Swimming with sharks – the truth about Safety In Numbers

You knew it was dangerous. But you let people go swimming anyway.

Let us imagine a situation on a holiday resort island – call it ‘Amity Island’ – which is subject to increasing numbers of shark attacks on its numerous beaches. The authorities are rightly concerned, and commission a study, which discovers a correlation between the numbers of people swimming at different beaches, and the number of shark attacks.

Namely – it is found that there are more shark attacks, per person swimming in the sea, at those beaches were there are fewer people swimming. And, conversely, it is (relatively) safer to swim at beaches where there are plenty of people swimming. On these beaches the rate of shark attacks, per person swimming, is lower.

It is then hypothesised that perhaps it is actually the numbers of people swimming at these more populated beaches that is somehow responsible for the lower rate of shark attacks. Perhaps it is something connected to the numbers of people in the water that discourages the sharks from attacking? In other words, the reason might be that the sharks’ behaviour is modified by the number of people in the water – lots of people mean they are less likely to attack.

No hard and fast evidence for this change in the sharks’ behaviour is discovered, but nevertheless a plan is put forward to encourage more people to swim in the water on those beaches where people are more likely to suffer a shark attack. After all, there is a proven link between the risk of shark attack, and the numbers of people swimming. Thus, if you can simply get more people swimming on the more dangerous beaches, then swimming becomes (relatively) safer for all. The more people that swim, shark attacks become less likely (for each person swimming).


Except, of course, it isn’t. This isn’t a very sensible way of dealing with the problem of shark attacks, for one particular reason. There might be confounding variables, which make the best beaches for swimming – hence the most popular – simultaneously the least likely to suffer shark attacks (per person swimming).

Another problem is that local knowledge of the beaches, and where sharks are likely to attack, might explain why fewer people are found swimming at the more dangerous beaches. In other words, sharks are keeping people away from certain beaches, rather than people keeping sharks away from others. Put simply, the correlation could run in the other direction.

If either of these explanations are true – and they could well be – then simply encouraging more people to swim on the most dangerous beaches is hopelessly misguided, perhaps even immoral.

It’s immoral in the further sense that it doesn’t actually reduce the absolute risk of shark attacks – which could be achieved by, for instance, erecting barriers to keep sharks away from the beaches completely, or similar measures. The ‘Safety in Numbers’ policy only aims to reduce the relative risk of people swimming. Even if it works, more people could actually die on beaches as a result, although each person swimming would be relatively safer.

Yet this kind of approach to the safety of cyclists and pedestrians has gained a worrying degree of traction over the last decade, despite a similar lack of evidence behind the theory. One of its exponents has been the Mayor of London, Boris Johnson, who by coincidence or otherwise has written

My hero is the mayor in Jaws. He’s a fantastic guy, and he keeps the beaches open, if you remember, even after it’s demonstrated that his constituents have been eaten by this killer fish. Of course he was proved catastrophically wrong in his judgment, but his instincts were right.

‘Safety in numbers’ is attractive as a theory because it requires very little financial or intellectual investment in trying to improve conditions for the most vulnerable users. All that needs to be done is to ‘encourage’ more walking and cycling, without the hard effort required of improving the environment itself (which, incidentally, does have a proven record of making walking and cycling safer).

The modern history of this idea can be traced back to a paper by P L Jacobsen, published in 2003 in the journal Injury PreventionLooking at various data sets at a variety of scales (the amount of walking and cycling in 68 different Californian cities, the distance walked and cycled in Danish towns, walking and cycling rates in a number of European countries, and the total amount of annual cycling in the UK and the Netherlands) and comparing them with fatality/injury rates, Jacobsen argued

A motorist is less likely to collide with a person walking and bicycling if more people walk or bicycle. Policies that increase the numbers of people walking and bicycling appear to be an effective route to improving the safety of people walking and bicycling.

The one graph in Jacobsen’s paper that might be of particular interest to UK cycling campaigners – who are often deeply attached to the Safety in Numbers concept– is this one –

Does “increasing the numbers of people walking and cycling” mean that fatalities are less likely, on the evidence of this graph? Well, not really. The fatality rate in 1950 is virtually identical to that in 1999, despite a vastly greater amount of cycling in 1950. Likewise cycling was much safer, relatively, in 1999 than it was in 1972, despite a virtually identical amount of cycling being carried out in the UK. And perhaps most damningly of all, from 1983 to 1999, people cycled less, and yet it became safer to do so (the graph runs down towards the origin).

Those who think that the best way to create safer cycling is just to get more people cycling should take a long hard look at this graph, because frankly it’s a complete mess, that shows no clear relationship at all between the amount of cycling, and relative safety.

However, Jacobsen himself is undaunted, stating that the relationship is ‘complex’, and ties himself in knots coming up with intervening variables that account for the way the graph shoots all over the place (intervening variables like, for instance, the amount of driving, or even seatbelts, which are not accounted for by the ‘Safety in Numbers’ hypothesis).

No such problems with another graph Jacobsen includes, comparing cycling rates with safety, but this time in the Netherlands, from 1980-98 –

By comparison, this relationship is more straightforward. More cycling, from 1980 onwards, corresponds more or less directly with greater relative safety.

While Jacobsen concedes (somewhat vaguely) that, over this period,

the Netherlands has implemented a range of policies to encourage people to walk and bicycle and make them safer. These efforts have succeeded in increasing bicycle use and decreasing risk.

