Prospect Theory

Abstract

Prospect theory is a descriptive account of how people choose among risky options, developed by Daniel Kahneman and Amos Tversky to replace the normative expected-utility model where the two diverge. Its central claim is that people evaluate outcomes as gains and losses relative to a reference point rather than as final wealth, that the value function is concave for gains, convex and steeper for losses, and that objective probabilities are transformed by a nonlinear weighting function overweighting rare events and underweighting common ones. These departures reproduce robust anomalies—the reflection effect, the fourfold pattern, framing effects and loss aversion—that expected-utility theory cannot accommodate. This article states the theory, works through its value and weighting functions with interactive demonstrations, and reviews the evidence, the 1992 cumulative revision, and the debate over how universal loss aversion is.

Keywords: prospect theory, loss aversion, reference dependence, probability weighting, framing

Expected-utility theory, axiomatized by von Neumann and Morgenstern and long the benchmark for rational choice under risk, holds that a decision maker should evaluate a gamble by the utilities of its possible final wealth states, each weighted by its probability. As a normative standard it is compelling; as a description of what people actually do it fails systematically. Kahneman and Tversky (1979) assembled these failures into a coherent alternative and named it prospect theory, a deliberately neutral coinage for a theory of choice among prospects—gambles with stated outcomes and probabilities. The theory is descriptive, not prescriptive: it predicts the choices people make, including the ones a normative theorist would call mistakes. Its influence reached beyond psychology into economics, finance, law and medicine, and it was the work cited when Kahneman received the 2002 Nobel Memorial Prize in Economic Sciences (Barberis, 2013).

## What Prospect Theory Is

Prospect theory replaces the single utility-of-wealth function of the normative model with three psychological components. The first is reference dependence: an outcome is not a final asset position but a change—a gain or a loss—measured against a reference point, usually the status quo but sometimes an expectation, an aspiration, or the way a problem is framed. The second is diminishing sensitivity: just as the difference between darkness and a candle is more striking than the difference between two floodlights, the subjective difference between winning \$100 and \$200 exceeds that between \$1,100 and \$1,200. This makes the value function concave over gains and, by the same logic applied downward, convex over losses. The third is loss aversion: the function is steeper on the loss side, so the pain of losing an amount exceeds the pleasure of gaining it.

To these Kahneman and Tversky (1979) added a treatment of probability. Rather than multiplying outcomes by their stated probabilities, the decision maker multiplies the value of each outcome by a decision weight derived from, but not equal to, its probability. The overall value of a prospect that yields outcome x with probability p and y with probability q is V = π(p)v(x) + π(q)v(y), where v is the value function and π the weighting function. Evaluation is preceded by an editing phase in which the prospects are simplified—outcomes coded as gains or losses around the reference point, shared components cancelled, probabilities rounded—so that the object finally evaluated may differ from the one presented.

## The Value Function

The value function v carries reference dependence, diminishing sensitivity and loss aversion in a single curve. It passes through the reference point at the origin, rises with a concave bend over gains and falls with a convex bend over losses, and is markedly steeper just below the origin than just above it. Tversky and Kahneman (1992) gave it the parametric form v(x) = x^α for gains and v(x) = −λ(−x)^β for losses, and estimated from choice data the median values α = β = 0.88 and λ = 2.25. The exponent below one produces the concave–convex shape; the coefficient λ above one is loss aversion, the factor by which the loss limb is steeper than the gain limb.

The demonstration below exposes both parameters. Reducing the curvature exponent deepens the bend in each limb—more diminishing sensitivity—while raising λ stretches the loss limb downward. At the default settings a gain of 100 registers as +57.5 units of value while an equal loss registers as −129.5, the 2.25 ratio made visible.

The kink at the origin is the theory's signature. Because the slope on the loss side exceeds the slope on the gain side at the reference point, a fair fifty-fifty bet to win or lose the same amount is unattractive: the prospective loss is weighted more heavily than the prospective gain even though the two are objectively symmetric. This single feature, loss aversion, does much of the theory's explanatory work.

## Probability Weighting and the Fourfold Pattern

People do not treat probabilities linearly. The weighting function π is regressive—it overweights small probabilities and underweights moderate-to-large ones—and it is steepest near the endpoints of zero and one. Overweighting of rare events explains why the same person will pay for both a lottery ticket, chasing an improbable gain, and an insurance policy, guarding against an improbable loss; underweighting of near-certainties explains the certainty effect, the disproportionate appeal of an outcome that is guaranteed rather than merely very likely. That effect was first exposed by Allais (1953), whose paradox remains the founding anomaly of the field: most people prefer a certain large sum to a gamble of higher expected value carrying a small chance of nothing, yet when the same two options are each made uncertain—so that the safe option loses its guarantee—the majority preference reverses, a switch expected-utility theory forbids because the common component stripped from both options should leave the ranking untouched. Prospect theory reproduces the reversal without special pleading: the weighting function's steepness near probability one lends the guaranteed outcome an extra pull that simply evaporates once it too becomes merely probable. The cumulative version of the theory replaced the outcome-by-outcome weighting of the 1979 model with a rank-dependent scheme that weights cumulative probabilities, which removed the earlier version's occasional violations of stochastic dominance and let the theory handle prospects with many outcomes (Tversky & Kahneman, 1992).

