Is it really 20 times more dangerous?

I’ve been seeing, more often lately it seems, fairly bold numerical statements about relative bike traffic safety; so to take this one from peopleforbikes for example (my emphasis added): “The problem is particularly glaring here in the United States, where bike injury and fatality rates are roughly 20 times those of the cycling-friendly countries of western Europe“.

These claims virtually always revolve around the Netherlands, with their renowned bicycle infrastructure; or perhaps Denmark (Copenhagen in particular).  One would imagine from the 20X factor that a cycling death there would be a practically-unheard of event in these places. But as usual there’s much more to the story. The statistics over which there is very little room for dispute or interpretation are as follows (I chose 2004 and 5 because of an article referred to later; the trend in later years has been down a bit, but it doesn’t affect the basic conclusions):

	2004	2005	2010 pop.	death/million pop.
Denmark	53	41	5,568,854	8.4
UK	136	152	62,041,708	2.3
Netherlands	157	151	16,696,700	9.2
Germany	475	575	81,757,600	6.4
US	727	786	309,300,000	2.4

As you can see, relative to the population size (per capita), cycling deaths in the Netherlands are far more common than in the US or UK. What everyone really wants to know, of course, is relative risk — John Pucher says in Making Cycling Irresistible: Lessons from The Netherlands, Denmark and Germany that in the US the average person cycles 0.1km/day, while in the Netherlands it’s 2.5 (figure 2); i.e. 25 more exposure, on average. This leads to Figure 10, with 2007 data, the takeaway of which is:

• Cyclists killed per 100 million km cycled; NL 1.1, US 5.8 (so US is 5.3X worse)
• Cyclists injured per 10 million km cycled; NL 1.4, US 37.5 (so US is 27X worse)

This is all well and good, assuming the risk exposure is correct, so I suppose it would be reasonable to claim something like the US fatality rate is about 5 times riskier than the Netherlands, or somesuch. The injuries, however, sound out-of-whack.

Moving on, the next part about injuries is really confusing; and possibly where the talk about “20 times safer” comes from (?? the OECD 2007 referenced in the article is linked below)… According to Figure 10, the injury rate in the US is twenty seven times greater than the Netherlands (1.4 versus 37.5 cyclists injured per 10million Km traveled), which he further states is “vastly understated”.

The serious injury to fatal ratio mentioned in Pucher’s article that the ratio tends to be about 10:1. I didn’t follow where their stats for the US came from. I have found (via a database derived from police reports) that the ratio in Arizona for cyclists is coincidentally almost exactly 10:1 (the query is below, it returns 879:85 for the period 2009-2012, the period for which i have detailed data. Serious injury is called INCAPACITATING_INJURY, versus, say, possible injury, or just injury). So I’m either not seeing, or misunderstanding any basis for the 20X claims. In other words, using Pucher’s units and benchmarks for risk exposure, it appears to me the serious injury rate is also somewhere in the neighborhood of 4 to 5X relative to the Netherlands; not 20x, not 27X and certainly not “vastly” more than 27X.

But, why is traffic in the US more dangerous? It probably has mostly to do with far higher MV speeds on surface streets; and this is interlocked with very different land-use patterns, including far far higher population densities in the Netherlands and Copenhagen, even compared to such biking model-cities like Portland, OR.

For other, similar-sounding, overstated claims, see, e.g. cycle-tracks-are-nine-times-safer-than-roads. 

References:

European fatalities by country: European Road Safety Observatory: Traffic Safety Basic Facts 2010 covers data from 1999-2008

US fatal (and injuries): NHTSA  Traffic Safety Facts: Bicyclists and Other Cyclists 2010 (covers data 2001-2010; there are versions of this document dating back to the 1970s)

EU population “est. July 2010” http://en.wikipedia.org/wiki/Demographics_of_Europe; US 2010 population http://en.wikipedia.org/wiki/Demographics_of_the_United_States retrieved 10/6/2013

Type of injury: SELECT eInjurySeverity, count(*) FROM 2010_incident i WHERE EXISTS (SELECT 1 FROM 2010_unit u WHERE u.IncidentID=i.IncidentID AND u.eUnitType IN (‘PEDALCYCLIST’)) GROUP BY 1;

This 2007 report put out by OECD is what Pucher refers to for the high number of injuries in US… I haven’t read it yet: Underreporting of Road Traffic Casualties (ITRAD Special Report. International Traffic Safety Data and Analysis Group)

see other articles tagged risk.  

OECD 2007 Report

So I started reading through the Underreporting of Road Traffic Casualties report, that was the basis for the Pucher article’s injury stats… I don’t see that it involved US in any way in particular. This was interesting, though:

Fig 3. bar chart showing a couple of dozen country’s traffic fatality per 100,000 population, and oddly doesn’t state the year — US is just under 15, and the Netherlands is something under 5 (4.5?). Just guessing by looking at NHTSA Trafic Safety Facts that it was something like 2005 when US was 14.67 (atrociously high, by the way; the US has fallen dramatically since then, the latest reported year was 2011 at 10.39 ).

Anyway, you can see by noting the fatality rates in my table, above, that in the US, bicyclist fatalities contribute a marginal amount to that overall number;  so 0.24 of the 14.67. Versus the other extreme of the Netherlands where 0.92 of the 4.5 are bicyclists. In the US, bicycle safety is a small part of a very big problem of overall traffic safety; in the Netherlands, bike safety is a large (well, significant) part of a smaller problem.

This glaring result is, of course, by looking per capita.