He then goes on to state that

it is improbable that the roadway design, traffic laws, or social mores, all of which change relatively slowly, could explain the relationship between exposure and injury rates. The more plausible explanation involves changes in behavior associated with changes in the amount of walking and bicycling.

In other words, the reason cycling has become safer is not because of changes to the physical environment; rather, the “more plausible explanation” is that drivers are behaving differently now that they are surrounded by more cyclists.

I’ll leave you to judge whether that is indeed more plausible; what is interesting is that this explanation of improved driver behaviour as a consequence of being surrounded by more walkers and cyclists is only a hypothesis. It is not substantiated by Jacobsen. He cites research showing that motorists drive slower when surrounded by more pedestrians (not, in itself, an indicator of better or safer driving), and also speculates that

motorists in communities or time periods with greater walking and bicycling are themselves more likely to occasionally walk or bicycle and hence may give greater consideration to people walking and bicycling.

The operative word here being ‘may’. There isn’t any research cited; this is pure speculation.

the most plausible explanation for the improving safety of people walking and bicycling as their numbers increase is behavior modification by motorists when they expect or experience people walking and bicycling.

And that’s it. It’s very thin stuff indeed for what has proved to be the foundation of nearly a decade of ‘safety’ promotion amongst transportation engineers, urban planners and walking and cycling advocates. Just last month I read Rob Penn arguing, in the Observer, that

Right now… we need to keep transmitting a positive message, a message that extols the utilitarian, democratic virtues of the bicycle, a message that encourages people to wheel their steeds out of the garage and into the sunshine. Which brings me back to my first point: if you increase the number of cyclists, it becomes safer to cycle[my emphasis]

This argument has been repeated so often it is perhaps a little unfair to single one person out for parroting it – this is just the most recent example – but the causal relationship posited in bold simply isn’t substantiated. Jacobsen only identified a correlation (and a deeply shaky one at that, if we re-examine the graph of cycling and risk in the United Kingdom). He did not establish a causal relationship – only assumed one! – and on top of that only speculated about a possible cause that might explain that relationship, namely improved driver behaviour.

That cause has not been established. In a critical appraisal of the ‘Safety in Numbers’ concept by Rajiv Bhatia and Megan Wier, published last year in the journal Accident Analysis and Prevention, the authors write

A popular explanation of [the non-linear relationship between numbers of walkers and cyclists and the rate of injury or death] is that an increasing numbers of walkers or cyclists directly influences hazardous behaviors of motor vehicle drivers; hence, the term “Safety in Numbers” (Jacobsen, 2003). However, this causal inference has not been subject to a thorough critical examination in the peer-reviewed literature.

In other words, we still don’t have any evidence that driver behaviour is actually modified or improved by the presence of more cyclists to the extent that there might be pay-offs in terms of the safety of the latter group.

Beyond serious concerns about the lack of evidence for a plausible mechanism behind the ‘Safety in Numbers’ effect, the authors also provide a succinct discussion of the serious challenges to the relationship itself, namely Temporal Direction and Confounding (these two problems were touched upon in my hypothetical introductory discussion of shark attacks).

Clearly the temporal direction of the relationship needs to established; we need to see whether more pedestrians or cyclists on a street subsequently ‘create’ more relative safety, rather than safer conditions leading to more walking and cycling. However, just as with the ‘explanatory mechanism’, the evidence for the relationship running in this direction is lacking. The authors write

to our knowledge, there have been no experimental or prospective studies that have evaluated changes in pedestrian volume on changes in injury frequency or rates.

Conversely –

research demonstrates that walkers have a strong value preference for lower speeds, lower traffic volumes, and greater buffers between pedestrians and motorists.

(It is unstated in the paper, which largely focuses on pedestrians, but research into the preferences of cyclists quite obviously finds similar results). In other words, the best evidence suggests that the relationship runs in exactly the opposite direction to that commonly assumed – namely, it is safety that is producing numbers.

So if it turns out that people are staying away from beaches where there are shark attacks, quite obviously it shouldn’t be sensible policy to encourage more people to swim on those beaches. Yet that is what the Safety in Numbers concept, as so commonly and crudely interpreted, suggests.

Likewise on the subject of Confounding, Bhatia and Wier are quite clear that safer environmental conditions could result in both more walking and cycling, and fewer injuries; traffic law and enforcement, and social norms, could have similar (although less well-established) effects. This is the most plausible explanation for the relationship exhibited in the Netherlands; an explanation Jacobsen was quick to dismiss.

The authors further argue that

Since, at the population level, more walking typically means less driving, a reduction in collisions per walker could also potentially be explained by lower traffic volumes, an established and potent determinant of collisions

In other words, more people walking and cycling may indeed result in greater relative safety, but not because of the modified behaviour of drivers. Instead, more straightforwardly, the relative safety comes from the fact that there are simply fewer drivers on the road.

It is this explanation – alongside a failure to change the physical environment for cycling, like in the Netherlands – that I think most persuasively explains why the increasing numbers of cyclists on Britain’s roads has not seen a corresponding reduction in the rate of injuries and fatalities, Indeed in recent years, the ‘Safety in Numbers’ relationship has gone into reverse – despite more cycling, the rate of KSIs (Killed or Seriously Injured) has increased, which is the exact opposite of what the proposed effect suggests should happen. As The Times noted recently

The rate of cyclists killed and seriously injured measured as a proportion of distance travelled rose by 9 per cent in 2011. It was the third consecutive year in which the rate of death and serious injury amongst cyclists had increased. The data undermined Government claims that it is becoming ever-more safe to cycle as increased numbers of cyclists take to the roads in Britain.

