Lies, Damn Lies, and Statistics

[Diver300]

How much less likely is it that a mile of cycling compared to a mile of driving will kill a pedestrian?

One of the comments in this:-http://www.bristolpost.co.uk/police-cyclists-could-soon-be-undercover-in-bristol-to-target-drivers-for-dangerously-close-passing/story-30070963-detail/story.html, which is a copy of what is said here http://www.cyclinguk.org/campaigning/views-and-briefings/pedestrians put it at a bit over 2 times less likely.

The figures were 1% of pedestrian fatalities due to cyclists, with 2.3% of urban, non-motorway miles travelled. I would guess that the vast majority of pedestrian road fatalities are in urban, non-motorway situations, so that seems a good measure.

In 2015, in the UK, 409 pedestrians died on the roads (https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/533293/rrcgb-main-results-2015.pdf).
That also shows 3.2 billion vehicle miles for cycling, and 247.7 billion vehicle miles for cars and taxis so about 1.3% is cycling. That is probably consistent with 2.3% of urban traffic.

The data from 18 years ago to 10 years ago (http://www.publications.parliament.uk/pa/cm200809/cmhansrd/cm090126/text/90126w0003.htm#09012627000041) shows 0.5 % of pedestrian deaths in that time due to cyclists, but the ratio appears to be increasing as the casualty rates for pedestrians was falling. The rates of pedestrian deaths due to cyclists was clearly subject to a large random variation, and was 1% in the final year of that report.
Further information is here:-https://www.whatdotheyknow.com/request/uk_accident_statistics_for_pedes which shows much lower figures for pedestrians deaths, but also falling.

The figures seem to show similar likelihood of pedestrian deaths per mile of cycling or mile of driving. It seems odd to me, as the cyclists is bringing so little energy to a collision compared to what a car brings. Are the figures right? Is it risk compensation, in that pedestrians will step into the road when they can't hear a car? Is pavement cycling so prevalent that the collision rate is so high that fatalities result?

Does it matter? Car drivers kill around 750 car occupants a year, while I am fairly sure that cyclists don't kill any, so adding in all victims makes the pedestrian deaths a small proportion.

[spindrift]

Discussed here:

http://road.cc/content/news/109269-are-drivers-and-cyclists-just-dangerous-pedestrians

Cyclists are just as dangerous to pedestrians as drivers. That’s the claim made by an article on The Times (link is external) website today.

According to transport correspondent Phillip Pank, analysis of the 2012 road accident figures published by the Department for Transport reveals: “When serious injuries are measured as a proportion of distance travelled, cyclists injured 21 pedestrians per billion km travelled in 2012 compared with 24 pedestrians injured by drivers.”

To steal a phrase from debunker of Bad Science Ben Goldacre, we think you’ll find it’s more complicated than that.

[spindrift]

Broadly, it's statistical innumeracy, it counts motorway miles and of course there are no cyclists on motorways.

It's a tricky field though, the risk of cycling. Risk per mile is probably best but it gets confusing when old people cycle a lot. Obviously that means it must be a GREAT place to ride but sadly the old people die and that skews the figures. Classed as a cycling death I think. You need a proper statistics person to explain it better than me.

[red marley]

It illustrates the perils of comparing very large absolute numbers (car miles travelled) with much smaller ones (bicycle miles travelled) when calculating a very unlikely event (pedestrian being killed by either). Taking this approach further you could conclude that riders of green 2012 Raleigh Mavericks with scratched paintwork are thousands of times more likely to kill a pedestrian than a car driver (substituting whatever the characteristics of the bicycles involved in those very rare events are). And that's before you consider the fact that the spaces in which motor vehicles accumulate many of those miles is almost entirely free of pedestrians.

Perhaps a more pertinent question for any pedestrian (which, lets face it is almost all of us) is in the unlikely event of being killed on the road how would it happen? The answer is overwhelmingly that it would involve a driver of a motor vehicle (about 98% of traffic incidents involving a pedestrian fatality).

[Kim]

Spacecraft miles probably compare favourably to bicycle miles in terms of death due to mechanical failure...

[yoav]

Apart from the motorway issue mentioned above, there's another reason why the "per mile" measure is inappropriate.  As cars generally travel faster and go further than bicycles, injury rate per hour of driving/cycling would be a more appropriate measure.

[Cudzoziemiec]

One of the comments in this:-http://www.bristolpost.co.uk/police-cyclists-could-soon-be-undercover-in-bristol-to-target-drivers-for-dangerously-close-passing/story-30070963-detail/story.html, ...

I never thought I'd have a reason to say this, but Thank you, Birmingham.

[hatler]

One of the comments in this:-http://www.bristolpost.co.uk/police-cyclists-could-soon-be-undercover-in-bristol-to-target-drivers-for-dangerously-close-passing/story-30070963-detail/story.html, ...

I never thought I'd have a reason to say this, but Thank you, Birmingham.

There's more than the usual amount of bile in the comments there. Good grief.

[PhilO]

Apart from the motorway issue mentioned above, there's another reason why the "per mile" measure is inappropriate.  As cars generally travel faster and go further than bicycles, injury rate per hour of driving/cycling would be a more appropriate measure.

That rather depends... If you're seeking to compare the risk of going out for an afternoon drive and going for a similar bike ride, then yes. OTOH, if you're seeking to assess whether it's better to do a 5 mile commute on two or four wheels, then the per mile figure is more meaningful.

