The Paradigm Shift from Neoclassical to Behavioral Economics: Foundations of Decision-Making Biases

The assumption of *Homo economicus*—the perfectly rational, utility-maximizing agent with stable preferences and unlimited cognitive capacity—has served as the cornerstone of neoclassical economics since the marginalist revolution of the late 19th century. Formalized by von Neumann and Morgenstern (1944) in their axiomatization of expected utility theory (EUT), this framework rests on four untestable but widely accepted postulates: completeness (all options can be ranked), transitivity (if A ≻ B and B ≻ C, then A ≻ C), continuity (indifference curves are smooth), and independence (preferences between two options are unaffected by the addition of a common third option). While EUT provides a mathematically elegant foundation for modeling choice under risk, decades of experimental evidence have demonstrated systematic and replicable violations of its core axioms, laying the groundwork for the paradigm shift to behavioral economics. This chapter traces the historical evolution of this shift, formalizes the conceptual boundaries of anchoring, loss aversion, and framing effects—the three core decision biases central to this volume—and outlines the pluralistic methodological toolkit that defines modern behavioral science.

1.1 The Collapse of Perfect Rationality: Empirical Challenges to Neoclassical Assumptions

The first systematic challenges to the rational agent model emerged not from economics, but from cognitive psychology, where experimental studies of human judgment revealed that decision-makers rely on mental shortcuts (heuristics) that produce predictable, systematic errors. The most influential early challenge came from Allais (1953), who demonstrated that individuals systematically violate the independence axiom of EUT via what is now known as the Allais paradox. Consider the following two choice sets:

• **Choice Set 1**: Option A = 100% chance of $1 million; Option B = 10% chance of $5 million, 89% chance of $1 million, 1% chance of $0.
• **Choice Set 2**: Option C = 11% chance of $1 million, 89% chance of $0; Option D = 10% chance of $5 million, 90% chance of $0.

EUT predicts that if a decision-maker prefers A over B in Choice Set 1, they must prefer D over C in Choice Set 2, as the common 89% chance of $1 million in Set 1 is equivalent to the 89% chance of $0 in Set 2. However, 70% of participants in Allais’s original experiment preferred A over B and C over D, a pattern that cannot be explained by EUT without abandoning the independence axiom. A second foundational challenge came from Ellsberg (1961), who showed that individuals violate the expected utility assumption of known probabilities by exhibiting ambiguity aversion: they prefer known probabilities over unknown probabilities, even when the expected value of the ambiguous option is higher.

These paradoxes revealed a critical flaw in the neoclassical model: its assumption of unlimited cognitive capacity and stable, context-independent preferences does not reflect the bounded cognitive resources and context-sensitive evaluation processes of real decision-makers. The formalization of this boundedness came from Herbert Simon, whose 1947 work *Administrative Behavior* laid the foundation for bounded rationality as an alternative to global utility maximization. Simon argued that decision-makers face three core constraints that make perfect rationality computationally intractable: 1. Information constraints: Not all relevant information is available or accessible. 2. Cognitive constraints: The human brain has limited working memory (Miller’s 1956 “magical number seven, plus or minus two”) and processing speed. 3. Time constraints: Decisions often need to be made under time pressure, precluding exhaustive search of all options.

To model bounded rationality mathematically, Simon introduced the concept of *satisficing*, a portmanteau of “satisfy” and “suffice,” where decision-makers search through options until they find one that meets a minimum acceptable threshold (the aspiration level $v^*$), rather than searching for the global optimum. Formally, for a set of $N$ options with values $v_1, v_2, ..., v_N$ and search costs $c_1, c_2, ..., c_N$, a satisficing decision-maker selects the first option $i$ where $v_i \geq v^* - c_i$, rather than the option $i^*$ that maximizes $v_i$ as predicted by EUT. Empirical estimates of the aspiration level suggest it is typically set at 70-80% of the maximum possible value of the option set, balancing the marginal benefit of additional search against the marginal cognitive cost of processing new information (Simon, 1956).

Simon’s bounded rationality framework shifted the focus of decision science from “how do rational agents choose?” to “how do real agents choose given their cognitive constraints?” This shift was formalized and expanded in the 1970s by Daniel Kahneman and Amos Tversky, whose collaborative work produced the first systematic, empirically grounded model of decision-making under risk: prospect theory (Kahneman & Tversky, 1979).

1.2 Prospect Theory and the Systematic Deviations from Expected Utility

Kahneman and Tversky’s prospect theory (PT) was developed to explain the empirical regularities observed in judgment and choice experiments that violated EUT’s axioms. Unlike EUT, which assumes that utility is defined over final wealth levels, PT posits that utility is defined over *gains and losses relative to a reference point* (usually the status quo), with three core components that produce systematic deviations from rational choice: 1. Value function: A S-shaped function that is concave for gains, convex for losses, and steeper for losses than for gains (the loss aversion parameter $\lambda$, estimated at ~2.25 in the original 1979 study, meaning losses loom larger than equivalent gains by a factor of more than 2). 2. Probability weighting function: A non-linear function that overweights small probabilities and underweights moderate to high probabilities, explaining why people buy lottery tickets (overweighting the tiny probability of winning) and purchase insurance (overweighting the small probability of a large loss). 3. Reference dependence: The value of an outcome is not determined by its absolute magnitude, but by its distance from the reference point, which is highly sensitive to contextual factors including framing.

1.2.1 Mathematical Comparison of Expected Utility Theory and Prospect Theory

To formalize the contrast between EUT and PT, consider a choice between a sure gain of $300 and a 50% chance of gaining $600, 50% chance of gaining $0. Under EUT, with a utility function $u(x)$ defined over final wealth $x$, the expected utility of the sure option is $u(x + 300)$, while the expected utility of the risky option is $0.5u(x + 600) + 0.5u(x)$. For a risk-averse decision-maker with a concave utility function, $u(x + 300) > 0.5u(x + 600) + 0.5u(x)$, so they will choose the sure option, consistent with the concavity of utility over gains.

Under PT, the value of the sure option is $v(300)$, where $v(x) = x^\alpha$ for gains ($x \geq 0$), with $0 < \alpha < 1$ (estimated at ~0.88 in the original study). The value of the risky option is $0.5\pi(0.5)v(600) + 0.5\pi(0.5)v(0)$, where $\pi(p)$ is the probability weighting function, estimated as $\pi(p) = \frac{p^\gamma}{[p^\gamma + (1-p)^\gamma]^{1/\gamma}}$ with $\gamma \approx 0.61$. Plugging in the estimates, $\pi(0.5) \approx 0.42$, so the value of the risky option is $0.42 *