Safety in Numbers

Summary

This page draws on published evidence about safety in numbers, and some thoughts of my own, after many years of research into factors affecting risk in cycling.

There is strong evidence of an association between numbers of cyclists and reduced risk:  there is a “safety in numbers” effect.

It is not known whether greater numbers in and of themselves reduce risk, or whether the risks fall due to other causes (confounding factors) that are not being measured. Most probably “safety in numbers” is an amalgam of effects.

Possible confounding factors would be: a) new cyclists being mostly low-risk utility riders, who dilute the effect of high-risk sporting riders; and b) better traffic conditions encouraging more people to cycle in safety.

However, signal detection theory and supporting experimental evidence do indicate that a true “safety in numbers” effect is likely to exist where cycling is growing from low levels.

Introduction

A wide range of observations shows that when there is an increase in the number of cyclists, the number of cyclists injured in traffic collisions does not rise as much [1]. Therefore, the average risk per cyclist is reduced. It has also been observed that places where cycling is popular have lower casualty rates than places where cyclists are rare. This effect of less risk with more cyclists is known as Safety in Numbers, or SiN for short.

(Photo by Pawan Sharma/Unsplash)

These findings have been used to reinforce other known benefits of encouraging more people to cycle.

However, cautionary voices [2] have warned that association is not the same as causality. Any change in cycling numbers will take place in a dynamic environment with many varying factors. In the absence of a clearly demonstrated mechanism for SiN, it is not appropriate to assume that just getting more people out on bikes will, by some “invisible hand”, automatically bring safer cycling.

Is There Actually a Safety In Numbers Effect?

The most comprehensive evaluation of evidence on SiN to date is a meta-analysis [3] of studies published between 1979 and 2014. This investigation concluded, on the basis of highly consistent results, that “a safety-in-numbers effect exists, but it is not clear whether this effect is causal, nor, if causal, which mechanisms generate the effect”.

The SiN effect is expressed mathematically in the following way: let us say that the number of cyclists doubles, but the number of injured cyclists rises by only 50%. In that case a relationship between numbers and injuries can be stated like this:

20.6 = c. 1.5

That is, there is a “power law” relationship between the increase in cyclists and the increase in injuries. In this case, the power law has an “index” of 0.6.

In practice, different studies report different powers of SiN.

The meta-analysis provided a power of 0.43 for cyclists and 0.51 for pedestrians. For motor vehicle traffic the power factor was 0.5. This means that, provided there is no change in motor traffic, a doubling of cyclist numbers would be followed by only 35% more injuries. The average risk of injury per cyclist would fall by about one third. If cycling doubled, and motor traffic fell by 20%, then the risk per cyclist would fall by 40%.

The debate surrounding SiN is not whether it exists: there are no sensible grounds for disputing an association between more cyclists and less risk per cyclist. The debate is whether risk falls because there are more cyclists, or whether there are more cyclists because risk falls for other reasons (i.e. confounding factors). The distinction is important. If SiN is to be a policy tool, the actual mechanism by which it works has to be understood. Otherwise, well-intentioned policies could yield negative outcomes.

The most likely confounding factors are:

Recruitment Bias

In places where cycling is uncommon, such as the USA, where just 0.5% of trips are by bicycle, a large proportion of cyclists are sporting riders on expensive racing bikes. They are comparatively traffic-tolerant. In contrast, cycling for utility is rare in most of the USA, with commuting to work or school amounting to only 15% of bicycle trips, compared to 35% in Denmark (see Chapter 1 of City Cycling [4] for further details).

(Photo by Angelo Pantazis/Unsplash)

 

The risks of sport riding versus utility riding have not been well studied. However, one statistic from the Netherlands suggests that the difference is large. Dutch state data in 2008 reported that 13.3% of hospitalised cyclists were wearing a helmet when injured [5], whereas helmet use in the Netherlands is negligible, somewhat less than 1%. Those who use helmets are almost exclusively mountain bikers or sporting riders.

(Photo Gary Neesam/Unsplash)

The inference is that sporting riders faced substantially greater risks than the mass of the Dutch population using their bikes for transport.

(Photo by StockSnap/Pixabay)

If growth in cycling were mostly due to more utility riding, then the low-risk commuters and shoppers would gradually dilute the higher risk sport riders, leading to an apparent SiN effect, despite there being no actual reduction in risk per cyclist.

Road Conditions & Infrastructure

Traffic speed has a significant effect on risk in cycling. Data collected by the British government in the course of its National Road Traffic Survey (a continuous program of counting traffic at thousands of locations in the British road network), reveal how strong the effect is.

Chart 1: Risk in Cycling (2011/12) on Different Types of British Road

DfT Risk by Road Class 2011 2012

Note: A-road = main or trunk road; B & C road = lesser road, lane or backstreet.

Chart based on data from Chart 30018 in Reported Road Casualties Great Britain 2012, Department for Transport (issued in 2013). 

This interesting chart shows that cycling in Britain on quiet roads with low traffic speeds bears about the same risk as cycling in the Netherlands or Denmark. That is, where British cyclists enjoy the same conditions as are typical of favourably engineered networks in NL and DK, they face the same low risks.