This shouldn’t really surprise us – if we trace the graph of fatalities/billion km travelled against the amount of UK cycling presented in the Jacobsen paper, we could easily be heading back towards 1983.

The problem is two-fold. There are still plenty of motor vehicles on the roads, and the physical environment has not changed. Clearly if everyone stopped driving, then an unchanged road environment would become considerably safer. You would only see single person accidents, or cyclist-on-pedestrian,  or cyclist-on-cyclist collisions, which are far less serious.

But despite more people cycling, motor traffic has barely declined at all. So there is still, approximately, the same volume of hazard that cyclists have to deal with, on roads that are just as dangerous for cycling on. It is deeply unrealistic, therefore, to expect the rate of injury and death to decrease.

At some distant point in the future we may hit a point at which sufficient numbers of people walking and cycling, and few enough are driving, that relative safety is achieved, without any change to the environment, but that is some way off, and the route to that point is unclear, and may be littered with the bodies of cyclists and pedestrians.

In a paper published in 2009, Rune Elvik argued that a doubling of pedestrian and cyclist volume, with corresponding mode shift away from driving, would not, alone, reduce the KSI burden, and may actually increase it. Indeed, he suggests that, without any change in the environment, it is only when the amount of driving is reduced by 50% (with corresponding mode shift to walking and cycling) that we may see a reduction in the total KSI burden. That is an enormously long way to go by ‘Safety in Numbers’ alone, with an  increasing KSI burden in every year that we attempt to get people to switch to bicycles. Freewheeler parodied this attitude in a recent post, writing

In a lively question and answer session the speaker was asked about recent rises in cyclists killed and injured on Britain’s roads. Sam said this showed that more people were cycling, which was very good news indeed. However, numbers were not yet great enough for the safety in numbers effect to work, which is why we must get more people cycling by spreading the message that cycling is safe.

But unfortunately this parody is not very far from the truth.

This brings us to a final problem with the ‘Safety in Numbers’ thesis, again touched on in my ‘shark’ discussion. Namely, that it is not particularly concerned with the total burden of injury or death, only with the rate at which deaths and injuries are suffered by a particular group of road and street users. ‘Safety in Numbers’, crudely applied, is quite content to see an increase in the absolute numbers of people being killed or seriously injured, if the rate at which they are being killed or seriously injured decreases.

This has serious implications for the way the safety of pedestrians and cyclists might be dealt with. Just like our resort of Amity might have decided to forgo the expense of erecting shark nets to keep the sharks away – which would have provided a higher degree of both relative and absolute safety – opting instead for the cheaper option of encouraging ‘Safety in Numbers’, so transport and urban planners might similarly decide to forgo the expense of redesigning streets in order to lower both the relative and absolute KSI toll, likewise plumping for the cheapest, and politically easiest, option. As Bhatia and Wier write

A danger with the SIN [Safety in Numbers] policy prescription is that it potentially shifts the cause and responsibility of the problem to walkers – the victims of pedestrian-vehicle collisions – and shifts attention away from fundamental change in roadway design, revision of speed policies, increased investment in pedestrian and bicycle infrastructure, gross disparities in conditions that support safe walking and bicycling in our communities, and the high absolute numbers of people being killed or injured while engaging in these sustainable transportation behaviours in the United States and elsewhere.

There is, I think, an analogous situation in Great Britain, where very little attention has been paid over the last decade to adjustment of the physical environment to make cycling safer. (There are signs – small signs – that this is about to change). Most of the attention has instead focused on behaviour, particularly the behaviour of cyclists, with attempts to train them to become safer, and also on how they act and dress. How much of this is down to widespread adoption of the ‘Safety In Numbers’ hypothesis, I cannot say, but it has certainly provided an outlet for those in positions of power who might be less inclined to grasp the nettle and make hard decisions about changing the physical environment for cycling. Again, without wishing to single anyone out, I notice that Transport for London’s Ben Plowden was sticking to this script at the recent London Assembly Transport Committee on Cycling Safety –

I’ve been cycling in London for many years. You used to see hardly any other cyclists on the road, but it’s very different now. I think you can get a “virtuous circle” where you get more cyclists, and improved behaviour by road users, leading to a “tipping point” where things change significantly. Now I see in some places 40 cyclists at every cycle of the lights. That must be having an effect on people’s perception.

Quite where this ‘tipping point’ might lie is, as will hopefully be apparent from the discussion here, really quite unknown, and the effects of ‘Safety In Numbers’ in the meantime very questionable. (The report by David Arditti on this meeting is worth reading in full because it gives a real flavour of the pervasiveness of ‘Safety in Numbers’ thinking).

Bhatia and Wier argue that the environmental determinants of injuries to vulnerable road users have been studied in depth, and it is clearly and consistently motor vehicle volume and speed, and roadway design, that are the important predictors. This evidence suggests, in their words, ‘an urgent need’ to focus on policies that reduce exposure to risk. It does not – quite clearly – suggest that we should employ policies that simply ‘encourage’ more people onto the streets, when those streets have relatively undiminished numbers of motor vehicles on them, and that are still designed in ways that create conflict.

In other words, ‘Safety in Numbers’ is a serious distraction from the actual business of making cycling safer. It is deeply unfortunate that instead of talking about the real problems facing the safety of cyclists, that debate about danger – both actual and perceived – is often discouraged by people who fear potential new cyclists might be ‘put off’. (It is also a handy way – if we are being cynical – for politicians to silence criticism of the dangers on the roads they are responsible for).

