a few months ago, i read stephen levitt's freakonomics, which brought me much frustration. it doesn't help that i have a conceptual aversion to the dominance of systems-level inferences about their components, be they social aggregate statistics of behaviors of individuals or ensemble properties of molecules. largely, it's that levitt's approach is meta-physical, drawing inter-generational inferences of the effect of abortion in the 1970s on the rates of crime in the 1990s.
now, it seems that the statistical significance of the original analysis may be lacking altogether (Foote & Goetz). related story in the wall street journal.