Regression to the Mean
mental-model
Source: Probability
Categories: cognitive-sciencesystems-thinking
From: Poor Charlie's Almanack
Transfers
A statistical phenomenon mapped onto prediction and judgment. Regression to the mean is the observation that if a variable is extreme on its first measurement, it will tend to be closer to the average on its second measurement — not because of any causal force pulling it back, but because extreme values are statistically unlikely to be repeated. The model reframes a wide range of apparent patterns as statistical artifacts rather than real effects.
Key structural parallels:
- Extreme performance is partly luck — any measured outcome is a combination of skill (or underlying tendency) and randomness. When someone performs exceptionally well or badly, the random component was probably favorable or unfavorable. On the next measurement, the random component is likely closer to zero, so the measured outcome moves toward the mean. This is not a force; it is arithmetic. The model teaches you to decompose observed outcomes into signal and noise before attributing causation.
- The illusion of effective intervention — a manager reprimands a poor performer, and the next quarter their numbers improve. A doctor prescribes a treatment when symptoms peak, and the patient gets better. In both cases, regression to the mean would have produced improvement without the intervention. The model reveals that we systematically overestimate the effectiveness of actions taken at extreme moments, because the natural regression provides a false confirmation.
- Punishment seems to work better than reward — Kahneman’s famous observation: pilots reprimanded after bad landings improve; pilots praised after good landings get worse. Both are regression effects, but the asymmetry creates a persistent illusion that criticism is more effective than praise. The model explains a deep bias in management and education toward punishment, grounded not in psychology but in statistics.
- The sophomore slump is predictable — in sports, business, and the arts, exceptional debut performances are reliably followed by less impressive second efforts. This is not because success breeds complacency (though it might); it is because the debut performance was partly lucky, and the luck is unlikely to be repeated. The model predicts the pattern without needing any psychological explanation.
Limits
- Not everything regresses — regression to the mean applies to variables with a random component measured repeatedly. It does not apply to deterministic processes, to first measurements of a fixed quantity, or to systems undergoing genuine structural change. A company whose earnings spike because it entered a new market may sustain the spike indefinitely. Treating all extreme outcomes as regression candidates is as much an error as ignoring regression entirely.
- The mean itself can shift — the model assumes a stable underlying distribution. But in many real systems, the mean is moving. A student whose test scores jump may have genuinely learned something; the new mean is higher. Climate temperatures that hit new extremes are not regressing to a fixed mean — the mean is rising. Applying the regression model to non-stationary processes produces false comfort.
- It can excuse inaction — if extreme outcomes are just noise that will self-correct, why intervene? The model can become a rationalization for doing nothing. A manager who dismisses a spike in customer complaints as “regression to the mean” may be ignoring a genuine systemic problem. The model tells you not to over-react, but it provides no clear threshold for when a signal is real rather than noise.
- Psychological reality matters even if it is statistically irrelevant — the model correctly identifies that the “sophomore slump” is partly a statistical artifact. But it is also psychologically real: people who experience public success do face new pressures (expectations, scrutiny, loss of underdog motivation) that may independently reduce performance. Dismissing the psychology because the statistics already explain the pattern is reductive.
- The model requires knowing the distribution — to predict regression, you need to know the mean, the variance, and that the variable is at least partly random. In practice, for most interesting real-world variables, you do not know any of these with confidence. The model offers clear predictions in textbook examples and murky guidance in real decisions.
Expressions
- “Regression to the mean” — Galton’s original term, now used broadly in statistics, sports, business, and everyday reasoning
- “Reversion to the mean” — the more common variant in financial contexts, applied to stock prices, earnings, and market valuations
- “The sophomore slump” — the pop-culture version, applied to athletes, musicians, and directors after exceptional debuts
- “Don’t confuse correlation with causation” — not identical, but the regression fallacy is one of the most common mechanisms by which spurious correlations arise
- “Things will even out” — the folk version, which is approximately correct for regression but dangerously wrong when mistaken for the gambler’s fallacy (confusing statistical tendency with compensating force)
Origin Story
Francis Galton discovered regression to the mean in the 1880s while studying the heights of parents and children. He noticed that exceptionally tall parents tended to have children shorter than themselves (though still above average), and exceptionally short parents tended to have taller children. He called it “regression toward mediocrity,” a term that carries an unfortunately judgmental tone for a purely statistical phenomenon. The concept entered psychology through Kahneman and Tversky’s work on judgment under uncertainty in the 1970s, where they demonstrated that failure to recognize regression effects was one of the most robust cognitive errors in human reasoning. Munger absorbed it as a core mental model, particularly for investing: apparent “turnarounds” in company performance are often just regression, and the wise investor accounts for this before attributing improvement to new management.
References
- Galton, F. “Regression Towards Mediocrity in Hereditary Stature” (1886)
- Kahneman, D. Thinking, Fast and Slow (2011), ch. 17 — the definitive modern treatment of regression fallacies
- Kahneman, D. & Tversky, A. “On the Psychology of Prediction” (1973)
- Munger, C. “The Psychology of Human Misjudgment” (1995), collected in Poor Charlie’s Almanack (2005)
- Secrist, H. The Triumph of Mediocrity in Business (1933) — the notorious case study in confusing regression with a causal process
Related Entries
Structural Neighbors
Entries from different domains that share structural shape. Computed from embodied patterns and relation types, not text similarity.
- Let the Buyer Beware (economics/mental-model)
- Loved One Is A Possession (economics/metaphor)
- Mental Accounting (economics/metaphor)
- Reciprocity (economics/mental-model)
- Responsibilities Are Possessions (economics/metaphor)
- Temperature Is Creativity (physics/metaphor)
- Odyssey (mythology/metaphor)
- Working Through (manual-labor/metaphor)
Structural Tags
Patterns: balancescaleiteration
Relations: causetransform
Structure: cycle Level: generic
Contributors: agent:metaphorex-miner