It may or may not be true that people are actually discouraged from taking up cycling by how it is discussed (I personally believe any such effect must be negligible in comparison to the everyday perception people have of the streets and roads they walk and drive along, and how inviting they might be for cycling). But if the only reason we want more people to cycle is to diminish the relative risk faced by current cyclists, then frankly I don’t think we should really care about whether they are put off. Because that’s simply not the right approach, both morally, and practically.

This entry was posted in Infrastructure, London, Safety In Numbers, The Netherlands. Bookmark the permalink.

38 Responses to Swimming with sharks – the truth about Safety In Numbers

  1. Many thanks for this – ugly graphs and all. You have written a well sustained and credible account of the wisdom I seem to have acted on over the last 30 years or so: avoiding the shark-infested beaches wherever and whenever possible. I take care on sections of road that feel wrong: Idismount and observe for a few minutes to see where the sharks are and what might be done to avoid them.

    As we have seen in recent years, there are plenty of intersections in the UK that are lethal and should be avoided. “I can cope” can be just foolish bravado. Local politicians should be encouraged to make improvements and be applauded when they do. National politicians must be harrangued into making resources available for the changes to be made.

  2. Andrea says:

    Excellent article.

    There is one other aspect of the SIN fallacy, worth mentioning: the increase in numbers is not uniform in time and space. If you ride outside the rush hour or in certain sections of London, you are vastly outnumbered by motor-vehicles.

    My experience (or my feeling) is that drivers’ attitudes at times and places where they have the upper-hand is worse than it would have normally been, to compensate for the frustration and envy suffered when cyclists have critical mass.

  3. Figure 5 looks like a shark fin – I think this is proof that is safe to swim/cycle 🙂

    Safety in Numbers is a stupid theory to apply to road safety as it is simply a percentage game. The more fish in the shoal, the less likely the predator is likely to eat a particular individual fish.

    2 beaches each with 1 shark that kills one swimmer day – Beach 1 only 1 person swims at a time. Beach 2 10 people swim a day. Which beach is safer? Beach 1 with 100% of people swimming per day killed, or Beach 2 with 10% of people swimming per day killed?

    You only have to have watched The Blue Planet to know that a large shoal of fish doesn’t put off the tuna/bird/seal from attacking the shoal and eating a fish. They don’t go “ooooh, there’s 4000 fish, they might get together to fight back; I won’t eat them today”. Mind you, a car driver might be less likely to shout abuse at 4000 cyclists. But if he decides to drive into you, there’ll be safety in numbers as he can’t run all 4000 of you down at once.

  4. disgruntled says:

    Has anyone laid alongside the figures in the second graph with the implementation of better cycling infrastructure in the NL? After all the dates (1980 – 1998) seem to coincide with the post Stop de Kindermoord years, and the turnaround in Dutch road design.

  5. Wyadvd says:

    Ok, ok axe nicely ground thank you

  6. A pleasure to read such a well-researched and argued article.

    Isn’t the difference, though, that *people* drive cars, and *people* walk, and *people* cycle? Same kind of mind, not different animals?

    Wouldn’t a better comparison be people in motor boats and people swimming? Thankfully, people aren’t out to eat each other like a shark is to fish. However, they may be watching who’s sunning themselves on their deck than where the boat is going.

    Over time, some of the people driving the motor boats may anchor up and hop overboard to swim themselves – a point you make in your article.

    I am not, though, disagreeing with your fundamental thesis: namely that a *much better* way to improve safety is – doh! – to have safer infrastructure for *all* [car,bike,walking] not just focused on one [car] mode. ie 20mph, segregated paths, tailored interesections/roundabouts, as appropriate for the situation.

    Aside: the nice thing with getting more people swimming [cycling] is that more people in boats [cars] will see that the water [road] is nice and come in for a dip [ride].
    Besides, there’s more people sunning themselves on the beach, where boat’s can’t go. 😉


  7. Very good read.
    The “safety in numbers” theory also misses the fact that it applies only at rush hour. I am more nervous riding in certain places outside of rush hour because I just know there will inevitably be a moron that will buzz me at 30+mph. Fewer traffic means they are (potentially) less likely to hit me, but because they go much faster they are much more likely to kill me.
    @disguntled: I don’t know what kind of data you would want to plot overlaid on fig 6, maybe kilometres of cycle path built? Anyway if you want to get the data from graphs you can use http://getdata-graph-digitizer.com/

  8. 3rdWorldCyclinginGB says:

    It’s quite interesting if you plot the two figures on top of each other to the same scale – you can see that the NL trend does NOT plot anywhere near the UK data. There are far fewer fatalities for the same distance in the NL data, which does show the trend expected in safety in numbers. You COULD also argue that there are two separate safety in numbers trends in the UK data (1950-1972 albeit the wrong way round chronologically, and 1972-1983). Those were the days. Unfortunately it seems that the UK has diverged significantly from any safety in numbers trend since 1983 and actually appears to have generated a situation where things get worse for cyclists as numbers increase. On the other hand, whether or not a negative distance-fatality correlation actually represents safety in numbers may be irrelevant. As long as you can generate such a correlation then I expect that cyclists would be in a good situation regarding their safety. The fact that we haven’t had such a correlation in the UK for years is very worrying indeed.

