Input Junkie

In large measure I agree with you. That's a problem with virtually any large scale assertion about society. It's not possible to do controlled experiments so correlation plus a plausible narrative is about the best one can do. Of course, if one wanted to be pedantic about this one could equally validly assert that no scientific hypothesis is provable. I can't prove that smoking causes cancer. It could be that both are caused by some third, independent, factor. Few physicians or biologists (except perhaps those in the pay of big tobacco) would argue that but it can't be disproved. In the original case we were considering, the hypothesis "obesity causes crime" is risible and demonstrably false as it is possible to demonstrate that that correlation is demonstrably false in many instances. The hypothesis "income inequality contributes to a higher crime rate" is at least not contradicted by the evidence. As a lifetime Bayesian I can live comfortably with my subjective prior!