The Guardian -v- Addison Lee
May 3, 2012 3 Comments
I was wondering why the London-based taxi firm, Addison Lee, had so many bad reviews all of a sudden on its iPhone app and so did a little googling. This Guardian article explains part of the controversy*, which concerns a quote by the company’s chairman:
“There has, as we all know, been a tremendous upsurge in cycling … These cyclists are throwing themselves onto some of the most congested spaces in the world … the influx of beginner cyclists is going to lead to an overall increase in accidents involving cyclists.
“The rest of us occupying this road space have had to undergo extensive training.
“It is time for us to say to cyclists, ‘You want to join our gang, get trained….”
The Guardian felt differently however, using as evidence the fact that motorists are responsible for up to 75% of collisions involving cyclists. Their source was an earlier Guardian article which contained this:
The data, which was analysed by the Transport Research Laboratory (TRL) …found that police attributed responsibility for collisions to … the driver solely in about 60%-75% of all cases, [with] riders solely at fault 17%-25% of the time … CTC said the report showed that the government needed to focus more on driver behaviour …”We believe this report strongly supports our view that the biggest problem for cyclists is bad driving.”
I don’t agree that the data supports that the view that the government should concentrate on driers [though this may be true]. The statistics reported relate to the probability that party x caused an accident given that an accident occurred. It can often be useful to analyse something with extreme numbers, so let’s do that.
Imagine there are 1,000,000 motorists, just 4 cyclists, and 3 collisions. Two of the collisions are the fault of the driver, and one is down to the cyclist. So 66% of accidents are caused by motorists, but this ignores the fact that there are relatively more drivers than cyclists (ok, a lot more given the numbers we picked). The bigger picture is that in this version of London a mere 1 in 500,000 motorists cause accidents, as opposed to a massive 1 in 4 cyclists.**
A related stat I remember: you have the same chance of dying on a 200-mile motorbike ride in the UK as you do of getting killed whilst serving for a single day in the British army in Afghanistan (a year or so ago).
—–
* The actual source of bad reviews seems more likely to be by black cab drivers not taking kindly to their unique privilege of being able to use bus lanes being challenged – a challenge which I think has some merit.
**When I first heard about Monty and his 3 doors I didn’t guess to change door, so happy to receive a challenge on this way of looking at the stats…
Marc Gawley 
You state “The bigger picture is that in this version of London a mere 1 in 500,000 motorists cause accidents, as opposed to a massive 1 in 4 cyclists”
I think the inference you draw, at least as I perceive it, is misleading. Rewrite the above sentence to “The bigger picture is that in this version of London a mere 1 in 500,000 motorists cause >>cycling<< accidents, as opposed to a massive 1 in 4 cyclists" This added adjective emphasises motorists have tiny exposure to accidents with cyclists while cyclists have large exposure to accidents with cars. The fact that in this version of London very few motorists "cause" accidents with cyclists is at least partly explained by the relatively tiny exposure to such accidents.
Hence I am not sure your inference (as I perceive it) is correct. Your version of London does not, to my mind, argue that bicyclists are "more often" at fault than drivers. There is missing exposure information. Maybe I am misunderstanding something.
Hi Piet, thanks for this. The exposure idea was something I’d thought about, and wasn’t quite sure if important or not – It’s true that there are way more cars than cyclists, and that therefore most motorists will never encounter a cyclist.
If we say exposure is an event in which an accident can occur, and that is when a car and bike pass each other in someway, then there are identical numbers of instances of cars passing bikes as of bikes passing cars. So I think the exposure in this definition is the same. What do you think?
I guess my issue is with “only 1/500 000 motorists “causing” accidents” ie the denominator in this calculation. Most motorists (in this hypothetical London) never encounter a cyclist to have an accident with and hence never cause such an accident.
My understanding is as follows. Suppose p is the probability, in a car/bike meeting, the car causes an accident and q is that for bikes. Then the probability of an accident in a bike/car meeting is 1-(1-p)(1-q)=p+q-pq. Given an accident, the probability of the car having caused it is then p/(p+q-pq), and similarly q/(p+q-pq) for bikes. The ratio, p/q, indicates the relative propensity for causing an accident given a car/bike meeting. In your example p/q=0.66. The fact that there are more cars appears irrelevant. The conclusion is that a car is twice as likely to be the cause of an accident as a bike.
Maybe my understanding is wrong or I am missing something.