I wrote about Nancy Krieger’s insightful American Journal of Public Health paper in a previous post. In this second of three posts, I will continue to unpack some of the content of her article, focusing on the distinction between correlation and causation.
Krieger’s paper is titled “Health Equity and the Fallacy of Treating Causes of Population Health as if They Sum to 100%.” In the first post in this series, Sherry Glied and I explained why this is a fallacy. Here, I want to draw your attention to the fact that Krieger’s title uses “causes” and I’m using “risk factors.” As she notes,
[T]the percentage of variation in outcomes that is ‘explained’ by particular factors is not equivalent to the proportion of risk causally attributable to these factors.
She is, and we are, well aware that correlation is not causation. Causal interpretations of risk factor associations is not always warranted.
For example, across the population, having relatively smaller feet is correlated with (a risk factor for) slower reading speed. That’s because, on average, children have both smaller feet and read more slowly than adults. We know this isn’t causal. We could equally say that slower reading is a risk factor for smaller feet. That’s true, but it’s no more causal.
The other point she’s making is that explaining variation and estimating effect sizes are different. I will return to this in a subsequent post. It’ll be a little mathy, but I’ll try to keep it simple.