Skip to main content
economics medicine-and-health political-science psychology

​
Framing effect

​
Description

Equivalent options are chosen differently depending on how they are presented. The classical demonstration (Tversky & Kahneman’s 1981 Asian Disease Problem): the same set of disease-intervention outcomes is presented to two groups, one as “200 of 600 people will be saved” and one as “400 of 600 will die.” The mathematical content is identical, but the populations choose the risk-averse option in the gain frame and the risk-seeking option in the loss frame. The frame did the work; the facts were constant. The diagnostic question — “if we presented this choice in the equivalent alternative frame, would the population’s choice pattern change?” — is the test for whether a framing-effect is operating. When the answer is yes, the choice is being driven by the frame at least as much as by the underlying outcomes. Three sub-types (Levin, Schneider & Gaeth 1998) carry slightly different mechanisms:
  • Risky-choice framing (the Asian Disease type) — gain vs loss framing of options under uncertainty; loss-aversion is the engine.
  • Attribute framing — single attributes described positively or negatively (“90% lean” vs “10% fat”); evaluative-valence shift even without risk.
  • Goal framing — the same action described as achieving a benefit vs avoiding a harm; influences persuasion power.
Framing-effect is the descriptive concept — the empirical bias observation. The prescriptive sibling is reframe, the deliberate cognitive or rhetorical move to supply an alternative frame. Curating both keeps the bias-vs-intervention distinction crisp; collapsing them loses information about what the catalog is doing.

​
Triggers

User-initiated: User describes a choice or evaluation that seems to depend on how the option is presented, or notices that wording-changes produce real behavior changes. Vocabulary cues: “framing,” “how you frame it,” “gain frame,” “loss frame,” “presented differently,” “depends on how you say it.” Agent-initiated: Agent notices that two presentations of the same underlying choice are producing different decision-patterns in the same audience. Candidate inference: “is the frame doing the work here? Would the choice flip if we presented this in the equivalent alternative frame?” Situation-shape signals: Survey-result interpretation; medical decision-aid design; product copy A/B tests; policy choice-architecture design; political messaging; financial-product disclosure; persuasion-effectiveness research; comparison-shopping interface design. The signal is strongest when the same population produces different outputs from arithmetically-equivalent inputs.

​
Exclusions

  • Wording-changes that genuinely change content — sometimes apparently-equivalent wordings actually carry different information (different defaults, different implicit reference classes); the resulting choice-difference is not framing-effect, it is a real information-difference. Diagnostic: a properly-controlled equivalent presentation would have to be objectively identical in content for framing-effect to apply.
  • Strong-prior populations where the frame washes out — domain experts, repeated-game players, and pre-committed deciders show weakened framing-effects; their priors dominate the frame. The bias is not universal in strength.
  • Decisions that explicitly average over multiple frames — when a deliberation process surfaces both frames before deciding (as good medical decision-aids and policy reviews do), the framing-effect collapses or substantially weakens. The phenomenon requires that one frame be salient and the other not.
  • Pure preference for one descriptive style — sometimes a population genuinely prefers one wording for non-bias reasons (clarity, accessibility, brevity); the resulting choice-difference is a presentation-quality issue, not a framing-effect on the underlying decision.
  • Catastrophic-stakes decisions where the loss-side saturates — like loss-aversion, framing-effect is a local phenomenon near the reference point; at extreme stakes both sides flatten and the asymmetry weakens.

​
Structure

Internal structure of framing-effect: a table of its component slots and the concepts that fill them.

