Three weeks ago I discussed a new paper by Boufous and Olivier that found the trend in the absolute number of cycling fatalities in Australia fell over the period 1991-2013 at an average rate of 1.9% p.a. (see Why is the number of cycling fatalities falling?).
The paper attracted some criticism. For example, commenter James Steward argued the trend is sensitive to the time frame:
If you start the range of years at 1998, however, the trend line has a positive gradient i.e. the number of cyclists who died in crashes increased over 1998-2014. On the other hand, if you include years as far back as 1989, the downward trend is even greater than 1.9% per year.… I’d say this claim of a decline is mostly bollocks.
The counts vary a lot from year to year, so of course it’s possible to select a shorter period that shows a different trend e.g. pick 2008-14 to show an increase, or 1997-03 to show a decrease.
Charting the numbers
So I decided to chart the entire data set on cycling fatalities held by the Bureau of Infrastructure, Transport and Regional Economics; to avoid charges of cherry-picking I haven’t excluded any years. The data is available for the period from 1989 to the end of June 2015.
Because I’m charting calendar year data, I’ve made the assumption that the number of deaths in the second half of 2015 will be the same as in the first half. That enables me to chart the entire 1989 – 2015 period. (1)
It’s clear from the exhibit that deaths plummeted between 1989 and 1992, coinciding with the introduction of mandatory helmet legislation in Australia. Whether the effect was due to fewer head injuries or fewer cyclists is a contested issue, but either way it’s reasonable to conclude the law was the key reason.
Thereafter the numbers vary from year to year, sometimes quite markedly, as you’d expect with such small counts. Nevertheless, even if the years when helmets were introduced are left out, the linear trend in fatalities is still downward over 1992 to 2015. (2)
In any event the really important message from this data is that cycling fatalities are falling in “real” terms i.e. it’s very likely the level of cycling increased substantially over the period but the number of fatalities declined or, if you prefer, were stable.
There’s a second issue raised in the Boufous and Olivier paper that’s worth exploring a little further too. They found deaths resulting from a collision with a vehicle declined over 1991-2013 while those that involved no vehicle increased. The latter category accounted for 3% of all deaths in 1991 but 22% in 2013.
Another recent study, Injury patterns and features of cycling fatalities in South Australia, gives some important insights into this finding. Kelly Olds, Roger Byard and Neil Langlois examined cycling fatalities in South Australia over 2002-2013.
What distinguishes their study is they looked at medico-legal autopsy reports to determine cause of death.
They note that 42 cyclists – of whom two were female – died over the period. Of these, 29 deaths were due to trauma resulting predominantly from collisions with motor vehicles. The other 13 were “collapses” due to an underlying medical condition, mostly heart disease, that resulted in death while cycling.
The average age of these who died as a result of crashing was 43 years and all but two were male. However the average age of those who died of natural causes was 55 years; they were all male.
A quarter of all deaths of cyclists aged 40-49 years were the result of natural causes. This rose to a half for those aged 50-59 years.
The findings indicate the risk of death from crashing with a car when cycling is not as high as the raw numbers suggest.
It also signals that older males need to be mindful that vigorous cycling – as with any form of exercise – can be fatal if the rider has a serious medical condition like heart disease. It’s likely though that the exercise benefit of cycling prevents many more premature deaths from disease that it causes.
BITRE have now put up the data for July 2015. Fortunately, there was only one cycling fatality in Australia during the month, raising the possibility the full year count might be well below trend.
The trend is flat over 1998-2015.