The framing effect describes how logically identical decisions produce different choices depending on whether outcomes are described as gains or as losses. Tversky and Kahneman (1981) test this in The Framing of Decisions and the Psychology of Choice. They presented logically identical problems, worded differently, to separate respondent groups. In their disease-control problem, 72 percent chose the certain option framed as lives saved (N=152). Only 22 percent chose the same option framed instead as lives lost (N=155).
What the Study Found
In Problem 3, 84 percent preferred a sure $240 gain over a 25% chance to gain $1000. For the mirrored loss version, 87 percent preferred a 75% chance to lose $1000 over a sure $750 loss. In a two-stage version of a $30-versus-$45 gamble, 74 percent chose the sure $30 (Problem 6). When the same probabilities were presented as a single-stage gamble, only 42 percent chose the equivalent $30 option (Problem 7). 88 percent still bought a play ticket after losing an unrelated $10 bill (Problem 8). Only 46 percent bought a replacement after losing the ticket itself (Problem 9).
Methodology
The data came from brief questionnaires administered in a classroom setting to students at Stanford University and the University of British Columbia. Each of the ten problems was presented to a separate group of respondents, ranging from 77 to 200 people per problem. A separate group of respondents completed a modified version of Problem 3 with real monetary payoffs, replicating the pattern found with hypothetical outcomes. Different respondents received different versions of each problem, so comparisons are between groups rather than within the same individuals.
Key Statistics
| Metric | Finding | Context |
|---|---|---|
| Certain option chosen, gain frame | 72% | Problem 1, N=152, "lives saved" wording |
| Certain option chosen, loss frame | 22% | Problem 2, N=155, "lives lost" wording |
| Sure $240 gain chosen over risky $1000 gain | 84% | Problem 3, Decision (i), N=150 |
| Risky $1000 loss chosen over sure $750 loss | 87% | Problem 3, Decision (ii), N=150 |
| Sure $30 win chosen (certainty-framed) | 74% | Problem 6, N=85 |
| $30 option chosen (identical odds, no certainty stage) | 42% | Problem 7, N=81 |
| Still bought ticket after losing unrelated $10 | 88% | Problem 8, N=183 |
| Bought replacement after losing the ticket itself | 46% | Problem 9, N=200 |
| Drove to save $5 on $15 calculator | 68% | Problem 10, low-price version, N=93 |
| Drove to save $5 on $125 calculator | 29% | Problem 10, high-price version, N=88 |
| Prospect theory value function | π(p) v(x) + π(q) v(y) | Overall value of a two-outcome prospect |
| Weighting function ratio property | π(pq)/π(p) is less than π(pqr)/π(pr) | Property of the decision-weight function π |
Why This Matters
Standard expected-utility theory assumes preferences stay stable across logically equivalent descriptions of the same choice. Wording alone reversed the preferred option in this study, independent of the underlying probabilities and payoffs. Prospect theory's value function and decision-weight function offer a framework for anticipating when risk attitudes shift with presentation rather than substance. Framing investment choices, insurance products, or trading rules as gains versus losses can shift client decisions without changing the underlying numbers.