The study talks of rate of subsequent cognitive decline post retirement. There's no fixed state, which once reached, people start retiring. I think the study does control for reverse causality. It is not a correlational study.