Combining a nonlinear value function with a nonlinear weighting function produces the fourfold pattern of risk attitudes: risk aversion for probable gains and improbable losses, and risk seeking for improbable gains and probable losses. The pattern is not a quirk of a few problems but a systematic prediction, and it is what makes the theory testable. The demonstration lets the two ingredients be dialled independently so the cell each combination lands in can be read off directly.

Table 1 sets out the four cells and the everyday behaviour each one predicts, with the illustrative choices Tversky and Kahneman (1992) used to establish the pattern.

The loss side carries a second signature, the reflection effect: a preference measured over gains typically reverses when the identical outcomes are recast as losses. A person who prefers a sure \$3,000 to an 80% chance of \$4,000—risk averse over gains—will often prefer an 80% chance of losing \$4,000 to a sure loss of \$3,000, becoming risk seeking over losses. The convexity of the value function on the loss side, not any change in the person, drives the switch.

## Framing and Reference Dependence

If value is defined over changes from a reference point, then whatever fixes the reference point governs the choice—and the reference point can be set by nothing more than the wording of the problem. Tversky and Kahneman (1981) demonstrated this with the Asian disease problem: two groups chose between programs to combat a disease expected to kill 600 people, one group seeing the outcomes framed as lives saved and the other as lives lost. The two framings were numerically identical, yet the gain frame produced a majority for the sure program and the loss frame a majority for the gamble—a framing effect, and a direct violation of the description invariance that any normative theory requires. The demonstration reproduces the design with an original scenario, presenting both frames in sequence so the equivalence is easier to catch than in the original between-groups study.

Framing effects are not confined to the laboratory. The same surgery described as offering 90% survival or 10% mortality draws different choices from patients and physicians; a price difference labelled a cash discount is accepted where the same gap labelled a credit-card surcharge is resisted (Kahneman & Tversky, 1984). Because the reference point is often left implicit, the person choosing rarely notices that a reframing has moved it. Kőszegi and Rabin (2006) later gave the reference point a theoretical foundation of its own, modelling it not as the status quo but as the decision maker's recent expectations, which makes reference dependence a prediction of the model rather than an assumption fed into it.

## Extensions and Applications

Loss aversion has proved the theory's most exportable idea. The endowment effect—people demand more to give up a good than they would pay to acquire it—follows directly, because parting with a possession is coded as a loss while acquiring it is coded as a forgone gain; Kahneman, Knetsch and Thaler (1990) demonstrated the gap experimentally with coffee mugs and ruled out the transaction-cost explanations. In finance, Benartzi and Thaler (1995) combined loss aversion with frequent evaluation of returns—myopic loss aversion—to explain the equity premium puzzle, the otherwise anomalous size of the historical excess return on stocks: an investor who checks a volatile portfolio often experiences many interim losses, each amplified by λ, and demands a large premium to bear them. Related patterns include the disposition effect, the tendency to sell winning assets too early and hold losing ones too long, and the reluctance of consumers to accept nominal wage cuts. O'Donoghue and Sprenger (2018) survey how reference-dependent preferences have been formalized and applied across these domains.

## Worked Example

The fourfold pattern can be computed directly from the two functions, using the cumulative-theory parameters α = 0.88, λ = 2.25, and weighting exponents γ = 0.61 for gains and γ = 0.69 for losses. Consider a lottery: a 0.001 chance of winning \$5,000, otherwise nothing. Its expected value is \$5, so a risk-neutral agent is indifferent between the ticket and \$5 in hand. Prospect theory is not indifferent. The weighting function turns the objective 0.001 into a decision weight of π(0.001) = 0.0144—the rare event is overweighted more than fourteenfold—while the value of the prize is v(5,000) = 5,000^0.88 = 1,799.3 units and the value of the sure \$5 is v(5) = 5^0.88 = 4.12 units. The ticket is therefore worth 0.0144 × 1,799.3 = 26.0 units against the sure amount's 4.12, so the model buys the lottery ticket: risk seeking over an improbable gain.

