Road Safety News

Drivers who speed 20% of the time increase their accident risk by 87%: Insure The Box

Friday 18th November 2016

Drivers who spend 20% of the time over the speed limit almost double their risk of an accident, according to new analysis from the telematics insurer Insure The Box.

Insure The Box goes on to say that ‘those who assume a little speeding can do no harm may be shocked to learn that going over the speed limit 10% of the time increases their risk of an accident by more than 42%’.

The analysis of three billion miles of driving data also reveals that smoothness of driving is a key factor in accident risk. Drivers who have ‘erratic driving behaviour’ such as swerving, speeding and habitually slamming on the brakes are far more likely to end up having a crash.

The findings were presented at the 2016 National Road Safety Conference, by Simon Rewell, Insure The Box’s road safety manager.

Insure The Box says it uses black box data to target customers who regularly speed with safer driving support.

Charlotte Halkett, general manager communications for Insure The Box, said:  “The driving data received from the black boxes installed in our customers’ vehicles enables us to directly target feedback and incentives to encourage our policyholders to improve their driving.

“For example, we have seen a reduction in speeding of 15% among customers who we had seen speeding on a regular basis. Our data shows that this is likely to reduce the number of serious or even fatal accidents on our roads.”


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 think everyone is either missing the point or is in denial of the reasonable and well-founded findings.

Billions of miles of data shows that drivers who exceed the limit 20% of the time double their risk of being involved in an accident. What it doesn't say is that these accidents are 'caused' by the excess speed; that is completely irellevant.

What can be reasonably assumed is this; because the drivers who speed 20% of the time have twice as many collisions as those who do not the risk, chance or frequency of collisions that have speed that is higher than is acceptable is increased. Because that is a reasonable and undeniable fact the seriousness of the injuries in these collisions is worse than is necessary.

What Mr Finney and others need to accept, I'm sure they don't know, is that it is not acceptable to carry speeds that are greater than is lawful and hence is unjustified into a collision whatever the cause of that collision.

As drivers who speed are more likely to collide with other vehicles, persons or property they should consider altering their speed and attitude.

Perhaps the amateur analysis and deniers of well-established fact should direct their efforts elsewhere.

Agree (1) | Disagree (2)

Nick, you are still missing my point (either intentionally or inadvertently!). What we see from ITB, I think, is a correlation between propensity to speed and accident risk. An insurance company might then be tempted to use "propensity to speed" as a proxy for the more elusive "accident risk". But we cannot derive from that that "drivers who spend a significant proportion of time over the speed limit (20%) almost double their risk of being involved in an accident" unless we can show that the speeding causes the increased accident risk (hence my question about accidents whilst speeding). What we *can* say is that "drivers who spend a significant proportion of time over the speed limit (20%) also have almost double the risk of being involved in an accident" (do you see the subtle difference). It is a case that correlation does not necessarily imply causation. We might more logically deduce from what we've seen so far that whatever it is (maybe a lack of respect for social norms, for instance) that results in the speeding also creates the accident risk. And what we certainly can't assume is that if we somehow manage to suppress the speeding (with speed limiting technology in the car, for instance) that the accident risk will decrease - because we've only treated a symptom of the underlying problem and not the underlying problem itself.
Charles, England

Agree (8) | Disagree (5)

As far as methodology goes, I've assumed that where insurance claims have been made following collisions, the vehicles' black box data have been analysed by the insurers and significant amounts of "..‘erratic driving behaviour’ such as swerving, speeding and habitually slamming on the brakes" may have been noticed over a period, which would cast doubts on the driver's ongoing ability to avoid collisions, if not directly trigger them. It would be necessary to relate that to the circumstances of the individual collision however and how much of a part the insured played in it.

I presume the insurers must have a model of what would be 'normal' driving i.e. not regular "‘erratic driving behaviour’ such as swerving, speeding and habitually slamming on the brakes" to compare with.

The assertion claimed by Insure the Box is nothing new and has long been suspected anyway - fortunately the technology now seems to be available to corroborate it.
Hugh Jones, Cheshire

Agree (5) | Disagree (5)

Nick, we don't even know what method was used. Was it directly from the data (with no processing involved) or was it, as Simon Rewell suggested, a result from one of their "risk modelling" exercises?

