Cycling comes to Google Maps in Australia

Professor Chris Rissell wrote an article for The Conversation this week with the self-explanatory title, Australian cycling boom? – Nope, it’s a myth.

This is a distinctly more moderate claim than the 38% fall in per capita cycling over 1986-2011 that he and his co-author, Chris Gillham, claimed a few weeks ago. But as I’ll show below, they’ve still got it very wrong.

Their claims are based on this article they wrote for the May issue of the on-line journal, World Transport Policy and Practice. I discussed it recently, noting it had a number of shortcomings.

These include errors of logic, reliance on problematic data and selective use of dates. I also noted the authors chose to compare the level of cycling against population rather than the more obvious and customary level of travel.

These shortcomings don’t demonstrate conclusively that the authors got it wrong, but they do suggest their conclusions should be interpreted with considerable caution.

However since writing that critique, I’ve subsequently gotten access to both the surveys they relied on: Day to Day travel in Australia 1985-86 and the 2011 National Cycling Participation Survey.

CORRECTION 8.30am 13/7/2012: I have to acknowledge I got the specific technical “flaw” I discuss below wrong. It isn’t a flaw. I’ve left the text under the fold for the record and apologise to the authors and readers. However I don’t resile from my other criticisms of the article.

These reveal there’s a more fundamental flaw in their analysis. The problem arises because the two surveys count cycling in different ways. The 1986 survey asked how many times respondents had cycled that day, whereas the 2011 survey asked how many times they’d cycled in the last week (and also in the last month and in the last year).

The authors therefore had to convert the 2011 weekly cycling rate to a daily figure so they could compare the two surveys. In their own words, they multiplied:

the estimated number of people in each state who had cycled in the past week by the averaged weekly number of bicycle trips for that state/territory, and then divided this number by seven to get the average daily number of bicycle trips.

Plenty of noise is introduced by these sorts of interpolating manoeuvres, but that’s not the key problem. Where they’ve gone wrong is in assuming the proportion of residents who cycle in a day is one seventh of the proportion who cycle in a week.

In the 2011 survey, 18% of respondents said they had cycled at least once in the past week (and 26% in the last month and 40% in the last year). So Gillham and Rissell divide the weekly level by 7 and in effect conclude only 2.6% cycle at least once per day.*

Of course that’s not right. Cycling levels aren’t distributed evenly by time period across the calendar. That’s obvious, just compare the annual and monthly numbers. Many cyclists venture out much more frequently than weekly. Some cycle on multiple days per week to get to school or work, or for exercise or training.

The error of logic is easily seen: If the monthly figure (say) is divided by 4 to arrive at a weekly estimate, it gives a figure of just 6.5%, considerably less than the actual figure of 18% reported by respondents. Or if the annual figure (say) is divided by 12 to obtain a monthly estimate, it yields a mere 3.3%. Again, that’s just a fraction of the 26% actually reported by respondents.

This is not a minor error. It means Gillham and Rissell very seriously underestimate the level of cycling in 2011. I don’t know what the daily figure is that they should’ve used, but I’d expect it would probably be around 10%, perhaps higher. Certainly it would be much more than the 2.6% they’ve effectively assumed.

I’ve attempted to replicate their method using their numbers for the ACT. They say the population of the Territory aged over nine years increased 72% over 1986-2011 while daily cycling by this group increased by only 47%.

Even if I make the extremely conservative assumption that just 3% of ACT residents aged over 9 years ride per day, the increase in the level of cycling over 1985-2011 matches the increase in the comparable population. If I assume 10% cycle per day – which seems plausible to me – then cycling grew more than five times faster in the ACT than population.

If I’m right, the Gillham and Rissell article is way off the mark and gives a very misleading impression of how cycling is faring in Australia. Indeed, it gives the completely wrong impression.

Anyone can make inadvertent errors. Once identified though, they should be corrected immediately simply because they’re wrong. In this case there’s the added concern that governments who’re already turning their back on cycling could use this sort of negative information to support their stance.

This issue should be cleared up. Unless the authors show I’ve misunderstood them, the article should be amended or buried as soon as possible.

________________________________

*Gillham and Rissell only looked at cyclists aged nine years and over so their actual numbers will be different, but the logic is the same.