An implication concerning SiN may be drawn from this chart. Where cycling is not popular, there are few, if any concessions in road design for cyclists. Riders are often forced to use busy roads, where risks can be high. This is particularly true for sporting riders going cross-country. As already noted, sporting riders are traffic-tolerant compared to those who ride only for local trips to work or to the shops.

Growth in cycling tends to happen after reductions in traffic, or reduced traffic speeds, or the provision of alternatives to busy roads. In this case, the fall in risk and the increase in cyclists could have a common cause: less threatening traffic conditions. As the above chart shows, reducing cyclists’ exposure to high speed traffic will have a material effect on average risk.

True Safety in Numbers?

Although it is likely that the above two confounding factors contribute to an apparent SiN effect, there is also evidence for a “true” SiN effect. The evidence suggests that true SiN would be especially powerful in places where cyclist numbers are rising from very low levels.

(Photo by Simon Rae/Unsplash)

A recent paper [6] proposed that “signal detection theory” [7] is a plausible mechanism for a true SiN effect. This theorises that probability of detection (e.g. a driver seeing a cyclist) depends upon:

(1) how clearly the target can be detected

(2) the observer’s relative frequency of experiencing the target

(3) the consequences of detection.

Experimental results show that rare targets are often missed [8], and very rare targets are highly likely to be missed [9]. In a driver simulation study [10], motorcyclists were detected significantly earlier when they were commoner, providing the driver additional reaction time to avoid collisions.

Another important factor is “inattentional blindness”: the degree to which unexpected objects or events are noticed at all. Inattentional blindness is especially likely in stressful situations. In one experiment, observers counted passes of a basketball among a handful of people, while a human in a gorilla suit walked through the game. Only half of the observers noticed the gorilla at all [11] (Note: there is also a Youtube video of this experiment, called “Selective Attention Test”).

The accumulation of experimental evidence on detection makes a strong case for there being a “true” SiN effect where cycling is growing from low levels of use.

Conclusions

There is strong evidence of an association between numbers of cyclists and reduced risk. That is, there is a Safety in Numbers effect.

It is not known whether greater numbers in and of themselves reduce risk, or whether the risks fall due to other causes (confounding factors) that are not being measured.

Possible confounding factors would be: a) new cyclists being mostly low-risk utility riders, who dilute the effect of high-risk sporting riders; and b) better traffic conditions encouraging more people to cycle in safety.

However, signal detection theory and supporting experimental evidence do indicate that a “true” SiN effect is likely to exist where cycling is growing from low levels.

One important benefit of SiN has been to correct official assumptions that cyclist casualties would increase in direct proportion to the number of cyclists. In the 1990’s, this view was a major obstacle to persuading the UK government to adopt the promotion of cycling as a policy target. It still remains an attitude commonly encountered in local authorities. At the time of writing (April 2018), positive trends in cycling remain the exception in the United Kingdom.

Another relevance of SiN has been to caution against measures that reduce the popularity of cycling, such as enforced helmet laws.

On the other hand, once the case for cycle promotion is accepted, SiN has limited application. Programmes to promote cycling are multi-faceted, employing a wide range of measures. This means there are too many changing variables to draw conclusions about whether SiN has been influential or not. Other measures to improve safety are well demonstrated, most notably lower traffic speeds and separate cycle routes along arterial roads. Proper enforcement of traffic laws, adequate sentencing of driving offenders, and strict liability legislation are also positive steps (none of which have yet come to pass in the UK).

References

 [1] Jacobsen P. Safety in Numbers: more walkers and bicyclists, safer walking and bicycling. Injury Prevention 2003;9:205-209

[2] Bhatia R, Weir M. “Safety in Numbers” re-examined: Can we make valid or practical inferences from available evidence? Accident Analysis and Prevention 43 (2011) 235–240

[3] Elvik R, Bjornskau T. Safety-in-numbers: A systematic review and meta-analysis of evidence. Safety Science, 2015. http://dx.doi.org/10.1016/j.ssci.2015.07.017

[4] Pucher J, Buehler R editors. City Cycling. The MIT Press. 2012.

[5]  Ormel W, Wolt KK, den Hertog P, .Enkelvoudige fietsongevallen. Ministerie van Verkeer en Waterstaat, 2008.

[6] Jacobsen P, Ragland D, Komanoff C. Safety in Numbers for walkers and bicyclists: exploring the mechanisms. Injury Prevention 2015;21(4):217-220

[7] Nevin JA. Signal detection theory and operant behavior. J Exp Anal Behav 1969;12:475–80.

[8] Wolfe JM, Horowitz TS, Kenner NM. Rare items often missed in visual searches. Nature 2005;435:439–40.

[9] Mitroff SR, Biggs AT. The ultra-rare-item effect visual search for exceedingly rare items is highly susceptible to error. Psychol Sci 2014;25:284–9.

[10] Beanland V, Lenné MG, Underwood G. Safety in numbers: Target prevalence affects the detection of vehicles during simulated driving. Atten Percept Psychophys 2014;76:805–13

[11] Simons DJ, Chabris CF. Gorillas in our midst:sustained inattentional blindness for dynamic events. Perception 1999;28:1059–74.