Choice Under Uncertainty

Choice under uncertainty is the part of microeconomics that asks how a person should—and how a person actually does—choose when the consequences of each option are not known in advance.

9 frameworks, from 1654 to today · Microeconomics · Economics

still a working view ended · each leads to its entry below

Overview

Choice under uncertainty is the part of microeconomics that asks how a person should—and how a person actually does—choose when the consequences of each option are not known in advance1. A farmer deciding which crop to plant, an insurer setting a premium, a saver allocating a pension, a doctor choosing a treatment: each must weigh outcomes that may or may not happen. The field's history is a long argument about what "rational" choice under risk means, and about what to do when real people depart from the rational benchmark. Its central tool, expected utility theory, became one of the most successful theories in all of economics—and its most productive failures became the foundation of behavioral economics.

Gambling, Probability, and Expected Value

The mathematical study of risk began with games of chance12. In the mid-seventeenth century, correspondence between Blaise Pascal and Pierre de Fermat (1654) laid the foundations of probability theory while analyzing how to divide stakes in an interrupted game12. Soon after, Christiaan Huygens published the first printed treatise on probability (1657), and Jacob Bernoulli's Ars Conjectandi (posthumous, 1713) gave the law of large numbers and the first sustained treatment of using probability to make decisions.

From this work emerged the expected value rule: the value of a gamble is the sum of each outcome's value multiplied by its probability1. A coin flip paying 100 or nothing is worth 503. The rule is simple, and it was enormously influential—it underlies actuarial practice and much of early probability. But it had an obvious problem, noticed early: people do not value gambles by their expected money1. The St. Petersburg paradox, discussed by Daniel Bernoulli in a 1738 paper for the Imperial Academy of Sciences in St. Petersburg, made the point sharply14. A coin is flipped until heads appears; the pot starts at 2 and doubles each flip. The expected payout is infinite (½·2 + ¼·4 + ⅛·8 + …), yet few would pay more than a modest sum to play. Expected value could not explain this.

Bernoulli's Utility and the Marginalist Tradition

Daniel Bernoulli's answer was that people value the utility of money, not money itself, and that utility rises at a diminishing rate. The first dollars matter more than the thousandth. This idea—diminishing marginal utility—made the St. Petersburg gamble worth a finite amount and gave a psychological account of risk aversion. It also connected choice under uncertainty to the broader marginalist revolution in economics, which by the 1870s (Jevons, Menger, Walras) was building value theory on marginal utility.

For decades, though, utility remained a vague, hedonistic quantity. Economists could not measure it, and comparisons across people seemed meaningless. The decisive move came from the ordinalist revolution of the 1930s, associated with John Hicks and R. G. D. Allen, which argued that consumer theory needed only rankings of bundles, not measurable utility. That raised an awkward question for risk: if utility is only ordinal, what does it mean to take an expected value of it? Adding utilities presupposes that they can be compared and weighted—a cardinal notion the ordinalists had rejected.

Von Neumann–Morgenstern and the Axiomatic Expected Utility

The resolution came from an unexpected direction: game theory. In Theory of Games and Economic Behavior (1944), John von Neumann and Oskar Morgenstern needed a numerical payoff for each player to define equilibrium. They showed that if a decision-maker's preferences over lotteries satisfy a small set of axioms—completeness, transitivity, continuity, and a substitution or "independence" condition—then those preferences can be represented by a utility function whose expected value ranks lotteries correctly. Utility was back, but now as a constructed index derived from choices among gambles, not a psychological feeling.

The independence axiom is the heart of it: if you prefer lottery A to lottery B, then mixing each with the same third lottery C (with the same probability) should not reverse your preference. This is what licenses the probability-weighted sum. The result was a triumph: it gave economics a rigorous, choice-based theory of risk and made expected utility the default model for insurance, finance, and macroeconomics. Later axiomatizations by Leonard Savage and others refined it, and the framework became so standard that "expected utility" and "rational choice under risk" were nearly synonymous.

