I love the animal tweets that Steve Stewart-Williams does, but a lot of his non-animal tweets are excellent also.
OK, I wasn’t expecting that: Car-seat laws function as contraception. They raise the cost of having a third child, because most cars can’t fit three car seats in the back. In 2017, the laws saved 57 lives in the US but led to 8,000 fewer births.
That’s been open on my computer for the best part of a month, but it refused to allow itself to be deleted. Too interesting.
“…the laws saved 57 lives” How could he possibly know?
Ah, it’s an estimate.
I guess by comparing two graphs, one of how things were going, and two of how they actually went. But it does sound suspiciously exact, doesn’t it?
Plus, what about the number of accidents caused by drivers who wear seat belts feeling personally safer, and hence mowing down more pedestrians? What if there had been spikes in the middle of steering wheels, pointing at the driver’s body? Might that have “saved” even more lives?
But I still think the claim that is being made is interesting, because of the much bigger size of the guess number of how many babies weren’t born, because of the two seats three seats thing.
In other words, as you just said while I was writing all that: “It’s an estimate.” Estimate being science-ese for “guess”.
I don’t think describing it as a guess is necessarily fair. As it says in the abstract, different US states have passed infant seat laws at different times. This enables statistics to be compared trying to eliminate other potential causal factors and estimate how much of an impact a particular measure has had. That will also take out the argument about some people driving more dangerously because they feel safer since this will be a net effect. The science of this kind of statistical analysis is quite robust but the accuracy of the resulting estimate depends heavily on the quality of the data and of course the care of the scientist to be objective and not look for the answer he wants to find. There are also good tools for estimating the error range based on the quality of the data. It would be interesting to know what the error ranges are on the 57 and what proportion of infant car deaths that represents and hence whether the figure is in any way significant. Hopefully that is in the full report though I can’t be bothered to read it to find out. The issue of looking for the answer you want to find is an important one due to the social sciences “reproducibility crisis” which is the discovery that on the rare occasions when different scientists conduct different experiments to try to validate or refute a finding they very frequently find they do not get the same result. This would be a major issue in physics. In the social sciences I get the impression a small group of people care a lot and too many don’t. I would like to be told if that is wrong and the issue is now being taken seriously across social science academia.
I confess I only read the tweet, not the article. It was the mere fact of car seat contraception that intrigued me, not the details of exactly how great the effect was.