10.1 Behavioral Finance

Behavioral finance studies how real decision makers systematically depart from the textbook investor. It sits in Theory and History, 38% of CMT Level I in the 2026 Program Guide, because a technician needs a reason that recorded prices can contain more than noise. If crowds mis-weight gains, losses, and news, then trends, overshoots, and failed breakdowns are not mysteries — they are what biased groups do with money. Independent OpenExamPrep material for these CMT Level I topics treats that claim as a hypothesis about people, not as a warranty that a pattern pays after costs, and not as a CMT Association publication.

Two assumptions traditional economic research relied on

For decades, models of markets rested on two working assumptions. CMT Level I expects you to name both, then show how prospect theory breaks them.

1. Rational agents. Decision makers were modeled as having consistent preferences. They ranked outcomes the same way from one sitting to the next, they did not reverse a choice just because a brochure used different words, and they maximized expected utility of final wealth. If a gamble raised expected utility, they took it; if it lowered expected utility, they refused it. Risk aversion entered as a concave utility-of-wealth curve: each extra dollar of wealth added a little less pleasure than the last, so people would pay to avoid a fair coin-flip. The object of choice was terminal wealth, not a gain or loss versus last Tuesday's fill.

2. Efficient information processing. Agents were assumed to use relevant information without systematic bias. They updated beliefs the way Bayes' rule says they should: new evidence shifts the posterior in proportion to how diagnostic it is. They did not cling to an old story, overweight a vivid headline, or treat a round number as a law of nature. Combined with competition, this second assumption underwrote the Efficient Markets Hypothesis taught earlier in Theory and History: if information is processed without bias and traded on, prices fully reflect available information.

Expected-utility theory (von Neumann–Morgenstern, building on Bernoulli) is the math of those two assumptions. A $10,000 hole versus a purchase price and a $10,000 hole versus a high-water mark are the same object if final wealth is the same. Real traders do not experience them as the same object. That gap is the door behavioral finance walks through. The two assumptions were a null model, not a photograph of every desk. Prospect theory and the bias lists only have meaning as departures from that null.

Prospect theory (Kahneman and Tversky)

Daniel Kahneman and Amos Tversky proposed prospect theory (1979; cumulative prospect theory 1992) as a descriptive account of risky choice. Three exam-ready claims:

1. People evaluate outcomes as gains and losses versus a reference point, not as final wealth. The reference can be a purchase price, a peak, an index level, a forecast, or "what I was promised." A stock bought at $50 that prints $48 is a loss even if the holder's lifetime wealth is up.
2. The value function is concave in the domain of gains (risk aversion when sitting on a paper profit) and convex in the domain of losses (risk seeking when sitting on a paper loss). Concavity in gains means the pleasure of going from +$100 to +$200 is smaller than the pleasure of going from $0 to +$100, so people tend to lock in gains. Convexity in losses means people will gamble to break even rather than take a sure smaller loss.
3. The loss side is steeper than the gain side. Losing $100 hurts more than gaining $100 pleases. That asymmetry is loss aversion.

A classroom numbers example: offered a 50/50 chance to win $110 or lose $100, many people refuse even though expected value is +$5. Mild expected-utility risk aversion can refuse some positive-EV bets, but the pattern across many such bets is a steeper loss lobe. Later estimates often put the loss-aversion coefficient near 2 (a figure around 2.25 is widely cited from Kahneman and Tversky's later work): losses carry roughly twice the psychological weight of equal-sized gains. Do not treat 2.25 as a trading constant. Treat "steeper in losses" as the testable shape.

Disposition effect — prospect theory on a chart

The disposition effect (Shefrin and Statman) is the footprint: investors sell winners too soon (concave gains) and hold losers too long (convex losses, gamble back to the reference). Technicians see it as volume when a name "gets back to even," as resistance at obvious cost bases, and as slow declines that refuse to be marked to market. The chart is not a personality quiz. It is a crowd using purchase price as a reference point.

Worked tape: a trader buys 200 shares at $40. Price falls to $32. Expected-utility cares about wealth. Prospect theory codes an $8 loss versus $40. The same trader will often refuse a clean $32 sale and will add risk if a bounce toward $40 looks like a chance to "get out even," which is convexity in the loss domain. A different name bought at $40 that rallies to $52 is more likely to be sold on the first red day, which is concavity in the gain domain. That pair of behaviors is loss aversion plus the S-shaped value function, not a new indicator.

