Everyone who’s ever done research has run into cases where the data fail to match up to prior expectations. As the saying has it, there’s nothing so tragic as a beautiful hypothesis slain by an ugly fact. I’ve just run into something of this kind in relation to the debate over road safety that’s been going on for some time on this blog. I’m still thinking about how to interpret the data I’ve found, but for the moment I’ll just report it.
A statistic that often comes up in the debate about road safety is the death rate per million vehicle-miles or vehicle-kilometres. This statistic is commonly used by defenders of US policies since it’s the only one that makes the US look relatively good. The rate of road deaths relative to population in the US is about twice that in Australia, and other leading countries (about 16 per 100 000 as opposed to 8 per 100 000). On the other hand, measured in terms of vehicle miles travelled per person, Americans appear to travel about twice as far, cancelling out the difference.
I’ve pointed out several times that this statistic takes no account of vehicle occupancy rates. Consider two countries in which people travel the same average distance per year, cars are equally likely to crash and occupants are equally likely to die in a crash. The only difference is that in the first country no-one travels as a passenger while in the second every car has one passenger. For nearly all relevant purposes, the two countries have the same performance on road safety. But, on the “deaths per vehicle-kilometre” measure, the first country has half the rate of the second.
In making this point, I assumed both that the US would have relatively low occupancy rates and that occupancy rates would be declining over time. However, looking at these statistics from the National Household Travel Survey, it appears that occupancy rates in the US have actually been rising since 1990, as can be seen from the fact that person miles of travel have been rising more rapidly than vehicle miles travelled. Also, although I can’t get really comparable data, it’s not obvious that occupancy rates in the US are lower than in Australia.
I have some doubts about this data. For a start, the revision of the 1990 data raised estimated person-miles and vehicle-miles by about 20 per cent, and the survey reports warn against comparisons over time. The same warning obviously applies even more strongly to international comparisons. By contrast, death rates can be compared fairly reliably.
Looking at the actual numbers, I find it surprising that increasing numbers of cars per household and increasing proportions of one-person households can be accompanied by higher vehicle occupancy rates.
Finally, on the comparison between Australia and the US, it’s startling in both absolute and relative terms that the average American is estimated to travel 24 000 km per year (450 km per week) as opposed to about 12 000 km for Australians. The two countries seem similar in most relevant respects, and a difference of 100 per cent surprises me. In absolute terms, the US figure seems very high, bearing in mind that more than a third of the population is under 16 or over 65, and presumably travel a fair bit less than this. To get this kind of average for the entire population, working age adults would have to be averaging something close to 100 km a day, which would presumably imply a couple of hours per day in the car for people in urban areas.
But the fact that the data surpises me doesn’t mean it’s wrong. I plan to investigate further and report on what I find.
John Quiggin 
Correct me if I’m wrong, but isn’t carpooling a bigger deal in America than it is here? For example, in some cities they have dedicated carpool lanes, something which I haven’t seen in Australia. The benefit is so significant that people are prepared to pick up total strangers in order to get in the faster lane — and the practice has got an unusual name: slugging.
Some interesting information on the practice.
I don’t know whether it is so prevalent that it would affect the statistics, but encouraging car-pooling was apparently US government policy from the 1970s onwards.
Car pooling is done here too. Sydney is full of T3 lanes at peak hour, which means taxis or cars with three or more people, and drivers do pick up complete strangers (mostly from bus stops, I think) In the US I have seen T2 lanes, and was solemnly informed by some locals that some drivers had been busted by the police with life-like dummies in the seat next to them. That might be apocryphal though.
-p
I didn’t realise Sydney had those lanes. Has slugging (impromptu carpooling involving complete and utter strangers) taken off there, or is it peculiar to the US?
In order to make meaningful comparison of US and Australian road deaths, won’t you need to take into consideration such factors as:
– road conditions (wet, dry, icy, snowy, etc)
– road surface
– length of journey
– age of vehicle
– age of driver
– etc.
I’ve always had a scunner against “deaths per million miles travelled” because of what I see as an abuse in the case of air travel safety. I find it misleading in that case because the risk structure of travel is not constant over the journey — for most journeys you have a dangerous bit at the beginning where you’re getting out of a densely populated area, then a long safe bit in a sparsely populated area, then a short dangerous bit as you get back into another populated area. Aeroplane flights obviously have lots more miles in the middle section, hence deaths per mile underestimates the “risk per journey”, which I would argue is the thing we really care about (as in, the answer to the question “If I get into this thing now, how likely am I to get out alive?”).