Now mirror the problem into losses: a 0.001 chance of losing \$5,000, otherwise nothing, against a sure loss of \$5—the position of someone deciding whether to insure. The loss-side weight is π(0.001) = 0.0084, again a large overweighting of the rare event, and the values carry the loss-aversion multiplier: v(−5,000) = −2.25 × 1,799.3 = −4,048.4 and v(−5) = −2.25 × 4.12 = −9.27. The risky option is worth 0.0084 × (−4,048.4) = −34.1 units, far worse than the sure loss's −9.27, so the model pays the premium: risk averse over an improbable loss. The same overweighting of small probabilities that sells the lottery ticket also sells the insurance policy—two of the fourfold cells, produced by one weighting function acting on opposite sign outcomes.

Discussion

Prospect theory succeeded where earlier critiques of expected-utility theory had not, because it did not merely catalogue anomalies but replaced the model with one that generated them from a small set of assumptions. Its reach is genuinely interdisciplinary: reference dependence and loss aversion now appear in asset-pricing models, in the design of contracts and default rules, in health-communication guidelines and in the behavioural-economics case for choice architecture. The theory is also falsifiable and much tested, which is why the debates around it are sharp rather than vague.

Those debates are real. The status of the reference point remains partly unresolved—whether it is the status quo, an expectation, an aspiration or a socially supplied comparison changes the predictions, and Kőszegi and Rabin's (2006) expectations-based model is one influential answer rather than a settled one. Loss aversion, the most cited component, has drawn the sharpest scrutiny: some choice patterns attributed to it can arise from diminishing sensitivity alone, its magnitude varies widely across studies and stakes, and a prominent critique argues that the evidence for a general tendency to weight losses more than gains is far weaker than the textbook consensus implies (Gal & Rucker, 2018). Prospect theory is best read as a powerful and well-supported description of choice under risk, not as a complete or final one.

## Current Directions

Two lines of recent work bear directly on how firmly the theory's core parameters can be pinned down. Ruggeri and colleagues (2020) ran a large multinational replication of the original 1979 choice problems across nineteen countries and thirteen languages and found the qualitative signatures—the reflection effect, the fourfold pattern, the shape of the value function—reproduced broadly, though with meaningful variation in magnitude between populations, evidence that the patterns are robust while their parameters are not culturally invariant. Brown, Imai, Vieider and Camerer (2024) took the opposite, quantitative tack, meta-analysing several hundred estimates of the loss-aversion coefficient and reporting a mean well above one but with striking heterogeneity, sensitivity to elicitation method, and signs of selective reporting—so that the textbook figure of about 2 is better understood as a central tendency across noisy, method-dependent estimates than as a fixed psychological constant. The two studies frame the theory's live empirical question: its patterns replicate, but the search for stable, universal parameter values is still open.

Common Misconceptions

Prospect theory says people are irrational.

The theory is descriptive, not a verdict on rationality. It specifies the consistent rules by which people actually evaluate risky prospects (Kahneman & Tversky, 1979); those rules depart from the expected-utility norm, but they are systematic and predictable, not random error. Behaviour that violates a normative axiom can still be lawful.

Loss aversion and risk aversion are the same thing.

They are distinct. Risk aversion is a preference for a sure outcome over a gamble of equal expected value; loss aversion is the greater weight placed on a loss than on an equal gain. Prospect theory predicts risk seeking over losses and over improbable gains (Tversky & Kahneman, 1992), so a loss-averse person is often not risk averse at all.

The overweighting of small probabilities means people think rare events are common.

Decision weights are not beliefs. A person can know a probability is one in a thousand and still let it carry more weight than a thousandth in a choice; the weighting function transforms how a probability enters the valuation, not the probability the person judges to hold (Barberis, 2013).

Glossary

Allais paradox.

The choice pattern demonstrated by Allais in which preferences between two gambles reverse when a shared, certain component is removed from both, violating the independence axiom of expected-utility theory and motivating the nonlinear weighting of probability.

Certainty effect.

The disproportionate preference for an outcome that is certain over one that is merely very probable, a consequence of the steep weighting function near probability one.

Cumulative prospect theory.

The 1992 revision that weights cumulative rather than individual probabilities, removing violations of stochastic dominance and extending the theory to prospects with many outcomes and to uncertainty.

Decision weight.

The number by which an outcome's value is multiplied in evaluating a prospect; derived from its probability through the weighting function but not equal to it.

Diminishing sensitivity.

The declining marginal impact of successive units of gain or loss as they move further from the reference point, giving the value function its concave-over-gains, convex-over-losses shape.

Editing phase.

The preliminary stage in which prospects are simplified—outcomes coded as gains or losses, shared components cancelled, probabilities rounded—before they are evaluated.

Endowment effect.

The tendency to value a good more highly once it is owned, because parting with it is coded as a loss while acquiring it is coded as a forgone gain.

Expected utility theory.

The normative model in which a rational agent evaluates a gamble by the probability-weighted utilities of its final wealth states; the benchmark prospect theory was built to describe departures from.

Fourfold pattern.

The prediction that risk attitude is averse for probable gains and improbable losses but seeking for improbable gains and probable losses.

