Saturday 23 February 2013

In absolute terms - I love Copenhagen

"Bicycle Tracks and Lanes: a Before-After Study"by Søren Underlien Jensen, Paper, January 2008 on first sight is saying that cycle tracks and lanes are unsafe. But is it? I - intrigued - had a look at the paper.

Let's have a look at absolute numbers first. As with any study, that's the raw data to you and I. Absolute numbers are the totally unmitigated figures, not adjusted for context of traffic volume or distance travelled. Using absolute numbers without considering the surrounding circumstances can lead to hilarities like this one. Remember this? Once upon a time there was Norman Baker who had a 'cycle' chum Mike Penning who proudly pronounced that "The Netherlands could learn from UK on cycle safety" covered by bikebiz, roadcc and guardian.

Little Mikey was looking at total cycle deaths (UK 100 vs NL 300) rather than contextualising it with the number of people on bikes and/or the distance cycled. This silly man was actually saying "Cycling in the UK is safe (there aren't any cyclists)" or "UK is a very safe country for bear attacks and mosquito bites".

Absolute figures are important for before-after comparisons as total collisions numbers may actually go up due to increase in ridership and journeys.

Anyways, back to Denmark.


The absolute figures from Jensen study are as follows ref. Tables 3

Observed Observed

Category BEFORE AFTER Reduction
Crash All 2,987 2,911 3%
Injury 1,313 875 33%
Property damage only 1,674 2,036 -22%
Injury All 1,476 937 37%
Fatal 25 22 12%
Severe 757 665 12%
Minor 694 250 64%

Observed Observed

Category BEFORE AFTER Reduction
Intersection All crashes 2,010 2,171 -8%
All injury 938 636 32%
Links All crashes 977 740 24%
All injury 538 301 44%

Observed Observed

Category BEFORE AFTER Reduction
Pedestrians, all injuries Total 469 315 33%
At intersections 267 197 26%
At links 202 118 42%
Bicyclists and moped riders, all injuries Total 574 406 29%
At intersections 353 285 19%
At links 221 121 45%
Motorists, all injuries Total 433 216 50%
At intersections 318 154 52%
At links 115 62 46%

The numbers of slight concern would be the negative reductions (ie increases) in the right-hand column. I say 'slight' because injuries are down and only property got damaged. Now... this is where we should not forget that these absolute terms haven't even been adjusted for the increase in cycling which was a respectable
  • "20% increase in bicycle & moped traffic and a 10% decrease in motor traffic" on tracks
  • "5% increase in bicycle & moped traffic and a 1% decrease in motor traffic" on lanes
So how is this study used to "prove" cycle infrastructure is dangerous? Well. The Jensen study tested a prediction model (these fancy algorithms listed on early pages in the study) which already factored in a reduction in collisions. And the prediction model predicted lower reductions than were achieved in reality. That's it.

My advice, be skeptical when someone quotes yet another study that "proves" that cycle facilities are dangerous. First look at totals, then check whether adjustments for increase in bike traffic, distinction between collision or injury categories (such as fatalities, major, minor, near-hits) have been made.

And yes, it's important to get the design of cycle tracks and lanes right too. Remember: existence of real cycleways in the UK is few and far between. What the Jensen study is talking about are proper pedal routes, not the substandard sh*te that cyclists - condemned to second-class citizenship by the UK authorities - get here.


  1. This is a fantastic piece of sleuthing. It's quite clear that there's an overall increase in safety when you point it out, yet I wonder why Jensen chose to trumpet the "decline" of safety in his papers.

    1. Hi Examined Spoke - I don't think there were any "sinister" intentions by the research team. It is my understanding that the researchers just carried out a correlation test to quantify how WELL in practice cycle tracks and lanes perform against their initial theoretical assumptions. In effect a comparison of a theoretical model against field data. No great shakes. Good practice even.

      I am now reading about this study here (German) another study which might prove more challenging to see through.