Saturday, 31 December 2011

It's not just me, SWOV also thinks "safety in numbers" is untrue

The phrase "safety in numbers", as applied to cyclists, comes up a lot on this blog. It was discussed in this post and this one and this one. It's discussed here a lot because it has been a core belief of many UK cycle activists, and in particularly, organisationally, of the CTC, who never loose an opportunity to promote the idea as basically the "solution" to cycle safety in the UK. Get lots of cyclists on the roads, they say, or imply, and it will all be OK, very safe, without needing to change anything else fundamentally in the road environment.

I must acknowledge some change of tone from the CTC of late, for example, in this letter to the The Guardian written by their Campaigns and Policy Director Roger Geffen, in response to the Guardian article that I also criticised in my last post. Here Geffen does mention the role of road design in creating safe conditions for cycling, while (as usual) stopping short of calling for proper continental-type cycle infrastructure. But in general, CTC has always espoused belief in the "safety in numbers" effect as being key to generating a bigger, safer cycling culture. From this seems to flows their concentration on promoting training for cyclists (though they first started to promote training in the 1930s, long before "safety in numbers" had been heard of). If you train cyclists to cope with our road conditions, you get more of them on the roads, so the argument goes, so it becomes safer, and that encourages more cyclists out, and so on. And from the CTC's concentration on training comes the government's concentration on it. Whenever the government is asked what it is doing about cycling, basically what it talks about is training for cyclists.

There is a counter-view, that I espouse, as have others, particularly Freewheeler, and David Hembrow. This is to look at the post-war history of cycling in the UK, and to note that we have had 60 years of training for cyclists, corresponding to an ongoing collapse in cycling. In our view, training simply is not the solution. The international experience shows you do not get significantly more people on bikes without major changes to the road environment. Cycling in the UK fell from its immediate post-war modal share of about 15% to about 1% today, despite 60 years of training, for reasons that are not going away. You can encourage and train and enthuse as much as you like, but those reasons continue to hold our cycling rate to a very low level, even in the cities that are currently said to be experiencing a "cycling booom" such as London, where cycling is now (2010 figures) up to the dizzy height of 2% modal share.

The modern vogue of the "safety in numbers" theory for cycling, in the English-speaking world (and it does seem to be largely confined there), seems to be mostly down to a rather vague piece of research published by an American, P L Jacobsen in 2003, in the journal Injury Prevention. Here Jacobsen stated his conclusion:
The likelihood that a given person walking or bicycling will be struck by a motorist varies inversely with the amount of walking or bicycling. This pattern is consistent across communities of varying size, from specific intersections to cities and countries, and across time periods.
I describe Jacobsen's research as "vague" because (and you can read the full paper if you like), though it is full of statistics, it lacks sufficient discussion of the radically differing contexts of the figures from different places and times to make its conclusions meaningful. It has no proper discussion of cycle infrastructure, despite the fact that it uses statistics from the Netherlands, where the effects of segregated cycle infrastructure surely cannot be ignored. It says:
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.
without going into the details of these policies further. But it is these details that are critical.

It also says:
The possible explanations [for changes in cyclist casualty rates] are changes in human behavior, roadway design, laws, and social mores. However, insofar as the changes seen in the time series data occurred rapidly and with both increasing and decreasing amounts of bicycling, 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.
Here is a key unsubstantiated assertion: why is it improbable that changed roadway design could lead to rapid safety change? Surely changing roadway design is by far the fastest method of changing both behaviour and social mores and, consequentially, safety.

In Jacobsen's defence, it may be noted that his final conclusion is simply to say that:
Policies that increase walking and bicycling appear to be an effective route to improving the safety of people walking and bicycling.
And it is fair to say that that may be how it "appears" if one takes a rather blinkered, infrastructure-excluding view of the problem. And he does not say what those policies should be, he does not specify that they should be training or publicity as opposed to road redesign. But unfortunately this work has been often used as evidence for an idea that the good safety record of the Netherlands and Denmark is not due to their well-designed infrastructure, it is due to their having large numbers of cyclists.

Clearly Jacobsen only demonstrated a statistical link, not causality. His conclusion is open to the obvious objection that the causality may be (in fact, seems, on a wider view of the contextual evidence, and also on a common-sense view, far more likely to be) the other way round, that larger numbers cycling are due to higher perceived safety, which is linked to real safety, which is linked principally to road design (and users' behaviour as influenced by road design).

