Jeffrey Pfeffer and Robert I. Sutton, Hard Facts, Dangerous Half-Truths & Total Nonsense: Profiting from Evidence-Based Management (Harvard Bus. School Press 2006) state at 91 that while “intelligence is the most powerful predictor of job performance,” that predictor seldom correlates more than 0.4 with performance (See my post of Jan. 14, 2006 on the best techniques to predict the performance of new employees.). When law departments try to assess whether to hire a lawyer, IQ might be toward the top end of correlation. The proxies for IQ are to some degree undergraduate and law schools attended and academic performance at those schools.
What I found even more revelatory is the author’s statement that the amount of variation explained by a predictor – here, intelligence test scores to predict job performance – is the correlation squared. Thus, intelligence accounts for no more than 16 percent of the variation in performance, leaving 84 percent unexplained. More broadly, whenever a statistical correlation is given, it is easy to figure out how much of the independent variable (here, IQ) contributes to the dependent variable (here, job performance). Sometimes its good to be square.