An evidence-led approach requires that, at the very least, we do not jump to conclusions so let's wait until ItB publish their full report with their methodology. Then we'll see what conclusions might be drawn.
Dave Finney, Slough

Agree (4) | Disagree (4)

Charles (4 posts below this one)
I fear you are muddying the water somewhat, either intentionally or inadvertently!

This is a very straightforward but comprehensive piece of analysis which draws a direct link between drivers who speed and a higher propensity to crash. It carries much more weight than, for example, an attitudinal survey because it is based on real, hard data - billions of miles of the stuff.

Nowhere does Insure The Box, or anyone else, attempt to suggest that the crashes happen while the driver is speeding so I suggest we disregard that and focus on the important message this analysis delivers - that drivers who speed are at a higher risk of crashing and therefore (presumably) causing injury to themselves and other road users.

I hope Simon Rewell from Insure The Box will correct me if my assertions are incorrect.

My final post in this thread.
Nick Rawlings, editor, Road Safety News

Agree (5) | Disagree (7)

I have had people drive into the back of me on two separate occasions, I'm not in a hurry to repeat, nor instigate that.
David Weston, Corby

Agree (3) | Disagree (2)

Give it time David! Your next (proper) collision could be just around the corner - metaphorically, or possibly in reality.
Hugh Jones

Agree (2) | Disagree (8)

As a speed violation specialist, I do admit that there is a link between those who break the speed limit and an increased amount of accidents - for all the reasons that Hugh has mentioned. (the rest I have repeated elsewhere ad nauseum)

What would be interesting would be for Insure the Box to release all the data under a non-commercial licence, so that we can determine what the accident rate of the people is for those are somewhere near the 25%ile, 50%ile, 85%il, 95%ile and 99%ile of recorded speed on any given road.

And in addition, what is classed as an accident? I have had a quick glance through the PDF that was marked as a source in the referring article, and couldn't find anything. I spend somewhere in the region of 50% of my usual journey to work in excess of any given speed limit - and to show for it I have reversed into a pure black fence at hm... half a mile an hour at the depths of night. Would this incident with this black fence count as an accident in the context of this report?
David Weston

Agree (7) | Disagree (3)

Nick, let me further explain the reason for my question. You say the people at Insure The Box have identified that link. But they haven't told us enough to support your statement that "drivers who spend a significant proportion of time over the speed limit (20%) almost double their risk of being involved in an accident". The increased risk may not be *caused* by the speeding at all, although it *might* have the same cause as the speeding. There is a distinct, if subtle, difference between the two cases. What we need to know to be able to better distinguish which is the case is whether the increased 'accident' risk is *because* they are speeding or whether the increased risk applies whether they are speeding, or not. If it's the former, then measures to reduce their speeding would result in fewer 'accidents', whilst if it's the latter then even if their cars were somehow modified to prevent them from speeding, there would be no reduction in their 'accident' rate. And a clue can be found from the proportion of accidents that happen whilst speeding.

This is probably quite a difficult distinction to accept for anyone who already believes that the data supports their preconceptions that speeding causes the increased 'accident' risk (I'm not suggesting you fall into this category, Nick), but clearly that relationship has not yet been shown to exist.
Charles, England

Agree (11) | Disagree (6)


I'm not sure that's the point. The people at Insure The Box have analysed 3 billion miles of driving data and identified a link between drivers who speed and an increased accident risk. They're not saying that the accidents happen while they are speeding, simply that, on average, drivers who spend a significant proportion of time over the speed limit (20%) almost double their risk of being involved in an accident.

This is probably quite difficult to accept for anyone who believes speeding is not linked to collisions and casualties, and therefore not dangerous (I'm not suggesting you fall into this category, Charles).
Nick Rawlings, editor, Road Safety News

Agree (9) | Disagree (6)

Simon, you still didn't tell us what proportion of the extra 'accidents' that the people who speed 20% of the time have happen whilst they are actually speeding. That is an important pointer to help understand the 'accident' relationship with speeding.
Charles, England

Agree (5) | Disagree (6)

So a number of additional points have been raised to which I wanted to provide a brief response.

Our data shows that people who speed 20% of the time have more accidents. Other factors may contribute to the accident, such as distance from the other vehicle.

The police are aware if a car is fitted with a telematics box. When they check the MIB records it shows the name of the insurer. Through our good relationship with the NPCC (formerly ACPO) they have learnt about the various providers of black box insurance.