I suspect that something similar may be at work in comparing US to Australia in deaths per mile; my guess is that the median journey length may be radically different.
John, one possible reason that Americans drive further is that there’s actually somewhere to drive *to*. Distances between major cities In Australia make driving a losing proposition in comparison to flying.
Also, public transportation is less developed there by population-relative measures. Compare LA’s public transport (10 million odd?) to Perth’s (800,000) and you’ll see that Americans *have to* drive further.
Apart from that, I’m not surprised in the least by your results. Speed is a vastly overrated factor in accident rates. Road quality is the #1 factor, and Australia’s long haul roads are the worst of any developed country I have ever visited. (And that’s all of them)
Deaths per head of population is surely the key figure and everything else is just for working out where the problem is. If Americans drive twice as much, it is twice as important that they get road safety right.
Re Yobbo’s point, Sweden has some atrocious driving conditions but if I recall correctly is the safest per mile in Europe. I suggest that states have a mildly varying appetite for road risk and wildly varying car use requirements which will be managed to match the risk appetite. So it isn’t exactly just driving conditions. Certainly people drive more slowly on unlit motorways than they do on lit ones. Speed may be more important when it’s 40 in a 30 zone than when it is 90 on a motorway.
INJECTING SOME ECONOMICS…
Tom Nankivell
When Jack says “If Americans drive twice as much, it is twice as important that they get road safety right”, he may be correct, but the real question is: what is “right”?
From an economic viewpoint, “right” is when net benefits are maximised.^ This is not when safety is maximised. Safety measures – lower speed limits, more stringent enforcement, stronger vehicles etc – entail costs. These must be compared to the benefits they bring, with the aim of generating an optimum level of speed, enforcement, safety features etc.#
Driving twice as far certainly does NOT mean that one ought be twice as safe per unit of travel, so that total exposure to risk remains unchanged. After all, if going slowly imposes costs, going slowly for twice as long imposes twice* the costs! On the other hand, as one travels farther, the cost of safety features is disipated over extra passenger miles, which means that the optimum amount of safety features per vehicle increases. However, if kilometres travelled in the US are significantly higher than kilometres travelled in Australia, it would seem entirely appropriate that the US have a higher road toll per person than Australia. Indeed, if they had the same per capita road toll, then this would indicate that their road safety regime was too stringent relative to Australia’s.
__________
^ distributional issues aside
# if new to this debate and interested in more on the notion of optimum vehicle speeds etc, see the ORR paper at http://www.pc.gov.au/orr/reports/submission/complian/index.html
* actually, value of time and frustration incurred may vary as time spent travelling increases
hi john,
“To get this kind of average for the entire population, working age adults would have to be averaging something close to 100 km a day, which would presumably imply a couple of hours per day in the car for people in urban areas.”
the sad thing is that commutes of this type are not uncommon in america. i used to live on long island, out on the east end. some folks would have a one-two hour commute into manhattan. and this was considered normal. sure lots of folks took the 5.30 train into manhattan, but i suspect that you might be underestimating the american determination to get to work and especially by their beloved vehicle! an interesting statistic would be to compare this with stats for the average length of commuting. that data is available i believe, because every couple of years, the papers over here speak about studies that show that the average length of commute is getting longer.
its completely ridiculous…
the government should stop subsidising transport. (putting general taxation money into public transport and roads, which i think occurs in australia although id have to chase the references) and by releasing land out in the styx (in Sydney at least).
Tim Blair’s come out in support of higher speed limits. You might want to respond to his post, John.
Driving hard
In economic terms, time spent travelling is a really big deal. In their book Time for Life, based on the 1985 US Time Use Study, Robinson and Godbey estimate that the average adult American spends 30 hours a week in paid employment and 10 hours …
Driving hard
In economic terms, time spent travelling is a really big deal. In their book Time for Life, based on the 1985 US Time Use Study, Robinson and Godbey estimate that the average adult American spends 30 hours a week in paid employment and 10 hours …