Reference-Dependent Preferences

Core Idea

Reference-dependent preferences formalize the insight from prospect theory that people evaluate outcomes as gains or losses relative to a reference point rather than as absolute levels. Koszegi and Rabin (2006) developed the most influential model, proposing that utility has two components: consumption utility (standard utility from the outcome itself) and gain-loss utility (additional utility or disutility depending on whether the outcome exceeds or falls short of a reference point, typically rational expectations). This model generates predictions about labor supply (taxi drivers work less on high-wage days), consumer behavior (demand patterns respond to reference prices), and risk attitudes (which depend on the stochastic properties of expectations). Reference-dependence is now a core building block of behavioral economic theory, extending prospect theory from static laboratory gambles to dynamic economic settings.

Explainer

Prospect theory demonstrated that people evaluate outcomes relative to reference points, but it left a critical question unanswered: what determines the reference point? In the original laboratory experiments, the reference point was usually the status quo or an experimentally controlled endowment. But in dynamic economic settings — labor supply, consumption, investment — the reference point is not fixed or obvious. Reference-dependent preference models, particularly Koszegi and Rabin's, address this gap by providing a systematic theory of reference point formation.

Koszegi and Rabin proposed that the reference point is determined by the person's rational expectations about outcomes. If you expect to earn $200 today, earning $250 generates gain-loss utility from the $50 gain relative to expectations, while earning $150 generates loss-related disutility from the $50 shortfall. The total utility has two components: standard consumption utility (you enjoy spending $250 more than $150) and gain-loss utility (the pleasant surprise of exceeding expectations or the painful disappointment of falling short). Loss aversion means that the disutility of falling $50 short exceeds the utility of exceeding expectations by $50.

This seemingly simple modification has rich implications. In consumer demand, it predicts that price increases from an expected level reduce demand more than equivalent price decreases increase demand — an asymmetric demand response around the reference price. This has been confirmed in field data: consumers respond more strongly to price increases than to price decreases of the same magnitude, controlling for the price level. In labor markets, it predicts target-earning behavior when workers have daily income reference points — they work fewer hours when wages are high because they reach their target faster, and more hours when wages are low because reaching the target requires more effort.

The expectations-based reference point also explains patterns in risk attitudes. If you expect a certain outcome, any risk relative to that expectation involves potential losses as well as potential gains — and loss aversion makes the downside loom larger. This produces risk aversion around the expected outcome. But if you already expect a risky outcome (a gamble), your reference point incorporates the distribution of possible outcomes, and loss aversion is partially pre-digested into expectations. This means that risk attitudes depend not just on the gamble itself but on whether the risk was anticipated — a prediction that standard expected utility cannot make.

The broader significance of reference-dependent preferences is that they bring prospect theory into general equilibrium analysis. Original prospect theory was a theory of isolated gambles in laboratories. Koszegi and Rabin's framework makes it applicable to any economic setting where agents form expectations — which is essentially every setting. Labor supply, consumption-savings, portfolio choice, bargaining, and industrial organization can all be analyzed with reference-dependent preferences, generating predictions that differ from standard models in specific, testable ways. This has made reference-dependence a working model in applied microeconomics, not just a behavioral psychology curiosity.