At Insure The Box we are using the data in a constructive positive way to encourage our policy holders to change their driving behaviour and we target our communications to those we believe are at an increased risk of having an accident because of their driving style. While we are in the early stages of this programme and analysis, we are already seeing encouraging results.
Simon Rewell, Insure The Box - London

Agree (3) | Disagree (5)

If I could possibly pre-empt what Simon's reply to Charles might be and refer him (Charles that is) to my first comment on this story, carelessness and recklessness are prominent in the everyday driving habits of the regular speeder, hence ther higher probability of them being involved in a collision. They might not be 'speeding' at the time, but could - as Paul Biggs intimated - simply be going too fast, or too close, not looking where they're going, or 'erratic driving behaviour' as mentioned in the news story etc. i.e. risky behaviour leading to little or no safety margins. One can see this for one's self everyday on the roads.

As an aside to Simon, I don't believe the police can accurately and reliably determine the actions of a driver leading to up to a collision through forensic investigations and as far as I know, would only do this anyway in the event of a fatal - or likely to be - fatal collision, which the majority of collisions are not. If the police do not request the data, is it because they may simply not be aware that the vehicle(s) may be equipped with the 'black box' I wonder? On the other hand, a driver possibly facing an erroneous prosecution may well be glad of the data!
Hugh Jones

Agree (8) | Disagree (3)

Whilst the telematics devices can record speed and I presume by that they mean speeds over the default limit for whatsoever road or street one is driving, the thing it cannot tell us is just how close he was from vehicle in front. This concern is relative to all speeds no matter what that may be. Is or was the distance the recorded driver gave to the car in front sufficient or was the driver behind him tailgating? Many things are related to speed such as vision ahead which can be severely lacking if the vehicle is too close to the vehicle in front. Further, the driver cannot only see less and appreciate less of what's transpiring ahead but others cannot see him if he is hidden by other vehicles, particularly bigger ones.

Speed is not the only issue here. Merely one of them. One can be travelling at a slow speed and still become involved in a collision given that safe space is not appreciated.
Bob Craven Lancs

Agree (6) | Disagree (0)

Might help if Mr Rewell told us what the baseline 'risk' was for all those times that people were not exceeding the speed limit. If that risk is just one chance in a million then an 87% increase is hardly worth worrying about.
Duncan MacKillop. No surprise - No accident.

Agree (5) | Disagree (8)

Thanks for replying Simon. You state: "We produced this driving risk model ... and calculated ... risk levels" so can you please clarify: What exactly is the basis of your speeding/crash claim? Is it:

1) your pure data can be split into 2 groups and those that speed >20% (group A) have 87% more crashes (than group B)
2) your modelling suggests that those that speed >20% also have 87% higher crashes

If it's 1, then you've shown correlation but there will be other factors that could account for the results, such as the obvious example I gave earlier.

If it's 2, though, then all you've done is make a model that gives the results you set it up to do. You may not have included factors that lead to either or both of the effects you're comparing, and you haven't even demonstrated a correlation!

Proving your results would be easy, just run RCT scientific trials. If the product you sell really does save lives, let's prove it and then everyone can be confident in supporting it.
Dave Finney, Slough

Agree (5) | Disagree (4)

Simon, you haven't answered the pertinent question: what proportion of the accidents of those who speed 20% of the time happen whilst they are speeding? So far, we have seen no reason to believe higher accident rates are caused by speeding rather than there being some other cause for them (which maybe causes the speeding too).
Charles, England

Agree (9) | Disagree (4)

...and in addition I should note that if the police ask for telematics data, we will support them however this request is subject to the Data Protection laws. Therefore we will require that due process is completed, including the production of a court order.
Simon Rewell, Insure The Box - London

Agree (9) | Disagree (0)

Hugh raises an interesting question. In general terms when there is an accident, the police do not request telematics data as their forensic investigation can normally provide the data they need.
Simon Rewell, Insure The Box - London

Agree (2) | Disagree (5)

Simon: As one who has doubts about the reliability and accuracy of the contributory factors of collisions as currently logged by the police for STATS 19 purposes, can I ask - where a customer who has a black box fitted has been involved in a collision, is the data which may be retrieved passed to the investigators so that what the driver(s) were doing in leading up to the collsion could be more accurately ascertained?
Hugh Jones, Cheshire

Agree (4) | Disagree (0)

I am pleased to see that our results regarding drivers and speeding are generating debate and it’s good to read your comments and opinions.

