This paper discusses problems that are common to both the epidemiologic risk-factor approach and the demographic variable-based approach to studying population health. We argue that there is a shared reluctance to move away from a narrow variable-based thinking that pervades both disciplines, and a tendency to reify the multivariate linear procedures employed in both disciplines. In particular, we concentrate on the difficulties generated by classical variable-based approaches that are especially striking when one neglects selection processes and the use of strategies to minimize its effects. We illustrate these difficulties in terms of the so-called "Hispanic Paradox", which refers to comparative health advantages that some Hispanic groups appear to have. We find that much of what is conceived by demographers and epidemiologists as a paradox may not be paradoxical at all.