Loss aversion

Loss aversion is the steepness fact named as its own bias. It is both a feature of the prospect-theory value function and an emotional bias in the taxonomy below. In markets it shows up as refusing to sell a loser because the sale would make the loss real; walking a stop farther away after a loss (risk seeking in the loss domain); demanding more compensation to take a short that could gap than a symmetric distribution would require; and reacting harder to a 10% drawdown than to a 10% rally in the same strategy.

Loss aversion is not the same as classical risk aversion. A classically risk-averse expected-utility agent dislikes variance of wealth. A loss-averse agent dislikes being below the reference point more than they like being above it. That is why one person can be conservative with profits and reckless with losses in the same afternoon.

Belief-perseverance biases (cognitive)

Cognitive biases are faulty reasoning. Belief-perseverance biases keep a story in place after the evidence has moved.

Conservatism and representativeness can pull opposite ways. Conservatism is sticky priors (underreaction). Representativeness is over-weighting a vivid resemblance (overreaction to a short sample). Both can be true on different names and horizons, so behavioral finance does not give a single "always fade the news" rule.

Confirmation wrecks process: you draw the trendline that fits the trade you already want. A practical counter is to write the invalidation before the entry. Hindsight poisons backtests and war stories. After a crash, every prior bearish divergence looks prophetic. Before the crash, the same chart had bullish divergences too. Level I wants the name and the mechanism.

Information-processing biases (cognitive)

These are mistakes in how numbers and stories are coded, even when the person is trying to be fair.

Anchoring is why measured-move targets and round numbers attract orders. The number does not have to be correct. It has to be salient. Framing is why a "failed breakdown" and a "bullish reversal bar" can be the same open-high-low-close. Availability is why a once-in-a-decade air pocket dominates a risk meeting more than the boring monthly grind that actually pays the bills. Mental accounting is why a trader will not net a $4,000 loss in Name A against a $4,000 gain in Name B if the two live in different "stories."

Emotional biases

Emotional biases arise from feelings — fear, pride, regret, attachment — rather than from a calculation error. Loss aversion is listed again here because the taxonomy treats the feeling of loss as emotional even though it is also the slope of prospect theory.

Overconfidence in this unit is treated as emotional because it is fused with ego: the feeling of being right. It shows up as "it will tag $72 this week" and as underestimating the chance the setup is noise. Regret aversion produces both omission (I never bought, so I never failed) and commission herding (if we all own it, I will not be the only fool). Endowment and status quo keep dead names in the book.

Why emotional biases differ from cognitive biases

Cognitive biases are errors in thinking. Education, checklists, base-rate tables, pre-commitment of invalidation, and a second reader can reduce them. You can teach someone that three-bar samples are weak evidence (representativeness), that purchase price is a sunk cost (anchoring and mental accounting), or that they must look for disconfirming closes (confirmation).

Emotional biases are errors in feeling. Telling a loss-averse trader that the sale is "just a number" does not make the stomach stop. Education alone is a weak treatment. Useful responses are structural: smaller size, stops entered when the trade is opened, a rule that a loser is sold at a predefined level regardless of the story, delegation of the sell button, or a cooling-off period. CMT Level I does not ask you to become a therapist. It asks you to know why a lunch-and-learn fixes framing more readily than it fixes regret.

For a technician, the practical split is: when the crowd is making a cognitive error (anchoring on a round number, representativeness on a familiar flag), a rules-based plan can be written down. When the crowd is in an emotional regime (panic, endowment of a bubble narrative, status-quo paralysis), a lecture will not quickly reverse the tape. Sentiment tools in the next sections exist because feelings scale to the market.

Do not mix the categories on a stem. Loss aversion is listed with emotional biases even though it is also the slope of prospect theory. Overconfidence, regret aversion, endowment, and status quo are not cured by a lecture on Bayes' rule.

Key Takeaways

• Traditional research: rational expected-utility agents and unbiased information processing
• Prospect theory: gains/losses versus a reference point; concave in gains, convex and steeper in losses
• Loss aversion and the disposition effect (sell winners, hold losers)
• Belief-perseverance: conservatism, confirmation, representativeness, illusion of control, hindsight
• Information-processing: anchoring, mental accounting, framing, availability
• Emotional biases are harder to correct with education than cognitive biases