Regression to the Mean

Definition

When a result is unusually good or bad, a large share of that extremity is typically due to chance rather than a stable underlying trait, so the next attempt — with less lucky chance involved — will tend to be closer to average. Kahneman states his favorite formula for it as "success = talent + luck; great success = a little more talent + a lot of luck." The persistent human error is to attach a causal story (my praise caused the decline, my punishment caused the improvement) to what is really "the inevitable fluctuations of a random process."

In the Book

The chapter's founding anecdote is Kahneman's "eureka" moment teaching Israeli Air Force flight instructors: he argued reward beats punishment for skill training, and a veteran instructor objected that cadets praised for a good maneuver typically did worse next time, while cadets berated for a bad one typically improved — "proof," the instructor felt, that punishment works and reward doesn't. Kahneman shows this is regression to the mean: a cadet praised was probably praised because that attempt was luckily good, and would likely have regressed toward his own average regardless of praise; the same logic runs in reverse for the punished cadet. He demonstrates the same pattern live by having officers throw coins at a target with their backs turned — the best first throws were mostly followed by worse ones, the worst by better ones, with zero coaching involved. He extends this to a broader diagnosis of "perverse" feedback in life generally: because people tend to praise others after good performance and criticize after bad, "we are statistically punished for being nice and rewarded for being nasty" — an illusion that keeps reinforcing itself.

Why It Matters

Regression to the mean is one of the most consequential statistical facts routinely misread as causal, because the mistaken causal story (punishment works, my intervention fixed it, this trend will continue) is always available and always satisfying. It applies anywhere performance has a random component layered on a stable trait: sophomore slumps, "curse of Sports Illustrated" covers, a manager's coaching intervention that "worked" right after an employee's unusually bad month, or a business unit's improvement after a leadership shakeup. Recognizing it means asking, before crediting any intervention, whether the baseline was already likely to move back toward average on its own.