The author claims to have done that to account for the relative age differences between the women (generally younger models) and the men (generally older "successful" men) featured in the magazine. If that is the case (and not just an excuse) then you might expect a bearded man to be of the same generation as a wide-skirted woman that appeared 21 years earlier.

Even if you didn't shift the graphs, though, the data would still be striking. It'd likely remind me of a predator-prey graph (e.g. http://www2.estrellamountain.edu/faculty/farabee/biobk/graph... ) which is "easily" modeled with differential equations.

This is a completely valid statistical approach. The author isn't hiding anything, the graph is clearly labeled.

http://en.m.wikipedia.org/wiki/Cross-correlation