Safety in numbers: if and when so, why?

I write here in response to an online article in the Grist blog, which addresses the concept of “safety in numbers” among cyclists and pedestrians. As is all too common, the article takes this phenomenon for granted, and ascribes it entirely to changes in behavior of motorists. There is no mention of changes in behavior of cyclists and pedestrians themselves.

The concept of “safety in numbers” is often applied to animal behavior, and a couple of examples might be useful. Here’s one from the University of Rhode Island Sea Grant program:

A potential predator hunting for a meal might become confused by the closely spaced school, which can give the impression of one vast and frightening fish. Additionally, there is the concept of “safety in numbers”—a predator cannot consume an unlimited quantity of prey. The sheer number of fish in a school allows species to hide behind each other, thus confusing a predator by the alteration of shapes and colors presented as the school swims along. Of course, those on the outside edges of the school are more likely to be eaten than those in the center.

Here’s a video of such behavior.

A herd of animals may actively defend its members, as shown in this astonishing video.

In warfare also, there are issues of safety in numbers, discussed in terms of “swarm warfare“. In some cases, the opposite tactic is applied, for example, in dispersing troops to avoid their all being taken out in a concentrated attack.

Bicyclists and other road users, on the other hand, don’t generally interact like prey and predators, or like armies in battle. I mean, motorists may sometimes be careless. A few may intentionally take risks with the safety of bicyclists and pedestrians, or be in denial about the hazards they pose. Motorists, however, generally aren’t intentionally out to kill and aren’t contemplating having us for dinner. What, then, is the mechanism of so-called “safety in numbers”? When might it exist, and when not?

First, there are different kinds of numbers.

  • The general level of bicycling or walking in a population.
  • Different locations in the same area that have different numbers of bicyclists/pedestrians/motorists, but all from the same general population.
  • Different areas, with different populations and different rates of use.
  • Numbers which increase on particular days, for example in nice weather.
  • A population in the same area but changing over a span of years.

In an earlier post, I examined an Oakland, California study of pedestrian crashes at different intersections in the same community. The researchers’ text claims safety in numbers but the graphs show otherwise: the crash rate was moderately higher at intersections with more pedestrians. It was also higher at intersections with more motorists. It would take further analysis to determine how these two factors correlate with each other.

The well-known Jacobsen study of safety in numbers of bicyclists compared different communities at the same time. It also has been criticized (including in the Oakland study) for faulty math that shows a hyperbolic descending curve even if the input data are completely random.

In order to maximize safety, we need to know not only what happens, but also why. My own opinion is that there is generally an increase in safety over time as the number bicyclists and pedestrians increases. This occurs due to a number of different factors but also is impeded by others.

Some factors increasing safety:

  • Longer experience, on average, of cyclists
  • Cyclists’ better understanding of how to ride safely in a group
  • Higher average age of cyclists
  • Multiplicative effect of mentoring of cyclists by other cyclists (especially, children by parents)
  • Change in attitude of motorists
  • Presence of one cyclist directly increasing the safety of another by blocking traffic.
  • Better facilities design, resulting from a more refined understanding of how to provide for bicyclists’ safety.

Some factors decreasing safety:

  • Entry of new and inexperienced cyclists into the mix
  • One cyclist’s concealing another from view when there many
  • Sense of entitlement leading to scofflaw behavior
  • False sense of security — “follow-the-leader” behavior when riding in a group
  • Rush to build bicycle facilities, resulting in design compromises and inherent flaws
  • Facilities that favor the least skillful, but are overprotective and frustrating for the more skillful, promoting scofflaw behavior
  • Placing all eggs in the basket of “safety in numbers” and neglecting other approaches to increasing safety

There’s clearly much more research yet to be done in this field.

13 responses to “Safety in numbers: if and when so, why?

  1. I think the analogy with groups has some problems too, John. I’m sure in the cities Jacobsen used in his studies, the majority of cycling risk exposure in the “more cycling” cities still predominantly involved solo cyclists. I believe there IS merit to the concept that motorists get better at interacting with cyclists if they have more encounters with them. But I agree that there are a number of other factors and effects which are likely more important.