    However, we (and the powers that be) in the UK can witter on about statistics and try “innovative solutions” as much as we/they like, but the simple fact is that only the Dutch principles of sustainable safety can DEMONSTRABLY be shown to work – as the two graphs illustrate – the NL data are mostly off the bottom of the UK data on the fatalites axis. Until that is taken on board and applied in the UK everything else is just gambling with peoples’ lives – I’ve read quotes where an official has said they will try this or that, then consider revising it if it doesn’t work. Sounds very reasonable of them until you realise that what “not working” means is cyclists being injured or killed.

  9. Thanks for this analysis. Whenever I’ve looked at graphs, such as the one in Fig. 5, that claim to show how biking has gotten safer, I’ve thought I just wasn’t understanding the graph, or that the axes were confusingly labeled or something left out of the explanation. Your clear and cogent analyses statistical and other arguments is always impressive.

  10. You forgot to say that the swimmers have to be nice to the sharks and obey all the rules of the water they are swimming in (and ideally pay sea tax), before the authorities will act to make things better. Because it is the swimmers fault for getting in the way of the sharks, and therefore they are to blame if they are killed or seriously injured…

  11. David says:

    I completely agree that the only way to increase the modal share of cycling is through segregated infrastructure and thanks for the detailed analysis in this piece. Couple of points though:
    i) do you really want to hang so much on Jaws which, brilliantly well made picture as it is, is notoriously (and tragically) wrong about shark behaviour?
    ii) I’m interested in how things actually change. In the Netherlands I believe change came about because cycling safety was made a child protection issue. I don’t think the CEOGB strategy of focussing on the odd City lawyer or Murdoch journalist who gets left-hooked is going to create a mass consensus for change. However getting the highest number of cyclists possible under current conditions will create more of a constituency for change. In that regard while Safety in Numbers has nothing to offer in day-to-day safety, it does have something to offer as a basis for more support for long term infrastructure campaigning.

    • michael says:

      Even if there is _some_ validity in the safety in numbers theory, its clearly never going to get us very far on its own. At best sheer numbers in themselves might reduce the death-rate very slightly, but its an insult to cyclists to suggest that such a slight improvement should be the long-term target.

      The idea that white-van man and sportscar-berk are all going to cycle on roads in their spare time and thus become more considerate of cycllists is a ludicrous pipe-dream. Even if 2 out of 3 motorists becomes a part-time cyclist and gets more sensitive to bikes on the road, the remaining 1 in 3 can still kill plenty of cyclists. The worst drivers are also, I suspect, the least likely to either cycle or even know anyone who cycles.

      The statisitcal correlations shown in that article look very, very weak as a basis for protecting lives. Far better to actually have proper infrastructure that physically protects people than rely on dubious statisical correlations.

  12. wyadvd says:

    I have to say I am not an expert in research, but I do have a research based degree. If you choose to critique an article then your first stage is to look at the methodology of the research. You have not done that. In fact all you have done is state a series of opinions coloured by your personal axe to grind in regard to some minor conclusions of the paper.

    You have deliberately omitted graphs which are very convincing in Jacobsen’s paper, namely the ones that collate the data from a large number of different cities in California where cycling rates vary wildly from city to city. This to me seems very a very strong data source to demonstrate any safety in numbers effect.

    I do not seem to be able to paste images but here are links to the graphs which appear to show an almost perfect inverse parabolic curve:

    there is a similar graph sampling a large number of different Danish cities where cycling rates varied wildly between the cities at the time of the sample;

    Have any of the people commenting on this blog taken the time to read Jacobsen’s paper properly or do you just blindly accept the skewed opinions of this blogger?

    Have you considered that the BMJ is a very high quality journal that only publishes papers which meet very strict inclusion criteria as to data sources and methodology?
    Im not saying Im right or wrong, but can we have a bit of balance. Its good research from what I can see.

    • michael says:

      I do take your point about the BMJ, while a lot of non-reproducable research does get into even prestigious journals (in part I think because of the limitations of frequentialist statistics) I agree it would be very surprising if they overlooked fairly basic flaws.

      All the same, those graphs don’t in themselves seem to demonstrate which way round the link runs, i.e. is there more cycling in safer cities because they are safer?

      • Wyadvd says:

        I have since done a bit of further research on this paper and found that its chief detractor is in fact the arch nemesis John forester who spotted a mathematical anomaly in the way the data relating to the graphs I mentioned which would have produced a parabolic curve with almost random data! So I give up!

        • michael says:

          Oh. Thanks for doing the legwork I’m too lazy to deal with! Funny though that it would turn out to be debunked(?) by someone I don’t particularly agree with either!

    • michael says:

      Yes, when you say “This to me seems very a very strong data source to demonstrate any safety in numbers effect.”

      I would ask you why you say that, becauase I don’t see that it demonstrates the effect works the way round you say it does. If you have a researched-based degree you surely know that correlation is not necessarily causation?

      I’m not meaning to be sarky or confronational, because I haven’t read the full paper (am too lazy, quite honestly!) and if you have I’d be genuinely interested to know if there’s something more in it that addresses this point.

      • Wyadvd says:

        I admit the conclusions are very definite causative statements based on a largely statistical study. My main problem is that to me 20 deaths per billion km cycled makes cycling a very low risk activity to start with . I have about 300000 km left to cycle in my life time so I’m happy with that. I do much riskier things . Fat people have a 1 in 3 chance of dying a slow and painful death before they even look at a bike!

        • michael says:

          I partly agree with you. Its certainly annoying that the same government that hectors us for all being too fat, and then declares that obesity is purely a question of personal responsibility, also panders relentlessly to the motorist and in practice, occasional PR exercises aside, appears to want everybody to drive everywhere.

