Are motorcyclists DISPROPORTIONATELY likely to be in an accident? That is, given our low numbers on the road versus cars, are we statistically more likely to be in an accident? Technically, yes; but is the scale of the dispoportion being inflated to justify heavy fees on motorcyclists? Below is a little essay - read if bored or interested - that purports to show dodgy and fanciful statistics are possibly inflating the "disproportionate risk" case, to support heavy fees on bikers (who are constructed as a socio-medical 'drain'.
First, I got gouged today, when I paid the Victorian State Government their pound of flesh for car and bike registration. Down here in the Great Nanny State we have our rego fee and also the mandatory TAC (Transport Accident Commission), or insurance. The rego to insurance component of my overall fee was 10:1 for the bike and under 2:1 for the car. That is, for every dollar of registration I spent on my car I spent under $2 on the TAC fee. The rego for my bike was one quarter the rego for my car but for every dollar of bike rego I spent $10 of TAC fee. Hence the ratio of 10:1 for bike, under 2:1 for car. The TAC fee is not factored directly from the rego fee, but that 10:1 versus 2:1 is still high, and I suspect dodgy stats are used to distract people from gouging. Your own State probably has similar stuff going on when you have to pay rego and insurance.
Are the stats showing bike riders dispoportionately at risk from accident inflated? I'm going to say dodgy sampling and fanciful sampling (non-technical terms for statistical errors!) are undermining reasonable public debate by pushing possibly inflated disproportions.
1) DODGY SAMPLING: The Victoria TAC website notes that road fatalities for VIC riders over the past decade are 14% of fatalities. They add that in 2013 motorcyle-related fatalities were 17% of the toll, despite bikes being just 4% of registered vehicles in Victoria. Note the possible sampling error, because it is too vague. For instance, we don't know if we should directly compare the 17 and 4 because the 17 could include out-of-state visitors' bikes. Now if you actually compare stats they provide but do not sit next to each other like that 17 to 4, you find that the 5 year average of road fatalities is 180 car-related and 40 bike-related, but it also turns out that the biggest predictor of crash rates is being over age 70. But the really big instance of dodgy sampling, because it hinges on an equivocation, comes with claims such as "Australian rate of motorcyle deaths is approximately 30 times for car occupants" (QLD Centre for Accident Research). Now, the VIC TAC-website is playing on this "disproportionate accident/death per vehicle on the road" image as well. But, first, note the claim again inflates the proportion: in vehicle accidents, 'occupants' are passengers not drivers and make up a small percentage of victims. So it looks like death from riding and death from driving is being compared, but it is actually death from riding and death from passenger in a car (passenger deaths are small even compared to driver deaths), thereby illigitimately inflating the disprportion claim. For instance, why not compare occupant to pillion (that would be apple to apple).
2) FANCIFUL SAMPLING: If we move on to the Federal Transport Bureau Report of 2004 its says: "In 2001, there were 6.2 motorcyclist deaths per 10 000 registered motorcycles, compared with the OECD median of 3.4. This is significant considering Australia ranks favourably in its overall road safety record. In 2001, there were 1.4 road deaths per 10 000 registered vehicles, compared with the OECD median of 1.8". If we take this as true, what it actually says is that you are five times more likely to die riding your motorcycle as driving your car. But recall that "30 times" figure above, which is repeated elsewhere (google it!). How do you go from 5 times the rate to 30 times the rate? The Transport Safety Bureau Report of 2004 said that "the death rate for motorcyclists is still very high" (p. 184). Apparently, "motorcyclists had a risk of death per 100 million kilometres travelled of between 18 and 25 times that of motor vehicle occupants". Sounds awful, right? But let's overlook the obvious conflation of separate statisics (riders, drivers, passengers and pillions), because when was the last time you met a motorcyclist who had ridden 100 million kilometers? Exactly . . . they do not exist. How is this fanciful rider created? By trying to compare what would happen to a rider if they did travel the kind of kilometers you average out for the kind of extremely large samples of auto drivers that do exist. Put differently, the statisics for justifying DISPROPORTIONATE death/accident risk actually compare real averages for real sample populations of real car drivers (where you have very large numbers doing many kilometers on average) to inferred numbers for non-existent populations of non-existent riders. It's like saying "I saw some violence in a crowd at a Sydney soccer match", so if we take that instance and multiply it, not by all the games on Earth, but by all the games in the Universe if we imagine the Universe as just like Earth, this is how much violence is in soccer games in the Universe. If that false comparison does not leap out to you as problematic, just ask yourself what every rider knows: what would happen to car driver behaviour if there were more riders? We know riders in their cars are more aware of riders, for instance. So the comparison of real drivers to inferred riders ignores common sense, that driver behaviour toward riders changes with more riders. And the disproportion hinges on using a statistical counter-factual - something that does not in fact exist - just to increase its sample size to be able to obtain any result at all.
PS: yeah, I ate lunch in the office today