Safety in Numbers ( #sin )

Safety in numbers: more walkers and bicyclists, safer walking and bicycling P L Jacobsen 2003
Inj Prev 2003;9:205-209 doi:10.1136/ip.9.3.205

This all got me thinking about “Safety in Numbers” (sometimes shortened to ‘SiN’, a phrase coined in this context in a 2003 article by PL Jacobson), which is the hypothesis that states that simply getting more peds and cyclists makes things relatively safer — that is as numbers go up, there will be a non-linear increase in injuries — the general numbers bandied about are a doubling (a 100% increase) of peds/cyclists would expect to see only a 30% increase in injuries. I’ve always been uncomfortable with this conclusion; a very good, level-headed, and well explained rebuttal here. It isn’t a dismissal of the numbers, but rather questioning of the causation.

There’s also a long-form treatment of cyclists as drivers vs. cyclists as peds vs. cyclists as edge-riders touching on ethics and SiN written by Mighk Wilson at china-cups-and-butterflies-options-and-ethics. Many excellent points; he cleverly suggests the mainstream ABC (experienced Adult, Beginner adult, Child) cyclist types really should be supplanted by those three new categories (cyclists as: drivers, edge riders, and peds), but no catchy 3 letter acroynms.

Discussion of Jacobsen in a 2010 article by John Allen; see commentary by Furth, Gutierrez and Forester.

Helpful brief explanation of the cause/effect troubles bob shanteau in cyclists are drivers.

Here’s a newer discussion based on a 2016 report mentioned here, and appeared in the bicycledriving discussion group; John Forester explains:

The safety in numbers has finally been calculated properly, despite the reference to Jacobsen’s false discovery of it. Jacobsen, and his followers, always calculated a ratio of ratios using only three variables. The mathematic form was A/B compared to B/C. Regardless of the variables used, even random numbers, this will always result in a declining value that looks like decreased risk of cycling with increased cycling modal share.The report quoted below has done this calculation properly using three independent variables: number of casualties, number of cyclists, length of bikeways. These do show a decreased risk of cyclist injury per cyclist as the number of cyclists, of time and of the length of bikeways increase. Therefore, any criticism must be directed at this discovery. The first caution is the standard one that correlation (which this is) does not demonstrate causation. That is, for those with a predilection for the safety of bikeways, these data do not demonstrate that bikeways make cycling safer. They might, or they might not. In order to progress from the shown correlation we must have a causal relationship that provides the basis for moving from correlation to causation. There is no such causal relationship; bikeways are designed without regard to reducing car-bike collisions, and, in fact, in ways most likely to increase car-bike collisions. The authors of the report are probably aware of this criticism, because they appear to avoid making any claim that bikeways make cycling significantly safer. Instead, the burden of their argument is that bikeways attract more cyclists and that is what decreases risk. But there could be other factors equally strong. We are in a time of increased use of bicycle transportation, and of increased social concern about bicycling as time progresses. With the passage of time, cyclists are probably getting more skilled; with increased social concern about bicycling society, both cyclists and, particularly, motorists, are probably paying more attention to avoiding car-bike collisions. There may be other significant causes also. Since the data shown in this report say absolutely nothing about the safety of bikeways, there is no reason for vehicular cyclists to change their position. As before, our position is that operating according to the rules of the road for drivers of vehicles is the safest and best way to travel in traffic for those cyclists who choose that style. Therefore, there is no justification for laws, such as CVC 21202, that attempt to prohibit such operation.

.

Other International Comparisons

…of traffic overall, not bicycling specific. See this blog entry that discusses Leonard Evans Am J Public Health 2014 article:  Traffic fatality reductions: United States compared with 25 other countries. The US lags mightily in reducing traffic fatalities overall.

According to wiki’s List of Traffic Fatalities by County; US is 7.6 per 100BKm vs. Netherlands 4.9. Per capita is 11.6 and 3.9 fatals per 100,000 population (retrieved 3/18/2015; quoting 2012 stats at the time) for US and NL respectively.

Strict Liability laws

I didn’t know quite where to put this so I plopped it here: the concept of motorist stict liability is sometimes misconstrued — here is a detailed explanation of such laws,  article 185 of the Wegenverkeerswet (Road Law), from bicycledutch.wordpress.com which is by the way an excellent resource for understanding all aspects of bicycling in the Netherlands. There’s a longer more broad view of the topic at cycling-embassy.org.uk.

In short it (where it exists) affects civil liablity only not criminal; and as such is more of a behind-the-scenes insurance company type of affair unlikely to affect anyone’s behavior very much one way or the other.

Journal of Safety Science

Here’s a cyclists are drivers! thread where Paul Schimek mentions a Journal of Safety Science article from ~ 8/2015 (it’s accepted but not yet published so i had a hard time getting it) about NL bicyclist traffic safety.   The Dutch road to a high level of cycling safety P. Schepers et al. Among other hypotheses they tested, they dimissed the legal liability framework as an explanation: “Notwithstanding the good reasons for liability laws related to vulnerability of cyclist and pedestrians and fairness, there is, as yet, no empirical and theoretical support for a link with cycling safety”.

They do attribute safety improvements to, among other things, to a “heavily used freeway network (that) shifts motor vehicles from streets with high cycling levels”, and note that this is not due to separated bicyclist infrastructire per se.

AZ and NL Overlay

Netherlands is significantly different than the US in terms of size, population, and particularly population density and land-use patterns.

Here’s what an overlay of Arizona compared to NL looks like.

NL has a land area of just over 13,000 mi2 and a population of 16,924,632 for a density of 1,300 per square mile. (wiki)

Arizona is 113,000 sq mi, and a population of 6,626,624 for a density of 57 per square mile. (per wiki, per 2010 census data)

Another fun comparison, Maricopa County in Arizona (the county that contains the Phoenix metro area ) is just under 10,000 sq mi, just a bit smaller than the country of NL’s 13,000. Maricopa County’s population was 4.168 million in 2015.