    Cycling groups rarely get involved in crashes with motorists. In my study of nearly a thousand long-form cyclist crash reports, only one involved a group (of three riders; motorist drive-out from a side street). (And of course cyclists often get injured on group rides in collisions with their fellow cyclists; often with injuries just as bad as in crashes with cars. But those don’t show up in police crash reports.)

  2. Good discussion, John. Thanks. I think all those variables are important to some degree, which is why I have my doubts that the studies so far are complete or compelling–even doing a principle component analysis on these variables, once hypothesized to be important, would require a lot of data and careful analysis.

    Fantastic video of the buffalo vs. lions. Wow.

  3. I wonder too if in some situations there is less variation in cyclist behavior (e.g. a tendency to mirror others or “follow the leader”) when the numbers are greater. This might also lead to more predictability and safety.

  4. “Higher average age of cyclists”

    I suspect this is a big component. Higher mode shares probably don’t come from getting more students on bikes. They come from getting more adults on bikes, especially older adults (beyond college age).

    These are the same people whose car insurance rates go down when they turn 25 and/or get married, because their behavior becomes less risky.

    Another possibility is cycling speed. As places with higher mode shares have built facilities, the average cycling speed on those facilities (due to design and/or congestion) is perhaps lower than a lone cyclist on an ordinary road. I wonder if there’s a correlation between higher cycling numbers and lower cycling speeds.

  5. I used to give the safety in numbers theory short shrift, but that presumes numbers high enough that motorists have an opportunity to witness and recognize safe operation as such. In many areas, most people believe cyclists should stay on the sidewalk and that is clearly not good for the cyclists that do operate safely.

  6. Steve, I never dismissed Peter’s hypothesis out of hand (but the scatter in the data and some other analytical flaws concerned me), but John hits the nail on the head. We have to analyze why numbers would increase safety, not take it at face value. Its not as simple as just pouring more cyclists onto the street, stirring, and saying “presto-chango” and suddenly cyclists are safe. If cyclists were few and far between but all were longtime “club” cyclists, I suspect there would be safety in scarcity. Like Forester, I would suspect one driver of crashes would be a short-frequency crash spike of inexperienced riders, followed by a drop off in crashes as the worst of these cyclists stopped riding and the rest gained skills.

  7. My criticism of the Jacobsen paper is not limited to the faulty graphical presentation (false advertising, I call it), but to other factors also. Jacobsen claims that the proportion of cyclists in the mix is the determining factor in the car-bike collision rate. Thought experiment: introduce many more cyclists into a city with low cyclist proportion and culture. Would the accident rate per cyclist diminish per Jacobsen’s formula? Probably not. The social and other factors that contribute to much cycling probably also contribute to a low car-bike collision rate. So far we have no means of measuring the strength of the individual factors.

  8. Predator / prey is relevant to understanding the behavior of cyclists and other users of the road because what we do, how we act, how we react, and what our reflexes save us from, are affected not only by our faculties of human reason and emotion, but also by the fruits of our evolution. Sure, we can act as high-functioning users of reason, economics, politics, and so on. We’re also animals with strong biological drives combined with specific, evolved traits which all affect how we interact in any situation, including in traffic, and particularly when there are perceived physical threats around.

  9. John Allen writes:

    >First, there are different kinds of numbers.
    > * General numbers of bicyclists in a population,
    >such that bicyclists are understood to be a presence on
    >the road.


    > Some factors increasing safety:
    > * Change in attitude of motorists
    – – –

    1. One way that “more cyclists” may change the behavior and attitudes of an increasing fraction of motorists may be that a given motorist is more likely to (a) be a cyclist, perhaps even a savvy cyclist, and/or (b) *personally* know or identify with someone — a relative, friend or coworker, or maybe a friend of one of them — that cycles, and who they can imagine being the cyclist they are passing or otherwise interacting with. This conjecture might be termed “reduced average degrees of separation”, e.g. if the motorist *is* a cyclist, then degrees=0; if they know and identify with no cyclists but someone “N” hops away in their social network does, then degrees=N+1. Hypothesis: in “cycling-hostile” cities the average degrees of separation will be substantially higher than in “cycling-friendly” cities, all other factors aside. BTW, I wouldn’t expect the “degree-reduction” to be uniform — it would depend on the topology of individuals’ and subgroups’ social networks.

    2) John Allen also writes:

    >Presence of one cyclist directly increasing the
    >safety of another by blocking traffic.