          And as a late-comer to cycling, one of the benefits is certainly that its a relatively enjoyable way to incorporate exercise into your everyday life.

          But raw statistical statements about ‘danger’ really miss the point. The point is that so much of the work of keeping those figures low falls on the cyclist him/herself. Cylcing is so much more unpleasant than it needs to be, due to the stress and effort of dealing with abysmal road-design and arrogant and irresponsible drivers. Most potential cyclists simply don’t cycle because it looks too stressful. The figures may be low partly because the most vulnerable and least confident/assertive potential cyclists stay off the road entirely.

    • Wyadvd –

      I think your comment is based on a slight misunderstanding. I am not disputing that a non-linear statistical relationship exists between the numbers of people walking and cycling, and their relative safety, at least in some of the data sets Jacobsen presents. I have omitted these graphs not because I am attempting to be devious, but because the relationship is not being questioned!

      I am taking issue with something rather separate – namely, the implications that have been drawn from that relationship. It has simply been assumed that the relationship runs in one direction (that numbers produce safety) when the best evidence suggests, instead, that the relationship runs in the other direction. Likewise that there could be confounding variables (for instance, a better environment that produces both more safety and more numbers). I suspect this is the real reason behind the confusing UK graph in Jacobsen’s paper, included here. Many other factors have a much more powerful influence on safety than mere ‘numbers’.

      There is also a lack of evidence for the causal mechanism that makes drivers behave better (the assumed reason for more safety), and finally there is the problem that ‘Safety in Numbers’ presents those in power with a reason to do nothing concrete or substantive for walking and cycling safety. This is happening in London right now.

      Please read my post again with this in mind!

  13. michael says:

    As someone who was a non-driving non-cyclist pedestrian and user of public transport for decades before finally risking taking to two wheels, I have to say I snort with derision at the notion that its ‘talking about the dangers of cycling’ that puts people off.

    I was detered because the roads I walked beside and tried to cross every single day looked scary and unpleasant, Abstract stats and propaganda had nothing whatseover to do with it. Having taken the plunge, I find its not quite as bad as I feared, but, yes, still pretty bad..

  14. Dear As Easy,
    I have given support and praise for a number of your posts, so here’s an occasion to disagree with you and your commenters (apart from wyadvd).

    I think you have it quite wrong: here are a few points;

    1. The origin of SiN goes way back before Jacobsen. Adaptive behaviour by motorists was spotted by the father of road safety academic research, Reuben Smeed, in the 1960s. His papers were analysed by John Adams to show changes over time in road casualty numbers, and how they occur in ways which are often spontaneous and not the result of official “road safety” interventions. Adams, with Gerard Wilde and others, also shows how risk compensation – adaptive behaviour by road users – occurs in numerous specific cases. Take a look at “Risk and Freedom” and “Risk” by Adams, downloadable at http://www.john-adams.co.uk/books/ . Also see my “Death on the Streets: cars and the mythology of road safety” http://rdrf.org.uk/2011/06/death-on-the-streets-cars-and-the-mythology-of-road-safety/ .
    2. There is a lot of evidence to show SiN in evidence; In London over the last decade cyclist casualty (KSI) rates have fallen quite obviously. Is this because cyclists have become more careful? There has been a lot of Dutch type infrastructural engineering? No. the finger points to the effect of an increased cyclist presence, particularly where there was already a noticeable amount of cycling, in inner London. Taking the number of cyclist deaths involving HGVs , this stays approximately the same, with cyclist journeys in the areas where they have primarily occurred nearly trebling, and with an increase in road freight. Again, the finger points towards the effects of increased cyclist numbers, as part of campaigns to get lorry drivers to be behave better towards cyclists.
    3. Does this mean that increased cyclist numbers is enough? Of course not: who has ever said so? We need it to be part of a wider programme of reducing danger at source, namely inappropriate use of motor vehicles, which can include vehicle and highway engineering and law enforcement as part of a cultural change seeing motor danger as the principal problem. Some people may indeed use SiN as an excuse – so do many use segregationsim as an excuse for not cycling or supporting cycling.
    4. Take a look at “The non-linearity of risk and the promotion of environmentally sustainable transport”, Rune Elvik, Accident Analysis and Prevention 41 (2009) 849–855
    5. “But if the only reason we want more people to cycle is to diminish the relative risk faced by current cyclists, then frankly I don’t think we should really care about whether they are put off. Because that’s simply not the right approach, both morally, and practically.” A lot of people want to cycle, and giving them confidence to do so is fine by me: it gives them a load of health and other benefits etc., quite apart from reducing overall casualty rates among all cyclists (A qualification: SiN is a general tendency and seems to happen with a significant cyclist presence in the first place, as well as being assisted by lower speeds and probably other forms of reduced motor danger).
    6. A lot of us are fed up with being “dangerised”: I have spent thirty years being told that I am, in effect, asking for trouble by cycling. While I have been hurt (outside London), and while there is the monster of unaccountable road danger present, I still don’t like the dangerisation, which slots in all too neatly with the relentless victim blaming , pressure to wear helmets, hi-viz etc.
    7. Accepting SiN is only the first step: reduced cyclist casualty rates (per journey/distance travelled/type spent cycling) is a better measure of cyclist safety than aggregated cyclist casualties’. But even then there are better indicators: some cyclist KSIs are not the result of motor danger, and in my view are qualitatively different from those that are. But supporting SiN is part of the danger reduction approach we need, in my view.

    • michael says:

      “2. There is a lot of evidence to show SiN in evidence; In London over the last decade cyclist casualty (KSI) rates have fallen quite obviously”

      Do you have a link for the data that you get this from, please?