Savage and Subjective Probability

Von Neumann–Morgenstern assumed the probabilities were known—objective chances, like dice5. Most real uncertainty is not like that. A firm does not know the chance its product fails; a doctor does not know the chance a treatment works. Leonard Savage's The Foundations of Statistics (1954) extended the theory to subjective expected utility: if a person's preferences over acts satisfy a set of axioms (including a "sure-thing" principle), then there exist both a personal probability and a utility function such that the person behaves as if maximizing expected utility with those personal probabilities65. Probability became a property of the decision-maker's beliefs, revealed by choices, not of the world5.

Savage's framework, building on earlier work by Frank Ramsey and Bruno de Finetti, is one of the deepest results in decision theory6. It also carried a strong normative claim: a rational person's beliefs should obey the probability calculus, and their choices should be consistent with expected utility. The "Dutch book" argument—that inconsistent beliefs or preferences expose you to a set of bets you are bound to lose—gave this claim teeth1.

Allais, Ellsberg, and the Cracks

Even as the theory consolidated, anomalies accumulated. In 1953 Maurice Allais presented a pair of choice problems in which most people, including eminent decision theorists, violated the independence axiom. The pattern—now called the Allais paradox—showed that people overweight outcomes that are certain relative to merely probable ones, a "certainty effect" the expected utility model cannot accommodate.

A second challenge came from ambiguity. Daniel Ellsberg's 1961 thought experiments (published in 1961 in the Quarterly Journal of Economics) presented urns with known versus unknown proportions of colored balls. Most people prefer to bet on the urn with known odds, even when the unknown urn offers the same or better expected payoff7. This ambiguity aversion violates Savage's sure-thing principle. Together, Allais and Ellsberg showed that the elegant axioms were descriptively false for many people—and raised the question of whether they were normatively required5.

Prospect Theory and the Behavioral Turn

The decisive break came from psychology. In 1979, Daniel Kahneman and Amos Tversky published prospect theory, a descriptive model built from experimental evidence rather than axioms89. Its central features: people evaluate outcomes as gains or losses relative to a reference point, not as final wealth levels; losses loom larger than equivalent gains (loss aversion); and probabilities are transformed—small probabilities are overweighted, moderate and high ones underweighted—so that the decision weight attached to an outcome is not its objective probability. The value function is concave for gains and convex for losses, producing risk aversion in gains and risk seeking in losses.

Prospect theory explained the Allais pattern, the certainty effect, and a mass of framing effects that expected utility could not. It became the empirical workhorse of behavioral economics; Kahneman received the Nobel Memorial Prize in 2002 (Tversky had died in 1996). A later refinement, cumulative prospect theory (Tversky and Kahneman, 1992), fixed technical problems with how the original handled many-outcome gambles by applying probability weighting to cumulative distributions, making the model consistent with stochastic dominance.

Rank-Dependent Utility and the Weighting of Probabilities

Prospect theory's probability weighting had a mathematical ancestor and descendant. In 1982, John Quiggin proposed rank-dependent utility, in which outcomes are ranked and decision weights are assigned by a transformation of cumulative probabilities rather than to each outcome separately10. This preserved the expected-utility structure in a generalized form while allowing systematic departures from linear probability weighting—and it avoided the violations of dominance that plagued naive weighting schemes. Rank-dependent utility, developed further by Chew, Epstein, Segal, and others, became a general framework within which prospect theory's weighting function could be embedded. It showed that one could keep much of the axiomatic apparatus while relaxing the independence axiom in a disciplined way.

Maxmin Expected Utility and the Ambiguity Literature

Ellsberg's paradox demanded a theory of ambiguity, and the leading response came from Itzhak Gilboa and David Schmeidler in 1989: maxmin expected utility. A decision-maker who is unsure of the probabilities is modeled as holding a set of possible priors and evaluating each act by its worst-case expected utility over that set11. This "pessimistic" rule rationalizes ambiguity aversion and connects to earlier work by Schmeidler on non-additive probabilities (capacities) and by Peter Wakker and others. Alternative models followed—variational preferences (Maccheroni, Marinacci, Rustichini, 2006), smooth ambiguity (Klibanoff, Marinacci, Mukerji, 2005), and the "alpha-maxmin" family—each trading off tractability, explanatory power, and normative appeal. The ambiguity literature remains active, with applications to asset pricing (the equity premium puzzle), insurance, and climate policy.

Behavioral Belief Formation

A parallel line of research asked not how people weight given probabilities but how they form beliefs at all. Psychologists had long documented systematic biases: overconfidence (people's confidence intervals are too narrow), the hindsight bias, and the tendency to update beliefs too little or too much relative to Bayes' rule. In economics, the behavioral belief formation literature—associated with figures such as Matthew Rabin, Colin Camerer, and later Stefano DellaVigna and Ulrike Malmendier—studied how people extrapolate from recent experience, overweight salient events, and hold biased priors. The "law of small numbers" (people treat small samples as representative) and projection bias (people underestimate how their preferences will change) became standard explanatory tools. This work fed into behavioral finance, where biased beliefs help explain bubbles, underreaction, and overreaction in asset markets.

The Present

Today, choice under uncertainty is a plural field. Expected utility remains the benchmark—the model against which alternatives are measured and the default in most applied work, from asset pricing to mechanism design. But the descriptive literature has produced a menu of models—prospect theory, rank-dependent utility, maxmin and smooth ambiguity, reference-dependent preferences, and behavioral belief formation—each capturing some departures from the benchmark while leaving others unexplained12. A central open question is which model to use when: the field lacks a single unified descriptive theory, and the normative status of the alternatives is contested. Some argue that ambiguity aversion is a rational response to genuine uncertainty; others treat it as a bias to be corrected. The rise of machine learning and large-scale data has also opened new ways to estimate preference parameters and test models outside the laboratory. What began as a puzzle about a coin-flipping game in St. Petersburg has become one of the most empirically rich and theoretically contested areas of microeconomics.

References

  1. Decision theory (Choice under uncertainty, introduction, History, Probability theory), Wikipedia
  2. Expected value (History), Wikipedia
  3. Expected utility hypothesis, Wikipedia
  4. St. Petersburg paradox, Wikipedia
  5. Subjective expected utility, Wikipedia
  6. Savage's subjective expected utility model, Wikipedia
  7. Ellsberg paradox, Wikipedia
  8. Prospect theory, Wikipedia
  9. Loss aversion (History), Wikipedia
  10. Rank-dependent expected utility, Wikipedia
  11. Itzhak Gilboa, Wikipedia
  12. Generalized expected utility, Wikipedia

The frameworks of Choice Under Uncertainty

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1654-1738

    Expected Value Theory

    How should one calculate the average outcome of a random event?

    Fly to its tower in Nousfarer

  2. 1944-Present

    Expected utility hypothesis

    How should one choose between uncertain outcomes?

    Fly to its tower in Nousfarer

  3. 1954-Present

    Subjective Expected Utility

    How should one assign value to uncertain outcomes?

    Fly to its tower in Nousfarer

  4. 1979-Present

    Prospect Theory

    How do people actually make choices under risk?

    Fly to its tower in Nousfarer

  5. 1982-Present

    Rank-dependent expected utility

    How should one weigh different possible outcomes?

    Fly to its tower in Nousfarer

  6. 1989-Present

    Ambiguity Aversion Models

    How should one decide when probabilities are unknown?

    Fly to its tower in Nousfarer

  7. 1989-Present

    Maxmin Expected Utility

    How should one make decisions when facing unknown risks?

    Fly to its tower in Nousfarer

  8. 1990-Present

    Behavioral Belief Formation

    How are beliefs about uncertain events formed?

    Fly to its tower in Nousfarer

  9. 2005-Present

    Smooth Ambiguity Model

    How should one choose when probabilities are uncertain?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

Expected Value TheoryExpected utility hypothesisSubjective Expected UtilityProspect TheoryRank-dependent expected utilityAmbiguity Aversion ModelsMaxmin Expected UtilityBehavioral Belief FormationSmooth Ambiguity Model

Beyond the field

Where Choice Under Uncertainty's ideas came from in other fields, and what they led to there: the strongest 12 of 58 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Choice Under Uncertainty is a planet in Nousfarer

Its city has a tower for each of these 9 frameworks. Land, walk in, and climb one concept at a time.

Land on Choice Under Uncertainty

Related fields

Fields its ideas are linked to

More in Microeconomics

See where Choice Under Uncertainty sits on the Map of Human Knowledge