Perhaps I can clarify - Insure The Box now has over 3 billion miles of driving data and all the claims that result from that driving. This data has been collected using professionally-fitted hard wired black boxes, as well as our claims processes and systems, of which real-time Accident Alerts are a critical part.

The data has been collected over six and a half years, during which time we have sold nearly 700,000 policies. We therefore have huge amounts of data as evidence, along with the understanding to produce mathematical models predicting which driving styles will statistically result in having an accident.

These particular speeding results are from multivariate analysis of our telematics data, using actuarial software to mathematically analyse 100s of millions of miles of our driving data, and all associated resulting claims, over multiple years. We produced this driving risk model a wide variety of measurable factors and calculated the statistically-validated risk levels.

Our analysis has clearly evidenced that, all other factors being equal, speeding 20% of the time results in the risk level of an accident increasing by 87%. Therefore we are proactively encouraging our customers to particularly think about this behaviour as they drive.

We will continue to analyse our data, which is constantly growing, in order to gain further insights into driving behaviour so that we can promote safer driving habits and to help reduce accident levels.
Simon Rewell, Insure The Box London

Agree (7) | Disagree (5)

In answer to Paul, just as speeds over the limit were analysed, so, according to the press release, were "...‘erratic driving behaviour’ such as swerving, speeding and habitually slamming on the brakes" which would be an indicator of someone regualrly driving at inappropriate speeds.

I think it inconceivable that someone who drives at 'inappropriate' speeds does not also excede the posted limits - I think they're in the same group and I've no doubt the collected data woud refelect that.
Hugh Jones, Cheshire

Agree (3) | Disagree (6)

Appropriate speed for the prevailing conditions/road layout, regardless of the posted speed limit, is the key speed factor. Thus a driver not exceeding the speed limit could potentially increase risk if speed isn't consistent with the prevailing conditions. This sort of 'blind' data needs to be interpreted in conjunction with in-car video evidence to be valid.
Paul Biggs, Staffordshire

Agree (7) | Disagree (5)

Is there any evidence? I can't find any on their website or in their pdf. Is this a result from a scientific trial, or is it just an assumption from a statistic? There may be other factors, eg imagine 2 groups of drivers:

Group A have jobs and drive high mileage on motorways or in rush-hour traffic.
Group B are unemployed (or students etc), driving lower mileage on A and B roads outside rush-hour.

Motorways are our safest roads and rush-hour is the slowest traffic so group A might have fewest crashes per mile with the least speeding.

A and B roads have greater crash rates and the highest speeds may be outside rush-hour. Group B therefore may have higher crash rates and also the most speeding.

As such, the speeding may be unrelated to the crash rates, as both are functions of the same factors.

More interestingly, the pdf states they've had "nearly 100,000 Accident Alerts" so why not tell us what % of those alerts involved speeding in the 5 seconds prior to impact. Now that, if presently honestly, might actually provide useful information.
Dave Finney, Slough

Agree (14) | Disagree (5)

Without knowing what percentage of their 'accidents' occur whilst they are speeding it is impossible to know whether the speeding and the 'accidents' correlate with each other or whether both might be correlating with something else. We certainly cannot conclude from this, no matter how tempting it might be, that less speeding will result in fewer 'accidents'. It could be that it is a lack of observational skills that lead to speeding, and that that same weakness also leads to having more 'accidents' - but that the 'accidents' don't necessarily happen at the same time as the speeding.

It is easy to (consciously or subconsciously) misrepresent data to support one's prejudices - what we need to be able to do is to caste our own prejudices aside and recognise when data is being misrepresented in such a way. A wise person will try to understand what the evidence is *not* telling them - a foolish one will fall into the trap of assuming that correlation implies causation, especially when that interpretation aligns with their preconceptions.
Charles, England

Agree (18) | Disagree (8)

It confirms what a lot have believed for some time, but it's good that it's been highlighted and publicised. Information from black boxes when related to the 'user', is one of the best aids to collision reduction we've had for a long time. Regular speeders are a risk to themselves and others, not just because of the action of speeding itself, but because it arises from an attitude which, in turn, can lead to other risky behaviours behind the wheel and therefore to a higher risk of collisions.
Hugh Jones, Cheshire

Agree (10) | Disagree (10)