When I saw the headline (“Safety Gap Grows Wider Between S.U.V.’s and Cars”), I was all set to rail against the corresponding article. I figured they were going to claim again that small cars were less safe than SUVs, without bothering to compare the rate of traffic deaths in small cars in collision with other small cars against the rate when they collide with SUVs. If small cars are less safe, in all likelihood that’s only when they collide with SUVs.
Instead, the point of the article is that the probability of dying in a car accident is 11% higher if you own an SUV than if you own a regular car. I’ll be damned.
The New York Times again proves Paul Krugman correct, however: it strives for balance rather than for objectivity. E.g.,
Industry groups have insisted for years that S.U.V.’s are at least as safe as passenger cars, if not safer. One group run by industry lobbyists, called the Sport Utility Vehicle Owners of America, says on its Web site that it is a myth that S.U.V.’s guzzle gas or that their higher rollover rate makes them more dangerous for their occupants. Ron DeFore, a spokesman for the group, cited statistics from the insurance industry, which found last year that fatality rates for newer sport utility vehicles were markedly improved from older models.
“Most people have gotten a skewed vision about the S.U.V. and think they’re unsafe, and that’s just not true,” Mr. DeFore said.
This is the obligatory “what the other side says” paragraph. Rather than look for the real answer in an unclouded light, the Times seeks to find a representative from each side of the story and ask what they think. The trouble is, one side could be obviously wrong. And yet the Times always gives equal time to both. Krugman’s line was that if the Bush administration said the Earth was flat, the headline the next day would read “Shape of Earth: Views Differ.”
The article writer compounds the error with a non sequitur. The paragraph begins with a sentence rebutting the claim that SUVs are unsafe, then goes into a second sentence arguing that SUVs have gotten safer. How does the second sentence provide any support for the first? SUVs may have halved their rollover rates, but may still have twice the death rate of passenger cars. Yet the Times doesn’t bother to point out the error. Maybe the author didn’t realize it?
Why should we trust an industry lobbying group — which has an incentive to push high-margin SUVs — against a presumably dispassionate government agency collecting statistics?
Nonsense:
This year, the government started conducting rollover tests on a test track rather than merely analyzing the vehicle’s dimensions on paper to determine rollover risk, as it had done in the past.
I have an idea what “on paper” means: after an accident happens, they determine a) whether the car rolled over, and b) what its dimensions are, and maybe c) how the load inside the car was distributed. They plug these numbers into a mathematical model, which helps them relate the object’s center of gravity to its probability of rolling over. This is precisely what they ought to be doing. I continue to believe that newspapers are scared of math, and view it as somehow cold or inhuman.
The point of math and statistics is to make precise what had previously been vague. For instance, this paragraph:
Rollover risk, though, is only one part of the safety picture. In crashes between vehicles, heavier vehicles tend to perform better than lighter ones, which is one reason that the smallest cars tend to have the highest occupant-fatality rates. The ways that people who own different types of vehicles tend to drive them is also a factor, especially in the case of sports cars.
Here is where they ought to be using a statistical model: how does the probability of death vary with the weight differential between the two cars in the accident? (In a multi-car accident, we might use the difference between the heaviest and the lightest car.) In this model, we would either hold other factors — like the speed of the collision — constant, or we would aggregate over all collision speeds. There are well-known statistical methods to use here. Why are we relying a) on vagueness, and b) on secondary sources like industry lobbying groups to tell us what makes one vehicle safer than another?
Ditto:
But weight is not a simple proxy for safety. In a federal crash study this year, large passenger cars and station wagons, averaging about 3,600 pounds unloaded, were found to have a death rate of 3.3 for each billion miles traveled; they were second only to minivans, which had a rate of 2.76.
Don’t bother breaking cars into categories; model the death rate as a function of weight, and tell me — on average — how much the death rate increases for every 100 pounds of weight increase.
Finally, and unrelated to the newspaper article, there’s an economic question. A completely unregulated auto industry would externalize the cost of its unsafe vehicles: if people die, someone else pays for it. A well-regulated auto industry would be required to pay for auto recalls, would be subject to lawsuits, and so forth. So the statistical question for me is: of those 40,000+ auto deaths (and the substantially higher number of serious auto injuries), how many are the auto industry’s fault? That is, how many could not have been prevented, given the car’s design and reasonable care on the part of the driver? So now take the total cost of these deaths and injuries, and subtract the total amount that the auto industry paid out for lawsuits and product recalls. Is this difference positive? That is, how well have regulations forced the industry to internalize its social costs?
