Road Safety News

DfT provides full draft of letter about RTM at camera sites

Friday 13th December 2013

The DfT has made available the full text of a letter written by Tim Stamp, its head of statistics, about regression to mean at camera sites.

On 29 November an article appeared in Local Transport Today (LTT) about a new method to assess the effectiveness of speed cameras. The “four time period” method was devised by engineer Dave Finney to account for the effects of regression to mean (RTM) at camera sites.

Extracts from a letter written by Tim Stamp to Idris Francis were quoted in the LTT article which was subsequently covered on this newsfeed.

Anti-camera campaigners, including Mr Finney and Mr Francis, have long argued that much of the casualty reduction benefit attributed to cameras is, in fact, largely due to RTM.

Click here to download the full draft of Tim Stamp’s letter.


Comment on this story
Report a reader comment

What's your view - comment on this story:

I confirm that I have read and accept the moderation policy and house rules relating to comments posted on this website.
Your comment:
Your name and location:
Your email:

I would like to thank Richard Owen for not only providing the verified database of collisions at speed camera sites, but also commenting on my report. Mr Owen's original comments, along with my answers, can be downloaded from my website:

I notice Mr Owen has updated his comments on RSKC so I will endeavour to update my answers when I get time next year. Mr Owen's comments are very helpful because they may suggest areas where further explanation is required and he may also be articulating the opinions of those who don't understand RTM or, therefore, the FTP (Four Time Period) method.

Mr Owen might equally raise his issues with Professor Allsop, The Department for Transport and the RAC Foundation, all of whom seem to agree the FTP method is valid.
Dave Finney, Slough

Agree (6) | Disagree (3)

..and people say statistics are dull! Can't wait to find out what happens next.
Hugh Jones, Cheshire

Agree (5) | Disagree (2)

You will see that I have moderated your post below because it was well over our advisory word count of 150 words. It is still over the 150 level but closer to it.

I see in Tim Stamp's letter he talks about the need for "robust analysis" of a wider range of camera sites, and suggests that you may be interested in contributing to this process.

I also see he suggests that any such contributions "will carry most weight and influence if they are presented in objective and dispassionate terms". He also suggests that independent peer review would be "by far the best way to ensure this outcome".

On another point in your post, it is quite correct to say that Dave Finney's work is referenced in the Road Safety Knowledge Centre. In the interests of fairness, I should perhaps also point out the Dave's work was robustly challenged by Richard Owen, who was operations manger of the Thames Valley Safer Road Partnership at the time that data was provided to Mr Finney for his work. Mr Owen outlined his concerns in a response to the report which is also available in the Knowledge Centre - subscribers can find both pieces of work on this link:
Nick Rawlings, editor, Road Safety News

Agree (7) | Disagree (0)

Mr. Stamp's reply followed my meeting with him and three other senior DfT people to point out that ignoring RTM wildly exaggerates camera benefit.

My analysis of 4m+ injury accidents and 220,000 examples of sites/selection periods qualifying for cameras but most (obviously) not getting them, is possible because Stats19 provides all relevant data without any need to hold a Trial to decide where not to install them - that happens by default! It confirms nationwide the Finney's findings for 75 mobile cameras that (a) RTM fall happens instantly as the Site Selection Period ends (b) well before cameras can be installed (c) accurately returns numbers to pre-SSP levels (d) no qualifying group of sites qualifying in any 3 year period qualifies any other time (e) RTM being near instantaneous at the end of the SP is readily distinguished from camera effect (if any) which is a slower transition as drivers become aware of the cameras, over weeks or months.

My reply to Mr. Stamp is now nearing completion. I point out inter alia that the Finney Report has already been peer reviewed and published (by RSGB).

It is of course true that no analysis of what happens without cameras can, in itself, quantify camera effect. Instead my analysis confirms that Finney's method of comparing what happens after installation, not to numbers during the SSP but to numbers in the gap between the SSP and installation, is consistently valid across the country. The next step is to make that assessment instead of pretending that RTM is insignificant.
Idris Francis Fight Back with Facts Petersfield

Agree (8) | Disagree (6)

Do people understand Tim Stamp's letter? Obviously it would help if Idris Francis's analysis of national data were available but do Stamp's comments make sense?

I thought the primary job of the DfT (Department for Transport) was to oversee safety on our transport systems. I thought the primary role of the head of statistics at the DfT was to evaluate the evidence of safety systems to ensure systems are at least effective, if not cost-effective.

