Built to Regress to the Mean
| Peter Klein |
Of 35 “Excellent” companies studied in In Search of Excellence, 30 declined in profitability over the 5 years after the authors’ study ended in 1979. . . . Similarly, of 17 of the 18 “Visionary” companies studied in Built to Last, only 8 outperformed the S&P 500 market average for the 5 years after the authors’ study ended in 1990.
This is from Phil Rosenzweig’s The Halo Effect (Free Press, 2007) (I’m quoting this summary in CFO Magazine). Rosenzweig’s book reads like a primer on research methods for producers (and consumers) of popular management literature. Rosenzweig, a management professor at IMD, explains the problem of selection bias, the difference between correlation and causality, the need to compare rival explanations, the difference between absolute and relative performance, and more.
“Some of what I talk about in The Halo Effect is Research Design 101,” Rosenzweig tells CFO. “You gather your independent variables, independently of the thing you’re trying to explain. You don’t confuse correlation with causality, and you don’t confuse ends with means. You control for other variables. It’s basic stuff.”
But that basic stuff is hard to translate into a BusinessWeek best-seller.
Thanks for the link to Gary Peters, who notes that the book might be good reading for a doctoral seminar on research methods.