​
Relationships

Relationship neighborhood of framing-effect: a graph of the concepts it connects to and the concepts it is a part of.
  • reframe — descriptive vs prescriptive pair on the same axis. Framing-effect names the bias; reframe names the deliberate move that exploits or counters the bias. A competent practitioner uses knowledge of framing-effect to choose effective reframes.
  • loss-aversion — mechanistic foundation for risky-choice framing. Without loss-aversion’s asymmetric weighting, gain-vs-loss frames would not reliably produce choice-reversals. The two concepts together give the empirical phenomenon + the structural mechanism.
  • anchoring — sibling context-pollution effect at a different granularity (point-estimate magnitude vs choice between options); they often co-occur in real-world presentations that combine salient anchors with evaluative frames.
  • doctrine — professional disciplines install both-frame doctrines (medical informed-consent communication; regulatory disclosure standards; survey best-practices) as structural counter-pressure against framing-effect distortions.
  • cargo-cult — exploitative framing that wears the costume of substantive reframing is cargo-cult. The contrast keeps the discourse honest about when a frame is doing real analytic work versus performing schema-change without earning it.
  • context-asymmetry — the party setting the frame has a context advantage; framing-effect is a mechanism by which context-asymmetry produces concrete outcome differences.

​
Examples

Asian Disease Problem (Tversky & Kahneman 1981) · psychology

The Asian Disease Problem is the canonical experimental demonstration of the framing effect. Subjects were asked to choose between two programs to combat an outbreak expected to kill 600 people. When the options were described in terms of lives saved (Program A saves 200 for certain; Program B has a 1/3 chance of saving all 600 and a 2/3 chance of saving none), most subjects preferred the certain gain. When the mathematically identical options were redescribed in terms of lives lost (Program C: 400 die for certain; Program D: 1/3 chance no one dies, 2/3 chance all 600 die), preferences flipped — most subjects now preferred the gamble.Same outcomes, same probabilities, opposite choices. The result was the founding empirical case for the framing effect: identical content can produce opposite preferences depending on whether it’s framed in terms of gains or losses, with the gain frame inducing risk-aversion and the loss frame inducing risk-seeking.