As a result, when presented with my new FTP method, and 3 evaluations of real data using the method, I would have expected the DfT to “take charge” and commission the independent peer-reviewed analysis Tim Stamp suggests is required. Isn't this what we pay the DfT to do?
Dave Finney, Slough

Agree (7) | Disagree (6)

There are certain driver/rider behaviours that need to be inhibited by traffic laws for the benefit of all road users, but the enforcement thereof doesn’t have to be measurable or show up in accident records for it to be worth doing. Such driver actions can simply be undesirable, intimidating and anti-social and lead to unrecorded accidents (which obviously can’t be measured anyway). Speeding is the worst of these and the most common – when an offender is caught, why does it have to be justified? Why does it have to be ‘measured’? It’s an offence. If a driver is found to be over the drink-drive limit or otherwise driving dangerously, that's an offence as well - does the effect of prosecuting him/her really need to be ‘measured’ or justified? You don’t need scientific trials - just common sense.
Hugh Jones, Cheshire

Agree (7) | Disagree (7)

Hugh is correct that the effects of speed cameras are wider than just at their location but those effects are mostly negative rather than being "benefits". The late Paul Smith cited 40 such negative effects. Any report of speed camera effectiveness needs to show how benefits outweigh their disadvantages, rather than just look at some potential advantages, or whether RTM may or may not be relevant.
Eric Bridgstock, Independent Road Safety Research, St Albans

Agree (10) | Disagree (13)

Hugh, I agree speed camera area-wide effect likely more significant than site-effect but how would you propose measuring it? If you can't measure then are you suggesting safety policies should be based on opinion? If so, whose opinion?

Speed cameras when first deployed weren’t painted yellow, many were concealed and drivers didn't have GPS warnings. Your proposal was tried but those first 10 years of speed cameras saw the worst road safety improvements on British roads since the 1950s. Speed cameras might not have caused the problem but authorities have since denied speed cameras have any area-wide effect. To find out, implement “Recommendations":

You say speed camera site-effect "is practically nil" and your opinion is supported by the government’s largest speed camera report, and much of my research, but many RSPs (and certainly the DfT, RAC etc) disagree with you. The issue could be decided by running scientific trials.
Dave Finney, Slough

Agree (7) | Disagree (9)

The number of accidents at a given spot - or ‘site’ in this context - compared to the number of vehicles passing through is so low that the chances of any measure (not just a speed camera) which is primarily designed to address driver error, actually influencing accidents to any noticeable degree at one spot is practically nil and the authorities made a mistake in trying to claim that cameras would do this. Their main benefit can’t be measured at a single site as it extends much wider.

The answer is for the authorities to continue to use the cameras as an automatic traffic law enforcement tool, but just place them where people speed, without reference to local accident history so the question of site selection, regression to the mean etc. need no longer apply (and no more reports thankfully!). After all, a speeding driver can cause havoc anywhere just like a drunk driver.

Concealing the cameras would be also get around the often overstated claims that they are ‘hazards’, in causing distraction, sudden braking, etc. The police already use unmarked cars to detect traffic offences (including speeding) without reference to the accident history in their immediate vicinity and without causing accidents, so there shouldn’t be too much opposition to this.
Hugh Jones, Cheshire

Agree (11) | Disagree (9)

Duncan MacKillop compares speed cameras to new medical drugs, and so do I at 3.8.4 here:

On my “Regression to mean” page, RTM is demonstrated in Figure 4.1 using randomly generated numbers. In my “Mobile cameras” report, RTM can clearly be observed in Figure 8.1 to 8.5 with real data. In Idris Francis's national analysis (not yet published), RTM is, if anything, even clearer.

Although Tim Stamp agrees my FTP method is "straightforward and logical", I get the impression that, no matter how good the analysis, many will never believe an answer if it doesn't support their opinion. This is why, as Duncan suggests, speed cameras must be run within scientific trials (known as Randomised Controlled Trials). People will struggle to argue with the results of scientific trials!

BTW, I don't consider myself an “anti-camera campaigner”? I've never campaigned against cameras, I've only campaigned for competence and honesty in road safety.
Dave Finney, Slough

Agree (11) | Disagree (8)

Surely this would only apply on trunk roads? (I thought I'd get in first with that one)
Hugh Jones, Cheshire

Agree (17) | Disagree (0)

RTM is one confounding factor, but so is the Elephant hypothesis. This hypothesis states that with the complete lack of evidence to the contrary the results would be the same if you placed an Elephant at a site instead of a camera. This is quite common in drug trials where only some of the subjects are given the actual drug with the rest either getting a placebo, or nothing at all.

An analysis of the before and after rates in sites where cameras have been used compared with those sites where Elephants have been used would no doubt reveal some very interesting results. The fact that cameras are the only game in town colours the results from any studies in the same way that drug trials would be rendered useless without placebo and control groups also being analysed.
Duncan MacKillop, Stratford on Avon

Agree (11) | Disagree (6)