Anyway, all this is a just pre-amble to tell you that the Director of SWOV, the Dutch institute for road safety research, agrees with me on this. His discussion on the subject is contained in a column on page 3 of the Autumn 2011 Research Activities published by SWOV in English. I would rate Fred Wegman, Director-manager of SWOV, as the world's greatest expert on cycle safety, on the principle that you listen to those who have achieved rather than talked about a goal. I quote his article in full:
Safety in numbers: more cyclists, lower risks?
A swarm of birds or insects, a school of fish, a flock of sheep, a pack of wolves. Language has many possibilities to indicate a group of animals. Biology has taught us that a group offers protection and increases safety for the individual members of the group. It is more difficult for a predator to attack a group than one single animal. For this reason a predator uses the strategy to isolate an animal from its flock before attacking it.

This is also the thought that occurred to me when studying cyclist safety: why not compare the safety of one individual cyclist with the safety of a group of cyclists. The risk of the one single cyclists is greater than that of a cyclist who is part of a group. The American
researcher Jacobsen compared the casualties among cyclists in different countries and
tried to establish a relation with the amount of bicycle traffic. He concluded that “Policies
that increase the numbers of people bicycling appear to be an effective route to improving
the safety of people bicycling”. In other words: add more cyclists to traffic and cyclist
safety will increase. This is a popular idea among those who put effort into stimulating cycling and therefore Jacobsen’s conclusion is quoted frequently in these circles. However,
I believe that this conclusion is not correct. I will try to explain why.

If there is much cycling in a country, the risk for cyclists is indeed lower. Comparison of
statistics of different countries offers conclusive evidence. The risks in countries that have
a lot of cycling like the Netherlands and Denmark are (much) lower than in countries where cycling is a less important mode of transport. The explanation may be twofold. Firstly,
there are the expectations of the other road user. If a driver does indeed expect a cyclist
on the road, as is the case in the Netherlands and Denmark, the risk is lower. But a second
explanation is conceivable: if there are more cyclists, more safe cycling facilities will be
constructed (which in turn make cycling more pleasant). We have sufficient evidence that
cycling facilities (like bicycle tracks) reduce the risks of cycling. Not only do the Netherlands and Denmark have many cyclists, there are also many cycling facilities. 
I do not expect that just a greater number of cyclists will on its own result in a risk reduction for the cyclist. On the other hand, I do expect that more cycling facilities will lead to lower risks. Policy that only focuses on an increase in cycling and at the same time ignores the construction of more cycling facilities, will not have a positive effect on road safety. Unless, of course this policy also takes care of cyclists only cycling close to one another: in a swarm, school, flock, or pack of cyclists.

This is the clearest statement I have seen in print that Dutch road safety experts do not agree with the simple "safety in numbers" idea, despite the fact that Jacobsen's work relied heavily on Dutch statistics. Wegman is very clear:
We have sufficient evidence that cycling facilities (like bicycle tracks) reduce the risks of cycling. Not only do the Netherlands and Denmark have many cyclists, there are also many cycling facilities.
And as I have often pointed out, the real clincher to the argument is that there are no counter-examples. There is nowhere that has achieved good cyclist safety and mass cycling without these facilities (like bicycle tracks).

Elsewhere in this interesting publication, we read SWOV's opinion on the correct way of dealing with the lorry danger problem, source of so much grief in the UK:
In SWOV’s opinion, the ultimate solution for the blind spot problem is a structural separation of trucks and cyclists. How this must be organised and what the economic consequences will be, requires further study. For the time being, the solution can be found in separating cyclists and trucks at intersection[s], both in time and position.
On Trixi mirrors mounted on signal posts, the devices that are being promoted in London at the moment as a method of addressing the lorry blind-spot problem for cyclists, SWOV has this to say:
This mirror has been found to barely influence truck driver behaviour and is only effective while the truck is stopped in front of the mirror.
The overwhelming impression from the whole document is of the extraordinary seriousness with which the Dutch take cycle safety, and how, far from being satisfied with already having the safest cycling and highest cycling levels in the world, they are continually spending money researching, applying and evaluating better solutions. It is clearly this attitude that lies behind their cycling achievement. "Safety in numbers" is a too-simple concept that in no way summarises the complexity of the problem and the interlocking effects which occur that influence cycle safety.