    My observations in the inner Bay Area lead me to believe this. Not only _actual_ safety, but also *perceived* safety, which may lead less-courageous riders to either begin riding (because they’ll feel safer with at least one other cyclist likely to be along their line of travel on a given block) or to ride in the street instead of on the sidewalk. I suspect that as the number of cyclists per block rises, motorist passing behavior changes because there’s no payoff for returning to the right after a pass. In other words, the (loose) stream of cyclists creates its own longitudinal space even without a bike lane, once the cyclist spacing distribution decreases sufficiently.

    John Ciccarelli

    P.S. Have you signed the LAB election petition yet?

  10. Thanks for your thoughtful response to the oft quoted “there’s safety in numbers” mantra for bicyclists on the roads. You provide some good material for pondering!

  11. I think predator/prey is mostly a red herring. The way traffic functions is not based on that. I’m afraid that even bringing it up will play into the fearful-and-vulnerable crowd’s mind set, and that just makes it harder to have an intelligent discussion.

    Regarding safe behavior: since, as we all know, two thirds of reportable crashes do not involve a collision with a moving motor vehicle, any significant reduction in the overall crash rate has nothing to do with how motorists perceive cyclists.

    In communities that haven’t yet been polluted by “bicycle friendly” nonsense, I find it easy to cycle and to interact with motorists. It’s easy here in Coopersburg, PA. It was easy when I was riding in small towns in Mississippi and Louisiana a few years ago. It’s been easy everywhere I’ve gone over the decades. I reject the idea that I would benefit from a swarm of inexperienced cyclists ‘training’ the motorists.

    “Safety in numbers” has been scientifically discredited, but it lives on because it’s a surrogate for the real issue. The real issue is the sociology of changing many people’s behavior. Like any other behavior change, cycling has its early adopters and its crowd followers. Advocates of greater mode share want more cyclists. This sounds like a good way to get them. I think its’ doomed to fail because it’s impersonal, treating cyclists as a crowd, not as individuals, and because it’s scientifically bogus.

    The hot air that is being expended on safety in numbers could, if repurposed, train one person on every block in the program. Were that to happen, as Mr. Ciccarelli points out, everyone would know someone who’s a cyclist. This would make it more alluring for crowd followers to begin cycling, and ensure that more of the role models weren’t clueless.

  12. One disadvantage to getting a significant number of poorly qualified riders out there is that indeed, everyone will know someone who is a cyclist. Not only that, but everyone will know someone who is a cyclist and who looks like they need some tutoring. So much for convincing the public to support our roadway rights.

    I hear it all the time. I suppose the optimist will follow by saying we should reach these people and help them improve. The problem is that as others have said before, many of them think they learned everything they need to know about riding their bike by the 4th Grade.

    The second disadvantage to the “Safety in Numbers” argument is that it could encourage new cyclists to be passive, thinking that there mere presence is going to do other cyclists a favor. While as John Allen and John Ciccarelli point out, the presence of more cyclists may increase the safety margin (see discussions above), a far more powerful argument would posit that new cyclists should pair with experienced role models as per Mr. Schubert, so that newbies are not only under destroyer escort while they learn, but that they learn good behavior along the way.

    Thanks to all who signed the petition (today is the drop dead date, so you can still sign). I don’t think we are going to get the 5%, which I suspect is the LAB’s plan, but we certainly have demonstrated that a significant number of the membership is concerned with the elections. I invite all who are interested in helping formulate what will undoubtedly have to be “Plan B”. I’ll probably think of something to post on my own site later this week to see if its worth a discussion.

    Khal Spencer

  13. The premise behind “safety in numbers,” it seems to me, is that if only we get more people riding bikes–by whatever means–then they’ll be safer. The fallacy here is that if these folks have not been educated in vehicular cycling, they are likely to do what uneducated cyclists always do–run stop signs, run red lights, ride on the wrong side of the street, not look before they enter a street from a driveway, ride on the sidewalk, not give signals. That behavior does not create a safe cycling environment, nor does it incline motorists to treat us as responsible vehicle operators–but rather as children. And yes, that is childish behavior.

Leave a Reply

Your email address will not be published. Required fields are marked *

Please answer this to show that you are a human!... *