Behavioral Finance

Behavioral finance is the study of how psychological factors shape financial decisions and financial markets. The Royal Swedish Academy of Sciences, in announcing Richard Thaler's 2017 prize, describes it as the field "which studies how cognitive limitations influence financial markets," and names Thaler as one of its founders. It sits inside the broader discipline of behavioral economics: Thaler's prize was awarded "for his contributions to behavioural economics," and behavioral finance is the part of that discipline concerned with saving, investing, borrowing, and asset prices. Standard economic models had long assumed a decision-maker who is self-interested and capable of rational calculation. Behavioral finance replaces that assumption with evidence about the choices people actually make.

The field's origin is a single paper. Daniel Kahneman and Amos Tversky published "Prospect Theory: An Analysis of Decision under Risk" in Econometrica in 1979, proposing a model of choice under uncertainty built to match observed behaviour rather than to describe ideal behaviour. Kahneman received the 2002 prize in economic sciences "for having integrated insights from psychological research into economic science, especially concerning human judgment and decision-making under uncertainty"; Tversky had died in 1996. Thaler's 2017 prize recognised work extending the programme into economics and policy, including mental accounting, the endowment effect, fairness, self-control, and the nudge, a term he coined.

What makes the field useful rather than merely deflating is the word systematic. The 2002 prize announcement records that Kahneman "demonstrated how human decisions may systematically depart from those predicted by standard economic theory," and that human judgment takes heuristic shortcuts that likewise depart systematically from the rules of probability. Random error cannot be designed around; a bias that pushes almost everyone in the same direction can be. A retirement plan that knows its participants will overwhelmingly stay in whatever option they are placed in can choose that option carefully.

The findings are easier to hold onto in families than as a list of several dozen named effects. There is no official taxonomy of biases, and the grouping that follows is a convenience for readers rather than a standard classification. Errors of valuation concern how an outcome is scored: loss aversion, the endowment effect, mental accounting. Errors of evidence concern which information gets weight: confirmation bias, recency bias, the availability of vivid examples. Errors of confidence concern how much people trust their own judgment, including overconfidence and hindsight. Errors of timing concern how the future is discounted against the present: present bias, hyperbolic discounting. Errors of inertia concern what happens when no active decision is made at all: status quo bias, the default effect. A single real decision often involves several families at once, which is why "which bias was that" is usually the wrong question and "which family of error is this decision exposed to" is the useful one.

Two limits belong on the page alongside the findings. The field is descriptive, so it identifies where behaviour diverges from a model and does not supply anyone's goals or values. And the magnitudes are experimental estimates that move with stakes, framing, and population, so an effect measured in a laboratory is a reliable direction rather than a constant.

The method is straightforward and is what separates the field from folk psychology. Work out what a rational-agent model predicts for a specific choice, put that choice to people under controlled conditions, and measure the gap. Where the gap is large, repeatable, and points the same way across populations, it becomes a documented effect with a name. Applied work then changes the setting in which the decision is made, rather than urging people to try harder, and measures whether behaviour moves.

An illustration of the method, using an experiment reported in the Royal Swedish Academy's scientific background for the 2002 prize. Two options are offered, both involving losses. Option A is a 25% chance of losing $6,000 and a 75% chance of losing nothing. Option B is a 25% chance of losing $4,000 and a 25% chance of losing $2,000, with a 50% chance of losing nothing. The expected loss is identical: Option A works out to 0.25 × $6,000 = $1,500, and Option B to (0.25 × $4,000) + (0.25 × $2,000) = $1,000 + $500 = $1,500. Option A is simply a wider spread around the same average, so conventional risk aversion predicts it would not be preferred. Kahneman and Tversky found that seven out of ten people did prefer it. The finding is not that the participants were foolish. It is that a specific, reproducible feature of how losses are evaluated made the gamble with the bigger possible loss more attractive, and any model that assumes otherwise will mispredict what real savers do.

Pros

• Gives a shared, specific vocabulary for behaviour that would otherwise be dismissed as carelessness, which makes it discussable and correctable.
• Because the deviations are systematic, they can be anticipated in advance and designed around rather than merely regretted afterwards.
• Shifts the practical remedy from willpower to structure, which is why automatic enrollment, automated saving, and written rules outperform good intentions.
• Encourages appropriate humility about one's own judgment, including about investment decisions made under stress.

Cons

• It is descriptive rather than prescriptive. Identifying a bias says nothing about what a person should want or do.
• Naming a bias does not neutralise it. Awareness is a weak intervention compared with changing the decision environment.
• Effect sizes are experimental estimates that vary with stakes, framing, and population, so they should be treated as reliable directions rather than fixed constants.
• The vocabulary is easy to misuse. "That's just loss aversion" can dismiss a well-founded objection, and behavioral language can decorate a recommendation without improving it.