Note the title of the article above: high crime raises diabetes. My University of Colorado Boulder colleague Jason Boardman was more careful in his interpretation than his publicist. What the study actually shows is that genes associated with diabetes overall cause diabetes for people who live in high crime areas, but don’t clearly cause diabetes for people who don’t live in high crime areas.

Except for the issue of whether it is one’s own genes causing something or the corresponding genes in one’s parents (or less likely the same genes in one’s other relatives) causing something, the randomization of genes at conception makes it reasonably clear that when genes are correlated with something that the genes are causing that. In the future, when we have data with genes for father, mother and child, with the focus on what happens to the child, the evidence for genetic causes will be as strong as evidence from randomized controlled trials.

Of course, genes interact with the environment: they have different effects in different contexts. The causal chain from genes to outcomes that social scientists are interested in is typically a long one, so all kinds of things along the way modify what the ultimate effect of the genes is. Jason Boardman has very interesting evidence that genes that on average cause diabetes are doing most of their work by causing diabetes among people who live in high crime areas.

But this is NOT evidence that high crime causes diabetes. It is not even evidence that high crime causes diabetes among those who have a genetic vulnerability for high crime to cause diabetes. High crime might cause diabetes among those with a genetic vulnerability for such an effect; stress could be a causal pathway. But it is just as possible that other factors that tend to be associated with high crime in an area cause diabetes among those with a genetic vulnerability to such an effect. For example, places with high crime are often food deserts, in which it is much easier to find junk food than to find healthy food (such as nonstarchy vegetables—one of the few categories almost everyone agrees is healthy). And so the evidence may be showing that food deserts cause diabetes among those who are genetically vulnerable to such an effect.

There is an interesting possibility in which high crime could cause diabetes among those genetically vulnerable, not through stress but because high crime causes food deserts as storekeepers avoid high crime areas, and food deserts in turn cause diabetes. That is, even when it is fully legitimate to say that high crime causes diabetes among those genetically vulnerable, it might not be the causal chain that most easily comes to mind when one says that.

When looking for remedies, it matters what the causal chain is. If food deserts are an important part of the causal chain, the remedies one would look for would be different than if stress is the key part of the causal chain. So getting statistical interpretations right matters.

As a caveat, let me say that my discussion of statistical interpretation here is based only on information in the news article. The academic paper might have good counterarguments to the possibilities I raise. But if so, the journalist should have emphasized those counterarguments, since they seem crucial to backing up the title of the piece.