The unexpected result
I trust insurance company studies for one reason: they have a financial incentive to tell the exact truth. The bottom line motivates them to put their prejudices aside and go for the right answer.
So I was surprised by a link one of my sons sent to me, about an insurance company study which found: “High-powered cars less likely to crash.”.
Say, what? Wouldn’t it make sense that muscle cars get in more accidents? Well, sure it makes sense, but it also made sense to Aristotle that heavier objects would fall faster than light objects. It just didn’t turn out to be true.
The way science works is; you have an hypothesis, which is basically just a well-formed guess. You collect data and see how well it fits the hypothesis. If there’s a good fit, you perform analysis and peer review, and then you have a theory, which is a strong predictive model. If contrary data comes up, you need to figure out why it is different and possibly alter your theory or even create a new hypothesis.
That is why unexpected results are so interesting. They lead to stronger theories.
…Progressive, the third-largest auto insurer in the United States… found cars with more than 200 horsepower actually generate 17 per cent fewer claims than those propelled by less than 200 horsepower. However, when the more powerful vehicles are involved in accidents, the dollar value of claims averages 22 per cent higher. Progressive said the effect is consistent: the more powerful a car model, the fewer but more costly its claims.
The insurer’s news release made no mention of factors that might skew the statistical correlation, such as the tendency of high-powered cars to be newer, more expensive and driven by more affluent owners.
I could pose an hypothesis that the reason for this correlation is that owners of more powerful cars are richer, which means they tend to be better-educated, which has been correlated with lower accident rates. To turn that hypothesis into a predictive model (a theory) would require more study.
Here are some other contrary results I find interesting:
- Giving stimulants to some children calms them down
- When bicycle helmets were made mandatory in Britain and Australia, head injuries went up while bicycle ridership went down
- Air bags kill some people
- Only a few models of SUV’s are really safer than normal cars
- Civil-service tests in Boston were found to have an inverse correlation to on-the-job performance in the Boston police department
- Speeding in town is likely to make little difference in your arrival time and may (if a lot of people do it) actually result in overall average slower traffic
- Recycling (except for aluminum) isn’t necessarily a direct benefit to the environment. The recovery process is often toxic and very, very expensive.
Does anyone have other examples to add to the list?