Let’s say that you think that blueberry scones specifically are causing the obesity crisis. You collect two pieces of data very carefully: how many blueberry scones are available in each county, and how many obese people in each county report eating blueberry scones at least once a week. You discover damning evidence: the higher the number of blueberry scones in a given county, the higher the number of obese people who reported eating blueberry scones. Ergo, blueberry scones are directly and positively correlated with obesity, and you stand by your position that we’d have far less obesity if the scones were banned.
But… wait a second.
I mean… of course those things would be correlated, right? And you’re asking the totally wrong question. If you want to make the case that blueberry scones are driving the obesity crisis, then you need to be comparing the availability of blueberry scones with obesity in general. If people in low-scone counties are just as obese but simply getting that way from eating different foods, then scones clearly aren’t the issue (or at least, not alone). It’s even possible that low-scone countries have higher rates of obesity in general, because maybe there’s some other food that’s even worse for you that people eat more of without scones as an alternative.
The point is that it’s a weird take to claim that scones are associated with scone-related obesity as if that was some sort of gotcha. Cars are associated with vehicular deaths too, but unless the total number of deaths are higher in areas with more cars, then you don’t know the whole picture. If you took an island nation and removed all vehicles of any kind, you’d have fewer fatalities from car accidents. But you’d also have more people dying because they couldn’t get prompt medical attention. Do I know that those numbers would wash out? Nope. But I wouldn’t make the claim that they don’t without checking – and measuring the right things.
The Opportunity Machine 