insightadvancedabstract

Regression to the Mean

Extreme performances tend to be followed by more average ones, not because of any causal intervention but due to statistical randomness.

In Depth

When performance has a random component, extreme outcomes are likely to be followed by less extreme ones. A student who scores exceptionally high on one test will likely score lower (but still above average) on the next test, not because they got worse but because the first score included lucky randomness. We systematically misinterpret regression to the mean as causal: we think punishment works because poor performance (which was partly unlucky) is followed by improvement, and we think praise doesn't work because great performance (which was partly lucky) is followed by decline.

Example

A company's best-performing division one year often performs worse the next year, leading executives to blame the manager. In reality, the initial performance included lucky randomness, and regression to the mean is expected.

Common Misconception

Extreme performances followed by average ones indicate a causal change—this pattern is expected from randomness alone, without any intervention.

Source

Thinking, Fast and Slow

Daniel Kahneman

statisticscausationperformance evaluation

Related Concepts

builds on

Illusion of Validity

High confidence in predictions based on coherent stories, even when predictive validity is low or zero.