      I am open to be persuaded if the data is out there, but I am confused that what you state appears to be entirely inconsistent with, for example, the graph here:


    • michael says:

      Also my objection is more about moral philosophy than statistics, I am unconvinced that its moral to use people as human speed bumps. Cyclists are not there as a means to try and persaude motorists to behave themselves (the logic of things like choke points and ‘shared space’ seems to be precisely that, use cyclists – and pedestrians – to try and persaude drivers to slow down, and if they don’t, well, who cares, its only a dead cyclist, there are plenty more of them available).

      The other aspect you mention is far more to the point – infrastructure, pedestrianisation, and proper enforcement of the laws of the road. If you have that the numbers will follow (and, true, if the numbers follow that then means more political power and hence more infrastructure, but that’s about political power in numbers not road safety as such)

      My opinion is that safety-in-numbers is a side-issue and at best a bonus positive-feedback. I also think its morally questionable to make that a primary plank of the approach.

    • michael says:

      There’s also the data (I think for the whole country) shown in the graph here:


      So I am very confused by your claim that cycling KSI rate has become ‘obviously’ lower in London over the last decade. All the data I can find says the opposite. As you have written an apparently well-recieved book (that looks to be a serious work judging from Amazon preview!) I have to assume you have some real basis for saying this, but I don’t see how it can be squared with the data from the DoT.

      • Wyadvd says:

        Cannot speak for Dr Davis, however absolute Ksi data can only be made meaningful in my opinion if we have accurate figures for the kms travelled by bicycle in London. These figures are difficult to come by, although I would imagine surveys have been done from which this data could be deduced. We need ksis per billion km travelled not just absolute Ksi figures.

        • Wyadvd says:

          Otherwise it’s called a Penningism

        • michael says:

          But both the links I gave are relative to miles traveled, NOT absolute figures, so your objection is surely already answered?

          (Dr Davis on the ohther hand does seem to have a possibly-valid response to those conflicting figures, one which will require more thought to decide about!)

          • Wyadvd says:

            Yes my mistake sorry. The other question to ask is when viewed alongside a graph of these figures spanning a few decades, would the recent figures look like a blip? And are the changes greater than might be expected from normal randomness ? Showing 3 years figures might not be valid at all!

            • michael says:

              Well, yeah. If the supposed ‘trend’ can be reversed by simply varying the start date by a year or so, then that suggests to me that there likely isn’t a long enough time period there to conclude anything of substance either way. But in that case it can’t be used as evidence _for_ the SiN principle.

              • Wyadvd says:

                Well far be it for me to say , but the upward recent trend in London casualties has absolutely nothing to do with SIN , valid theory or not. It has everything to do with cyclists having been corralled onto blue painted lanes that come with the expectation that they shouldn’t be in a right hand lane if they want to pass straight on. The expectation is that they should be on the left with all the left turning lorries. Great idea boris. I still believe in the safety in numbers principle regardless. Dangerous road engineering is present in London like no where else on earth, and that’s what’s behind the deaths .

    • Thanks again for commenting Robert. I’ll take your points in order!

      1) I am aware of the long history of SIN. That’s mainly why I referred to the ‘modern history’ of the idea! Sorry if I wasn’t explicit enough.

      2) Yes, safety has improved while numbers have increased. However I would be very wary of making such a straightforward connection, particularly now that the casualty rate has started increasing again, despite the numbers of cyclists continuing to rise. That is, numbers are no longer being accompanied by more safety. This plainly suggests that there are plenty of other variables in play here. ‘Smoothing traffic flow’ could be reversing the safety gains, something that ‘numbers’ isn’t capable of offsetting.

      Bus lanes were also introduced in great numbers over the period in question. Likewise there have been medical advances (perhaps a decade ago Mary Bowers, who was run over by a lorry, would have been killed). The picture across London and the UK over the last few years is far from clear; this should also be apparent from the long-term relationship between relative safety and numbers in the UK since the 1950s.

      3) People are saying that Safety in Numbers is all we need. This has happened, and is happening. It’s an open goal for those in power who simply don’t want to make difficult decisions about making cycling safer. Others are placing far more emphasis on the concept, again at the expense of measures that have a proven record of improving safety, both relative and absolute.

      4) I have! It’s cited here.

      6) No disagreement here. Although I would suggest that what you are talking about is the increasing (unpleasant) tendency for cyclists to be more and more responsible for their own safety, with attendant pressure, as you say, to behave absolutely correctly, and to wear particular clothing and equipment. That is quite different from pointing out that the dangers cyclists face are needless and excessive.

      7) I disagree – SIN is a distraction. We should just focus on the measures that will actually make cyclists safer. Be it stiffer penalties for driving offences, reduction of motor vehicle speeds, and changes to the physical environment. Talking about boosting ‘numbers’ as a means of boosting safety is an easy way for those responsible to do nothing beyond mere ‘encouragement’. There’s also the moral problem (referred to by Michael) of using cyclists as a means of ensuring their safety. We should just focus on reducing the objective and subjective danger they face.

