Foundationsbehavioral
Framing Effects: Same Facts, Different Choices
Overview
A medical treatment with '90% survival' is chosen over one with '10% mortality' — though they're identical. Gas 'discounts for cash' vs. 'surcharges for credit' — same prices, different politics. Framing effects (Tversky-Kahneman 1981) show preferences aren't just revealed, they're constructed by presentation. This lesson covers the classic demonstrations, attribute vs. risky-choice vs. goal framing, the mechanisms, applications in policy and marketing, and the deep question: which frame reveals 'true' preferences?
Learning objectives
- Define framing effects and reproduce the classic studies
- Distinguish attribute, risky-choice, and goal framing
- Explain the mechanisms: reference points and associations
- Apply framing: health, finance, climate, tax communication
- Evaluate the 'true preferences' problem
The classic demonstrations
The Asian disease problem (1981): '200 of 600 saved' → 72% choose the certain program; '400 of 600 will die' → 78% choose the gamble. Same outcomes, reversed majority. Attribute framing: '75% lean' beef outsells '25% fat' beef; '90% survival' beats '10% mortality' with patients and physicians. The mechanism is reference-dependence: 'saved' frames gains (risk aversion → certainty), 'die' frames losses (risk-seeking → gamble). Preferences aren't pre-stored and revealed — they're constructed on the spot, hostage to description. This single finding did more than any other to establish behavioral economics.
Rationality requires choices to depend only on outcomes, not descriptions. Framing effects prove descriptions drive choices — the axiom fails in the lab and the clinic.
Three kinds of framing
Attribute framing: one attribute described positively or negatively ('75% lean' vs. '25% fat') — the best-replicated, most robust type. Risky-choice framing: the Asian-disease pattern — certain vs. gambled outcomes across gain/loss frames, driven by prospect theory's reflection effect. Goal framing: the same action's consequences framed as gains of acting vs. losses of not acting — 'screening detects cancer early' (gain) vs. 'not screening risks late detection' (loss); loss-framed messages motivate detection behaviors more. Each type has different mechanisms and applications — lumping them obscures more than it reveals.
Loss-framed messages ('you risk missing early detection') outperform gain-framed ones for screening behaviors — a robust finding applied in cancer-screening and vaccination campaigns.
Mechanisms
Why do words move choices? Reference-point shifting: frames set the zero — 'saved' vs. 'die' locate outcomes in gain vs. loss territory, engaging different risk attitudes. Associative valence: positive frames evoke positive associations (lean = healthy), negative frames negative ones — fast, affective, System-1. Emphasis framing: highlighting different attributes changes their decision weight (job candidates evaluated on strengths vs. weaknesses). Information leakage: frames signal the speaker's beliefs ('surgeon chose 90% survival framing' implies endorsement). Multiple mechanisms mean framing is not one bias but a family — with different remedies per member.
Emphasis framing can be legitimate communication (highlighting relevant attributes). The bias charge applies when logically equivalent descriptions produce different choices — description invariance is the test.
Applications: policy, marketing, medicine
Policy: pension auto-enrollment ('you're in' vs. 'opt in'), organ-donation defaults, tax communications ('9 of 10 pay on time' — the UK's famous nudge), climate messaging (local health gains vs. distant planetary losses). Marketing: 'discount for cash' vs. 'surcharge for credit' (same prices; discounts poll better), 'only $1/day' reframing, free-shipping thresholds. Medicine: informed consent lives or dies on framing — physicians' presentation of survival vs. mortality statistics shapes treatment choices, raising ethical duties of balanced presentation. Every communicator frames; the question is whether they do it deliberately and honestly.
Adding '9 out of 10 people pay their tax on time' to letters raised payment rates — social-norm + framing, the Behavioural Insights Team's signature win, replicated worldwide.
The deep problem: which frame is 'true'?
If choices flip with frames, which choice reflects true preferences? This is behavioral welfare economics' hardest question. Options: the 'considered' preference (what you'd choose with reflection — but reflection has its own frames); the majority frame; or abandoning the idea of frame-free preferences entirely (Bernheim-Rangel's approach: only choices robust across frames count). Practically: require balanced presentation (both frames), use defaults that survive reflection, and prefer frames that aid comprehension over persuasion. The philosophical puzzle remains open — but policy can't wait for philosophy, so transparency and reversibility are the working ethics.
Key points
- Framing effects: logically identical options produce different choices depending on presentation — violating description invariance.
- Three types: attribute ('75% lean'), risky-choice (Asian disease), goal framing (gain- vs. loss-framed consequences).
- Mechanisms: reference-point shifts, affective associations, emphasis, and information leakage.
- