Fabula data: road deaths in Australia. Young or Old?

There are a 1001 ways to lie with data. Here's one. 

Have a quick look at this chart. It shows data about the 1300 people that died on our Australian roads in 2024, divvied up by age groups. This chart and its data are straight from the Australian government's Bureau of Infrastructure and Transport Research Economics, BITRE.  According to this chart middle-aged people (aged 40 to 64) are most at risk of dying on the road. This data includes any mode of transport, so whether on foot, driven, driving, or riding. All are included. I happen to be part of that 40-64 cohort, so I suppose –according to this graph– I should be somewhat more nervous and/or vigilant while out there getting to and from places on the road. But I can be comforted to know that it's on the government's radar: with data-driven good information we can prioritise our efforts and funding to where it helps the most the fastest the best.

Calendar Year data for 2024 on Road Deaths in Australia. Screenshot from BITRE Road Death dashboard (data currency, Apr 2025)

 

However ...

What is really happening here is that this chart confabulates a story that doesn't exist; Australia doesn't buck any global trends. Young people are still the most at risk on our roads. Much more than middle-agers. Have been forever. Still are.

Just zoom in to that category axis. Size matters. In this case the size of the age groups. The cohorts here group together different number of ages: the cohort from 0-16 is 17 years worth of Australians, while the age group 17-25 represents only 9 years. The table below shows you that the middle age cohort for 40-64 year-olds, with 25 cohort years, is the largest by an almost comical margin: almost 3 times the size of the smallest cohort, the 17-25 year-olds. In whole number 2024 Australians this is 3.1 million 17-25-year-olds, vs. 8.3 million 40-64-year-olds. I suppose I should mention that I didn't count the technically open-ended cohort of 75+. (The oldest person to die on the road in Australia in 2024 was 97, BTW) 

Why?

The only remotely valid reason that I could come up with to visualise this data against such wildly uneven age-cohorts is to  somehow even out population-wide age distribution, but a glance at population clock and pyramid data for 2024 at the Australian Bureau of Statistics here  shows this is not warranted at all. So the mystery of the uneven cohorts stands. 

Population pyramid for 2024. Australian Bureau of Statistics. 

 

I asked the Australian Bureau of Infrastructure and Transport Research Economics who compile this data and publish the road death graph, why they chose this categorisation. The answer I received mentioned "age categories are based on age-based risk" and "are in line with the reporting [...] to the National Road Safety Strategy 2021-30".

I'm not sure this makes any sense to me. Firstly, if risk is something to do with chance and severity of the outcome (and the outcome death is roughly equally catastrophic for young and old people), then what we want to do is estimate the chance of death@age in the near future by measuring the occurrence of death@age in the recent past. Which is precisely what is being misrepresented here by grouping so unevenly. Furthermore: if data is being reported to decision makers or policy makers running our National Road Safety Strategy, then I hope that they normalise that data on their end, because the age-category  story as told by BITRE's chart is misleading at best. 

Fix this? 

In the meantime the stocky data pro in me thought of some of the many ways to normalise this data and chart to help it tell the story that this data contains: do fatal incidents on our roads relate to the age of the dead road users and if yes, then how?

One solution to normalise this data for the varying-size age cohorts is to divide each cohort's number of victims by the number of years in the cohort. This does –however– give us the slightly abstract measure of death per cohort-year per cohort. In stead of that, I have just chosen to evenly group the victims in 15 year cohorts. This results in roughly the same number of cohorts, so a very similar visualisation, with the bonus that it now shows that young people are the most at-risk-of-death on the road, not the middle agers.

 

 

You can argue whether there are even better ways of normalising this sort of data. Should the exact age structure of the Australian population be reflected? Is there a structural difference in the distances that people drive at different ages? Perhaps road-tripping to far-away music festivals is a risk factor for young people. But then again, maybe the sales-representative job puts middle-agers back in the game. IDK. 

What I do know is that the plain age categorisation based on the uneven age categories, as chosen by BITRE, dramatically and arbitrarily changes the story that the data actually tells.