In one sentence
An anomaly is a systematic pattern in data or behavior that contradicts a widely used benchmark model, pushing economists to refine the model or the measurement.
Historical Context
The study of anomalies gained prominence in the late 20th century as economists began to notice patterns in decision-making that conflicted with the expected utility theory and other well-established models. Notably, the work of psychologists Amos Tversky and Daniel Kahneman on cognitive biases laid much of the groundwork for understanding economic anomalies.
What counts as an “anomaly”
In practice, economists call something an anomaly relative to a specific benchmark, such as:
- expected utility under risk,
- rational expectations,
- perfect competition with full information,
- efficient markets (in finance),
- frictionless consumption-smoothing models.
An anomaly can be resolved by:
- changing preferences (e.g., loss aversion),
- adding frictions (search costs, liquidity constraints),
- adding information problems (adverse selection, moral hazard),
- improving measurement (risk adjustment, missing variables, data revisions).
Canonical examples
Some widely discussed anomalies/puzzles include:
- Allais paradox / certainty effect: systematic violations of expected utility’s independence axiom.
- Present bias and under-saving: short-run impatience inconsistent with exponential discounting.
- Disposition effect: selling winners too early and holding losers too long.
- Equity premium puzzle: historically high equity returns relative to safe rates are hard to match in baseline models.
- Momentum / value effects (finance): return patterns not fully explained by simple risk factors.
flowchart LR
A["Benchmark model<br/>(e.g., expected utility, EMH)"] --> B["Observed pattern contradicts model"]
B --> C{"Interpretation"}
C -->|Preferences| D["Behavioral mechanisms<br/>(loss aversion, weighting)"]
C -->|Frictions| E["Constraints/market structure<br/>(liquidity, search, limits to arbitrage)"]
C -->|Measurement| F["Risk adjustment / data issues"]
D --> G["New predictions and tests"]
E --> G
F --> G
A key caution: data-mining and publication bias
Some “anomalies” shrink or disappear after:
- adjusting for multiple testing,
- using out-of-sample validation,
- accounting for transaction costs and liquidity,
- revising measurement choices.
This is why modern empirical work emphasizes replication, pre-analysis plans (in experiments), and robustness checks.
Related Terms with Definitions
- Cognitive Bias: Systematic patterns of deviation from norm or rationality in judgment.
- Expected Utility Theory: A theory that assumes individuals choose options with the highest expected utility.
- Prospect Theory: Describes how people make decisions in situations of risk and uncertainty, highlighting deviations from the expected utility theory.
- Allais Paradox: A situation that contradicts the expected utility hypothesis by showing that people’s choices violate the independence axiom.
- Limits to Arbitrage: Reasons mispricing can persist (risk, costs, funding constraints), even with rational traders.
- Publication Bias: The tendency for statistically significant results to be published more often than null results.