statistics example: treatment effect vs. selection effect

Malcolm Gladwell, author of Blink and The Tipping Point, has an article about the evolution of admissions at elite colleges in a recent issue of the New Yorker. Here is a salient distinction between two kinds of influences on experimental outcome:

Social scientists distinguish between what are known as treatment effects and selection effects. The Marine Corps, for instance, is largely a treatment-effect institution. It doesn’t have an enormous admissions office grading applicants along four separate dimensions of toughness and intelligence. It’s confident that the experience of undergoing Marine Corps basic training will turn you into a formidable soldier. A modelling agency, by contrast, is a selection-effect institution. You don’t become beautiful by signing up with an agency. You get signed up by an agency because you’re beautiful...

The nut of the discussion is: Do people who go to Ivy League schools make more money because of their educations or because of other factors? The correct way to figure this out is not to compare graduates but rather to begin on common ground, considering people who were accepted at both elite and non-elite institutions, and to see whether their lives after university were substantially different. Guess how that turned out...