Medical risk communication · medicine-and-health

“90% survival rate” vs “10% mortality rate” describing the same procedure systematically shift patient consent and physician recommendation. Decision-aid literature explicitly counters this with both-frame presentation.
same offer, framed via the free-trial frame vs the upcoming-charge frame; conversion rates often diverge by 1.5-2× in pricing-team experiments.
Druckman pushes back on the loose use of “framing” in political-science literature, distinguishing equivalency framing (Tversky-Kahneman type: logically-identical descriptions of the same lottery) from issue framing / emphasis framing (selecting which considerations are salient about a complex topic — “free-speech frame” vs “public-safety frame” for the same event). The mechanisms differ: equivalency framing exploits a specific cognitive distortion (loss-aversion), while emphasis framing operates by activating already-held considerations differentially. Conflating the two produces inflated effect-size estimates and bad replication targets.Inference: When a paper claims “framing effect of X%,” check whether the manipulation was equivalency or emphasis before generalizing. The two transfer differently to engineering practice — UI copy variation usually triggers equivalency-frame effects (when the underlying information is identical) while documentation organization usually triggers emphasis-frame effects (different priorities surfaced).
the same loan or savings product looks dramatically different under each frame; regulatory disclosure standards exist precisely to neutralize the framing-effect.
attribute-framing classic; the same product is evaluated more favorably in the positive frame even when the percentages are explicit.
George Lakoff’s Don’t Think of an Elephant! (2004) extended the framing-effect tradition from the lab into political communication. Building on his earlier cognitive-linguistic work, Lakoff argued that political language doesn’t just describe issues — it activates underlying conceptual frames that already carry moral and political commitments. The book’s titular advice (you can’t argue against a frame by denying it; doing so reinforces it) became influential in American progressive political strategy.The book is the political-communication application of the same shape Tversky and Kahneman documented in the laboratory: equivalent factual content presented through different frames produces different responses, because the frame quietly imports presuppositions the listener adopts along with the description. Lakoff’s contribution was making the lab phenomenon legible to practitioners who needed to choose between competing ways of describing the same underlying policy.
Levin, Schneider, and Gaeth catalog three structurally-distinct kinds of framing effect that prior literature had lumped together: risky-choice framing (the Asian Disease type — gains-vs-losses descriptions of the same lottery flip preferences), attribute framing (positive-vs-negative description of a single attribute, e.g., “75% lean” vs “25% fat” shifts evaluation), and goal framing (positive-consequence-of-acting vs negative-consequence-of-not-acting changes persuasion). Different mechanisms, different moderators, different effect sizes; treating them as one phenomenon obscures replication failures and confounds policy-design choices.Inference: When designing or critiquing a framing intervention, name which framing type is in play before predicting effect size. A choice-architecture nudge for opt-in defaults is goal framing; a nutritional-label redesign is attribute framing; a public-health risk communication is usually risky-choice framing. The structural primitive framing-effect partitions cleanly along these three subkinds — analogical transfers across the partition boundary are weaker than transfers within.
choice-architecture framing of the same underlying choice; opt-out countries have dramatically higher consent rates. Sometimes treated as a default-effect; structurally a framing-effect on the action (“I have to act to leave the program” vs “I have to act to join the program”).
“do you favor X?” vs “do you oppose X?” reliably produces different distributions in the same population on the same underlying policy question; the survey methodology literature centers framing-effect as a key threat to validity.
George Lakoff’s frame-analysis work flagged the political contrast between calling a policy a “tax cut” versus a “tax relief.” The legislative content can be identical — the same change to the same line in the tax code — but the linguistic framing differs in what it presupposes. “Cut” frames the existing tax as a neutral baseline being adjusted downward. “Relief” frames the existing tax as an affliction, with the policy as the rescuing intervention; once “relief” is in play, opposing the policy means opposing relief.The example shows the framing effect operating at the level of vocabulary choice rather than choice-set framing. Identical fiscal content produces different audience responses because the frame quietly imports an evaluation along with the description. Lakoff’s broader point in Don’t Think of an Elephant! is that this kind of framing work, done well, can determine which side gets to set the terms of the debate.
Thaler and Sunstein move framing from descriptive psychology (Kahneman/Tversky) to prescriptive policy design: every choice has some framing whether the designer notices or not, and the choice architect is responsible for the frame they ship. Their canonical cases — opt-in vs opt-out organ donation, default 401(k) enrollment, salient vs hidden cafeteria placement — show large behavior shifts from changes that preserve choice and information but alter the default frame.Inference: “We didn’t pick a frame” is not an available position for any interface or policy. The question for designers is not whether to frame but which frame, with what reversibility for the user. Pairs cleanly with the catalog’s make-wrong-unrepresentable move at one end (structurally remove bad outcomes) and the framing-effect empirical primitive at the other (acknowledge the unavoidable cognitive lever). The policy lineage opens a target domain for analogical transfer — when a feature-toggle, a config default, or a setup wizard is being designed, the choice-architecture frame applies.
The canonical demonstration of framing-effect. Tversky and Kahneman’s Asian Disease Problem presented two groups of subjects with the same set of disease-intervention outcomes. One group saw the choice described in lives-saved terms (“Program A: 200 of 600 people will be saved; Program B: 1/3 probability all 600 saved, 2/3 probability none saved”); the other saw the arithmetically equivalent choice described in lives-lost terms (“Program C: 400 will die; Program D: 1/3 probability nobody dies, 2/3 probability all 600 die”). The populations chose the risk-averse option (the certain 200 saved) under the gain frame and the risk-seeking option (the gamble) under the loss frame. The mathematical content was identical; the frame did the work.Inference: The paper established framing-effect as a robust, replicable empirical phenomenon and supplied the diagnostic test — would the choice pattern change if the same options were presented in the equivalent alternative frame? When the answer is yes, the frame is doing causal work in the decision. The remedy in high-stakes settings (medical informed-consent, regulatory disclosure, policy review) is to present both frames before deciding — the bias substantially weakens when both presentations are salient at once. The 1981 paper is also where Tversky and Kahneman first connected framing-effect formally to the loss-aversion kink in prospect theory’s value function: the gain/loss reversal is reliable because the same outcome is weighted differently above and below the reference point.