Causality: how to compute causal effects with linear regression. It turns out the estimation is very simple, but the data must possess very specific properties, and we must know about those properties. These two requirements are often referred to as "identifying assumptions".
Usually we do not know if they are satisfied.
Comments