  15. Firstly, wyadvd. It should be obvious that I am saying that cyclist KSIs per journey/time /distance travelled is far superior to cyclist KSIs. This is a key demand from CTC.LCC,BCOG, RoadPeace and Sustrans in response to the TfL “Road Safety Action Plan” consultation document, which is still out to consultation until the end of the month. See http://rdrf.org.uk/2012/08/disgraceful-tfl%e2%80%99s-%e2%80%9ctowards-a-road-safety-action-plan-for-london-2020%e2%80%9d/

    Secondly, Michael, on the cyclist casualty rates. The latest collection of official statistics is here: http://www.tfl.gov.uk/assets/downloads/pedal-cyclist-collisions-and-casualities-in-greater-london-sep-2011.pdf Key pages are 6 and 7. SiN is shown by looking at:
    • KSI, which in effect means “Serious Injuries” as Killed area small proportion of these, and not “Slight Injuries”, which is a particularly unreliable measure.
    • From 2000 or so, when the cycling numbers moved up after the congestion charge was introduced.
    • Using the counts referred to in the report, which are cordon and TLRN counts. There are other counts done by Boroughs which some of us are trying to collate and systematise. The DfT London counts are notoriously unreliable.
    Other discussions, such as the one you refer to, do not do the above.(they use DfT indices, Slight Injuries, and go from a different date). Also, some “debates” refer to very short time periods.

    Let me summarise by saying:

    There are two different things here. One is the evidence about adaptive behaviour/risk compensation, which in the case of SiN is about motorists reacting to increased numbers of cyclists in certain situations (such where there is already a noticeable number of cyclists and typical inner city speeds). I think there is clear evidence here. It is related to discussions about worse motorist behaviour when there is reduced perceived risk for motorists, whether because of more crashworthy cars or a similarly idiot-proofed highway environment. Or one of the reasons for the absence of evidence for the effects of bicycle crash helmets. Or the long term patterns of road deaths as described in analyses of the Smeed curve. All of this is central to discussions about safety on the road.

    The other flows from this, and it is about what we want. You can call this the morality question. It is the key issue.

    As far as I am concerned, even if the casualty rate among cyclists falls, that still doesn’t necessarily meet my wishes. Even a well monitored casualty rate does not separate between those cases where, for example, an individual cyclist falls of her bike because she falls asleep, and one where a third party is responsible for a collision she cannot be expected to avoid. I am interested in the latter; the former may be irresponsible but is, as far as I’m concerned, a quite secondary issue. I want the danger to her to be reduced, and since we are always going to have some danger about, those responsible for it to be held accountable.

    There is no contradiction between me wanting to reduce road danger (which includes making those responsible for it accountable) on the one hand; and riding a bicycle in current conditions and supporting those who want to in their legitimate desire, particularly when the risk is low and them doing so reduces overall cyclist risk (given the provisos about the precise locations and conditions we are in). There is a difference between a paradox (an apparent contradiction) and a contradiction.

    Briefly on some other points raised by you and As Easy:
    • I really don’t know anybody committed to cyclist well-being who just thinks getting more cyclists out there is the answer. I haven’t met them. Government is not, in my view, seriously interested in reducing danger to cyclists – not is it really interested in getting significantly more people to ride bikes.

    • Thanks for saying my book was “well-received”. I got glowing reviews from people who support cycling and walking and sustainable transport: the “road safety”(RS) lobby either misrepresented it or ignored it. Which, since I think the RS lobby is an integral part of the problem of danger on the road, is hardly surprising.

    • As Easy keeps on talking about cycling becoming, or not becoming, “safer”. It is important to remember, in my view, that – unlike what the Daily Mail and similar suggest – cyclists pose little threat to other road users. (At least substantially less than the motorised do). I hope we can agree that cycling, in that sense, is pretty safe.

    I think it important to use “danger” in the transitive, as opposed to intransitive sense. Otherwise you miss out on the fact that danger (principally) comes from the motorised. Indeed the confusion between transitive and intransitive is a key component of RS ideology, and something to be wary of.

    Finally, I just do happen to think that an effect of having more cyclists about is to reduce some of the danger to all cyclists, and that it is part of a general road danger programme – so I disagree.

    Beyond this I am just going to repeat myself.

    • michael says:

      Thank you for a temperate and informed response.

      I still instincitvely react against the whole notion of using people as a means to persaude other people to behave better, with its desperate hope that the latter will choose to be nice rather than risk the life of the former. I just dislike the whole idea (something I feel everytime I approach a choke-point with cycle markings in the middle of the road apparently urging me to bravely ‘take the lane’, when there is an impatient motorist behind me clearly itching to overtake).

      As for the stats, there are clearly technical issues that you may well understand better than I do, so I will either have to take your word for it or do more work myself.

      But I would make the following two points

      1 the ‘go from a different date’ point at least is a very dubious one, surely, because if you can change the percieved ‘trend’ by simply varying your start date by a year or so that to me suggests you are not looking at a sufficiently long-term trend in the first place so as to say anything meaningful. I realise this is only a minor part of your response.

      2 The numbers of cyclists in London is still extremely low, so any supposed effect is only showing up in terms of very small numbers, and thus doesn’t seem very strong evidence – what happens for very small numbers doesn’t necessarily scale up. it could be hugely affected by exactly _who_ it is who is clocking up the increased milage, for example. If its a case of experienced riders riding longer distances, that could easily lead to a lower KSI rate even without any safety in numbers effect.

  16. Pingback: An expanded CicLAvia goes off without a hitch; a killer driver walks with a virtual pat on the back « BikingInLA

  17. John Steedman says:

    I’ve always though that the “smarter choices” policies are really aimed at growing the cyclist constituency / political pressure.

  18. Pingback: Would you spur swimmers to the beaches of Amity Island? | streets.mn

  19. Pingback: Is it really 20 times more dangerous? | Arizona Bike Law

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.