[Karla]

About 420 pedestraians got killed on average over the last five years, so that makes it 4 a year killed in a collision involving a bicycle (with no data for whether any were ped/bike/car pile-ups).  That's 4 pedestrians a year, out of a population of sixty whatever million.

You're really pushing the boundaries of meaningful statistics if you think you can generalise risk rates from 4 deaths a year.

[DuncanM]

About 420 pedestraians got killed on average over the last five years, so that makes it 4 a year killed in a collision involving a bicycle (with no data for whether any were ped/bike/car pile-ups).  That's 4 pedestrians a year, out of a population of sixty whatever million.

You're really pushing the boundaries of meaningful statistics if you think you can generalise risk rates from 4 deaths a year.

Is that a calculation, or is there a source for those numbers?  Last time I looked for data (a few years ago), the numbers of pedestrians killed in the UK by cyclists on the prior 3 years were 0,1,0.  4 per year seems an awful lot.
Cheers
Duncan

[hatler]

Is that possibly the difference between deaths of pedestrians involved in a collision with a bike vs deaths of pedestrians on the pavement involved in a collision with a bike ?

[Karla]

Total pedestrian deaths from Casualties involved in reported road accidents (gov.uk)*
I was taking 4 as 1% of that.

I do recall from last time I looked at these sorts of figures, that pedestrians getting killed off-road by bikes was more common than getting killed on-road (the latter being something like 1 a year, as you say), but it's pretty hazy so don't quote me on that.  Another source could be the difference between "killed in a collision involving a cyclist" and "killed due to the fault of a cyclist".


*Interestingly, the data is in Openoffice .ods spreadsheets, not Microsoft .xls/.xlsx - I wonder why?

[Cudzoziemiec]

One of the comments in this:-http://www.bristolpost.co.uk/police-cyclists-could-soon-be-undercover-in-bristol-to-target-drivers-for-dangerously-close-passing/story-30070963-detail/story.html, ...

I never thought I'd have a reason to say this, but Thank you, Birmingham.

There's more than the usual amount of bile in the comments there. Good grief.

Normal for the Post. The same bile has leaked out on to an article about the police tasering their own race relations adviser.

Undercover police WILL target Bristol drivers who pass cyclists too close. Bristol police Taser prominent member of their own race relations group in the face. One day, one police force, two headlines, two sets of assumptions and two very different approaches and outcomes whereby one section of the community enjoy special protection, the other a very extra special special treatment. Why ? Bristol a city divided by class, wealth and race and opportunity. Marvin how do you feel about this as you're in charge ? It's starting to look a bit like another tragic case who's tragic outcome was predicated upon crass assumptions about race, community and who the police had time for and believed, trusted and acted to protect or not. Bijan Ebrahimi RIP.
Read more at http://www.bristolpost.co.uk/investigation-as-bristol-police-taser-prominent-member-of-their-own-race-relations-group-by-mistake/story-30072082-detail/story.html#JAlmwYwDmHk1L4P3.99

Sometimes people will look for a link where there is none.

[Steph]

Two things in response to this:
Firstly, the famous "I'm not stopping" stitch-up was on-road, but the MP involved still deletes any comments saying just that, as she is suggesting it was off-road/on the footpath.
Secondly, the bikes-more-dangerous-than-white-van-man bollocks was pushed very hard by a certain 'fading away' organiSSation  that used to troll here and elsewhere. That says all I need to know about the weight of the statistical method involved.

[Diver300]

About 420 pedestraians got killed on average over the last five years, so that makes it 4 a year killed in a collision involving a bicycle (with no data for whether any were ped/bike/car pile-ups).  That's 4 pedestrians a year, out of a population of sixty whatever million.

You're really pushing the boundaries of meaningful statistics if you think you can generalise risk rates from 4 deaths a year.

Is that a calculation, or is there a source for those numbers?  Last time I looked for data (a few years ago), the numbers of pedestrians killed in the UK by cyclists on the prior 3 years were 0,1,0.  4 per year seems an awful lot.
Cheers
Duncan

I was under the impression that the numbers were in the 0 - 1 range, but the tabulated figures appear to be in the 0 - 5 range.

I can't find a reference, but a few years ago someone petitioned the government to reduce the 645 annual fatalities due to pavement cyclists. The petitioner had taken the rate of pedestrian road fatalities, on and off the pavement, due to motor vehicles and cyclists, and lumped it all on the least likely combination.

That shows that impressions can be wrong. 5 could be typical.

About 420 pedestraians got killed on average over the last five years, so that makes it 4 a year killed in a collision involving a bicycle (with no data for whether any were ped/bike/car pile-ups).  That's 4 pedestrians a year, out of a population of sixty whatever million.

You're really pushing the boundaries of meaningful statistics if you think you can generalise risk rates from 4 deaths a year.

With a pedestrian death rate from cyclist collisions running at around those numbers, the threat posed by cyclists to the average pedestrian is minimal, and so it appears that pedestrians are far less cautious of cyclists, which probably increases the accident rate per mile cycled.

e.g. https://www.youtube.com/watch?v=G4K8AjNIVPA where the pedestrian blocks a cyclist for proceeding behind him on a zebra crossing, seconds after ignoring a taxi that had done exactly the same.