Yes, if you are talking about the flock or the herd, there is safety in numbers. I was in the Vismarkt, Groningen city centre in September, on a study tour led by David Hembrow. I was sitting there at a café table with David and the others from the Cycling Embassy of Great Britain, looking at the scene, which was this (video by Stabiliser):

Yes there are huge numbers of cyclists here, "flocks" of them, and it feels very safe for them. But why are there such huge numbers here? It is because urban planning over the whole city has made it difficult to drive here and very easy to cycle. There are no separate cycle facilities here, but that is because they are not needed here. If you look at the rest of the city, as we did, you see the massive network of segregated cycle tracks and traffic-free paths that make it both subjectively and objectively safe to cycle from the suburbs and surrounding villages to the centre, whether there are many or few cyclists cycling around you, resulting in this massive concentration of cyclists in the centre, where all the routes converge, where there is no longer need of segregation. A person parachuting in from Mars, or London, to this spot, might say "Cycle facilities, what cycle facilities? No need, as it is so safe, as the place is full of cyclists". But they would fundamentally misunderstand the context. Similarly, Jacobsen's simplistic comparison of cycling safety with cycling levels critically overlooked context.

Segregated cycle track on a main road leading to Groningen city centre
Happy new year. Let us hope 2012 will be a safer year for cycling in London than was 2011.


  1. The sentence you've pulled from Wegman's argument misses out on a significant thrust of it, that it is increasing numbers of cyclists that provide the political impetus for increasing investment in infrastructure.

    While I would fully agree with you that the high levels of cycling in the Netherlands are due to the excellent infrastructure, I think that to argue that the direction of causality in the numbers v safety argument must be reversed is also missing something. I would argue that this thinking is flawed, in assuming that the causality is uni-directional even over time. I would suggest that actually causality here is bi-directional, with an increase in infrastructure & perceived safety helping to cause increases in numbers, and increases in numbers causing more spending on infrastructure & driver awareness & better perceived safety, which causes increases in numbers, etc.

    Thus the 2 factors have, I would feel, interlinked, mutually re-inforcing causality. To achieve results *both* aspects - infrastructure & numbers - must be advocated for and promoted. Only if both sides are pushed can the 2 together enter into that virtuous spiral of increasing safety & numbers which the Netherlands saw.

  2. I agree that causality must go both ways to an extent. It is clearly the case that the effect that Wegman talks about, the large numbers of cyclists creating the political pressure for better facilities, is very strong in the Netherlands. This is manifested in the fact there that large expenditure on cycling is in no way politically controversial – like large expenditure on the NHS in the UK, it is seen as benefiting everybody. You also see, furthermore, the pressure for the character of the cycle provision coming from groups that would have no influence on cycling provision in the UK, because of the social breadth of cycling there. See for example this piece by David Hembrow where grumpy pensioners criticise the steepness of a cycle bridge as being like "the Alps". This kind of thing drives the quality of provision there.

    We can't get anything like that pressure in the UK from the base we are at now of 1-2%, so we are more dependent on political leadership, which is in short supply. I don't totally discount the effect of promoting cycling through advertising, but I think that it will only have a real effect if it is simultaneous with obvious improvements in conditions for cycling. You can't do what TfL does, which is do the promotion, then say "It's still not worth providing decent cycle lanes because there still aren't enough cyclists." It requires more of a leap of political faith, which the Dutch made 40 years ago.

  3. David Arditti: Bear in mind that cycling rates in the Netherlands never collapsed to the pitiful levels in the UK (& Ireland - another country I'm interested in). I don't have the figures to hand, but I have a vague recollection from one of Hembrow's or Wagenbuur's posts that at its worst, dutch cycling levels were still order of 25% participation. Thus the dutch re-focusing then did not require quite the amount of political courage as it does in the UK today.

  4. Training should always be at the bottom of any safety pyramid.
    This is the order in which you should proceed.
    1. Assess the risk - look at the infrastructure we have and see what the risks are.
    2. Avoid the Risk - by redesigning the infrastructure.
    If you cannot avoid the risk completely.
    3. Minimise the risk - by doing what you can to reduce it
    and last of all...
    4. Warn about the risk and provide training.
    But too many people jump straight to 4 thinking that you cannot do anything other than 4.