Framing effect.

A change in choice produced by describing logically equivalent outcomes as gains rather than losses, violating description invariance.

Loss aversion.

The greater subjective weight of a loss than of an equal gain; the standard estimate is a ratio of about 2.25.

Myopic loss aversion.

The amplification of loss aversion by frequent evaluation of an investment's returns, offered as an explanation of the equity premium puzzle.

Reference point.

The baseline—status quo, expectation, or aspiration—against which outcomes are coded as gains or losses; its location determines the choice.

Reflection effect.

The reversal of risk preference when outcomes framed as gains are recast as losses, following from the opposite curvature of the two limbs of the value function.

Value function.

The function mapping gains and losses to subjective value; concave for gains, convex and steeper for losses, and passing through the reference point at the origin.

Weighting function.

The nonlinear transformation of probability into decision weight that overweights small probabilities and underweights moderate-to-large ones.

Key Researchers

Nicholas C. Barberis (b. 1971). Stephen & Camille Schramm Professor of Finance at the Yale School of Management; a leading figure in incorporating prospect theory into asset pricing, and author of a standard assessment of the theory's first three decades in economics. Faculty Page - ORCID

Colin F. Camerer (b. 1959). Robert Kirby Professor of Behavioral Economics at the California Institute of Technology; a behavioural economist and neuroeconomist who has quantified loss aversion and probability weighting across large bodies of experimental data. Faculty Page - ORCID

Daniel Kahneman (1934-2024). Eugene Higgins Professor of Psychology, Emeritus, at Princeton University; with Tversky he created prospect theory and received the 2002 Nobel Memorial Prize in Economic Sciences for integrating psychology into economics. Obituary - Wikipedia

Matthew Rabin (b. 1963). Pershing Square Professor of Behavioral Economics at Harvard University; with Kőszegi he built the expectations-based model of reference-dependent preferences, giving the reference point an endogenous foundation. Faculty Page - Wikipedia

Richard H. Thaler (b. 1945). Charles R. Walgreen Distinguished Service Professor at the University of Chicago Booth School of Business; documented the endowment effect and myopic loss aversion and received the 2017 Nobel Memorial Prize in Economic Sciences. Faculty Page - Wikipedia

Amos Tversky (1937-1996). Davis-Brack Professor of Behavioral Sciences at Stanford University; co-originator of prospect theory and of the heuristics-and-biases program, whose collaboration with Kahneman reshaped the study of judgment and choice. Memorial - Wikipedia

Peter P. Wakker (b. 1956). Professor at the Erasmus School of Economics, Erasmus University Rotterdam; a decision theorist whose axiomatic work clarified prospect theory's treatment of probability weighting and ambiguity. Faculty Page - ORCID

Frequently Asked Questions

What is prospect theory in simple terms?
Prospect theory is a psychological account of how people choose between risky options, holding that they judge outcomes as gains and losses from a reference point rather than as final wealth, weigh losses more heavily than equal gains, and distort probabilities by overweighting rare events (Kahneman & Tversky, 1979).

How does prospect theory differ from expected utility theory?
Expected utility theory says a rational agent evaluates gambles by the probability-weighted utility of final wealth states; prospect theory says people evaluate changes relative to a reference point, apply a value function that is steeper for losses, and replace probabilities with nonlinear decision weights (Barberis, 2013).

What is loss aversion?
Loss aversion is the finding that a loss produces a larger psychological impact than a gain of the same size; the value function is steeper below the reference point than above it, with a commonly cited ratio of about 2.25 (Tversky & Kahneman, 1992).

What is the fourfold pattern of risk attitudes?
It is the theory's prediction that people are risk averse for probable gains and improbable losses but risk seeking for improbable gains and probable losses, a pattern produced jointly by the value function and the probability-weighting function (Tversky & Kahneman, 1992).

What is a framing effect?
A framing effect is a reversal of preference caused purely by how equivalent outcomes are described, such as lives saved versus lives lost, which violates the description invariance that normative theories assume (Tversky & Kahneman, 1981).

Is prospect theory still accepted today?
Its core patterns are among the most replicated in the behavioural sciences, confirmed across nineteen countries in a large multinational study, though the exact magnitude of parameters such as loss aversion varies and remains debated (Ruggeri et al., 2020).

How is the reference point determined?
Often it is the status quo, but it can be an expectation or aspiration; one influential formalization treats it as the decision maker's recent expectations, making reference dependence a prediction of the model rather than an input to it (Kőszegi & Rabin, 2006).

How does prospect theory explain economic behaviour?
Loss aversion underlies the endowment effect and, when combined with frequent portfolio evaluation, the equity premium puzzle, so prospect theory has become a standard tool for explaining anomalies in markets and consumer choice (Benartzi & Thaler, 1995).

References