This section covers the systematic, predictable ways human judgment departs from the rational-actor model assumed by classical finance — the psychological errors that make real investors over- and under-react, hold losers, chase winners, and crowd into the same trades. The field's organizing distinction, formalized in Michael Pompian's widely taught taxonomy, is between cognitive errors (faulty reasoning and information-processing — anchoring, confirmation bias, availability/recency — which are at least partly correctable with better data, structure, and discipline) and emotional biases (rooted in feelings such as fear, regret, and arrogance — loss aversion, overconfidence — which are harder to debias because they protect the self rather than process information). The core tension running through every node below is the same: each bias began as a useful heuristic — a fast, low-effort shortcut that lets a person act under uncertainty — and becomes a liability only when it is applied where it doesn't fit. None of these is a pathology of the foolish; they are documented in professionals, Nobel laureates, and the authors who named them. The job of this branch is not to "predict the market" but to supply a diagnostic vocabulary for where human judgment — the user's, the crowd's, and the analysis system's own — is likely to be miscalibrated.

What this section covers (the map)

The six sub-topics are not a random list; they form a rough pipeline of how a flawed decision gets made and then defended:

• Anchoring & Adjustment — judgments get tethered to a salient but often irrelevant reference point (purchase price, 52-week high, a round number, last quarter's figure) and adjust too little from it. The entry bias: it sets the reference frame before any reasoning happens. (Tversky & Kahneman 1974; George & Hwang's 52-week-high anomaly.)
• Confirmation Bias — once a view is held, supporting evidence is sought and weighted as signal while disconfirming evidence is treated as noise. The defense bias: it converts a falsifiable thesis into an unfalsifiable conviction and protects the other biases from correction. (Wason 1960; Nickerson 1998.)
• Loss Aversion & Disposition Effect — losses are felt more sharply than equivalent gains (prospect theory), producing the best-documented market footprint: selling winners too early and riding losers too long. The exit bias. (Kahneman & Tversky 1979; Odean 1998.)
• Recency & Availability Bias — probability is judged by how easily examples come to mind, so recent and vivid events get overweighted; the engine of performance-chasing and return extrapolation. (Tversky & Kahneman 1973; Greenwood & Shleifer 2014.)
• Overconfidence — subjective confidence systematically exceeds objective accuracy; the cause of trading too much, underdiversifying, and undersizing risk. Decomposes into overestimation, overplacement, and overprecision. (Moore & Healy 2008; Barber & Odean 2000.)
• Herding — the social bias, where the error propagates between agents rather than within one mind: imitating the crowd over one's own information, via informational cascades and career/reputational concern. The microfoundation of bubbles and crashes. (Banerjee 1992; Bikhchandani & Sharma 2001.)

Anchoring, confirmation, and availability/recency sit on the cognitive side; loss aversion and overconfidence are usually classed as emotional; herding spans both (it can be a rational reputational calculation and an emotional FOMO response). These categories blur in practice — the boundaries are a teaching device, not a clean partition.

When it matters — and when it doesn't

These biases bite hardest when the quantity is uncertain, the decision is under time pressure or emotion, feedback is slow or noisy, and the actor is inexpert or unincentivized for accuracy. They matter much less where outcomes are fast, frequent, and clearly scored (which is exactly why calibration training and trade journals work) and among sophisticated, rules-bound institutions, which empirically show several of these effects far more weakly than retail. A second boundary is aggregation: a bias that is small in one head can dominate price when a whole market shares the same prior, which is why herding, recency, and overconfidence are the engines of bubble/panic dynamics. The honest limit is that naming a bias is descriptive, not predictive — knowing the crowd is overconfident or anchored tells you little about when a move will reverse.

Adoption, debate & evidence (field-level)

Behavioral finance moved from heterodox to mainstream over four decades — Kahneman's 2002 and Thaler's 2017 economics Nobels mark its acceptance, and the cognitive/emotional bias catalog is now standard in the CFA curriculum and most behavioral-finance texts (Pompian, Montier). But the corpus must hold two honest fault lines that the children detail:

1. Existence vs. magnitude. The lab phenomena (anchoring, availability, prospect-theory asymmetry) replicate strongly, but several headline magnitudes are contested. Loss aversion's textbook λ ≈ 2.25 comes from one small 1992 study; the best meta-analytic estimate is nearer ~1.96 (Brown et al. 2024), and a 2025 re-analysis argues the aggregate effect is fragile. Anchoring effect sizes are similarly debated as inflated by underpowered designs. Treat these as stylized averages, never personal constants. 2. Bias vs. rational behavior. Whether a behavior is an irrational bias or a sensible heuristic is genuinely disputed for several: confirmation bias may be a rational positive-test strategy (Klayman & Ha 1987); much measured herding may be efficient shared reaction to news, not imitation (Bikhchandani & Sharma 2001); some performance-chasing reflects real time-varying risk premia. The behavioral regularities (the disposition effect, overtrading's cost, return extrapolation) are far better established than the mechanisms claimed to cause them — most mechanism attributions are model-based inferences, not proven from P&L.

The recurring caveat across all six: an observed bias is rarely a tradable edge net of costs, and it is dangerously easy to "explain" any outcome post hoc as a bias (an unfalsifiable trap).

Sources

• Pompian, M. (2012), Behavioral Finance and Wealth Management — cognitive-error vs. emotional-bias taxonomy and behavioral investor types: Breaking Down Finance; michaelpompian.com.
• CFA Institute behavioral-biases reading (cognitive vs. emotional classification) — summarized at CliffsNotes.
• Primary-source attributions, evidence, and dispute flags are in each child node (Tversky & Kahneman 1973/1974/1979; Kahneman & Tversky 1992; Shefrin & Statman 1985; Odean 1998; Barber & Odean 2000/2001; Moore & Healy 2008; Banerjee 1992; Bikhchandani & Sharma 2001; Greenwood & Shleifer 2014; Brown et al. 2024).

Section-level dispute flags: (1) the cognitive/emotional split is a teaching device, not a clean partition — many biases span both. (2) Several headline magnitudes (loss-aversion λ, anchoring effect size) are contested and treated as stylized, not constants. (3) For several biases the behavioral regularity is well-established but the rational-vs-irrational interpretation is genuinely disputed — see the individual nodes.