lower rate in the next, because it is more likely to move closer to the norm than to move further away.
The main importance of regression to the mean in collision analysis is that long term trends are more significant that short term spikes. If an intervention at a single given location is judged purely in terms of collisions “before and after” when the before period exhibited an unusually high rate, a fall in collisions in the after period would be the more likely result and cannot necessarily be attributed to the success of the intervention.
There are ways to allow for this tendency to regress to the mean. One is to take large samples of many sites, and apply statistical corrections to allow for the natural tendency of sites to converge towards the mean. For instance, this approach has been used in large scale studies of speed camera deployments, which clearly showed greater decreases in collision rates than can be attributed simply to regression. Another approach is to identify ‘controls’ which have not been subject to the intervention but are otherwise comparable, and measure them over the same periods. Because the controls are subject to the same regressive tendency as the sample, this makes it possible to discern changes which may be attributable to the intervention.
Of course, regression to the mean is only likely to be a misleading factor where an unusually high (or low) rate was in evidence before the intervention was deployed: this is not true for many speed camera sites, which were installed long before selection criteria based on crash rates were introduced. However, it is good practice to work with samples which are as large and representative as possible, and to compare perceived casualty reductions to controls such as overall trends in crash rates in similar areas, in order to minimise any danger of being misled by regression.
Nick Rawlings, editor, Road Safety News
0
I spoke with Bruce on the phone back when he was working at TVSRP (Thames Valley Safer Roads Partnership). I can’t remember what I asked him specifically, but I’d like to thank him for any assistance he gave me. As I’m sure Bruce is aware, RTM (regression to mean) is my main area of interest so I’d be very interested to see the questions and answers, if I may?
John, you may be right to be sceptical “about the usefulness of road safety statistics and the way they are manipulated” and that KSI figures on their own are not the whole story, but the authorities are spending millions doing what you want, although possibly not exactly where you want!
Dave Finney, Slough
0
As a local campaigner for road safety and cycling I have become very sceptical about the usefulness of road safety statistics and the way they are manipulated by highway authorities. The reliance on KSI statistics as a measure of road safety is flawed and creates an inbuilt pro-motorist bias. The conventional road safety wisdom badly needs some fresh thinking.
1) Falling KSI levels do NOT prove that the roads are safer. They show only that vehicles are now safer for their occupants. And that the NHS is far better at saving lives of victims. Safer vehicles (SUVs) make drivers feel invulnerable but increase the risk to pedestrians and cyclists.
2) Using KSI figures on their own is misleading because the numbers are now quite low. A fuller picture emerges if all crashes, including slight injuries or no injuries, are included. But KSI is regarded as an infallible guide to road safety.
3) Many roads are so unsafe for cyclists and pedestrians that these vulnerable users avoid them altogether. Not surprisingly, these roads have a ‘good’ crash record.
4) The rigid requirement that only a serious record of crashes justifies road safety improvements is routinely used in Kent to block requests for traffic calming, pedestrian crossings, safety cameras, 20 mph limits outside schools and other improvements. In Kent this cost-benefit analysis doctrine is known as ‘CRM’ (crash remedial measures). It is now an obstacle to road safety.
John Morrison, Sevenoaks Cycle Forum
0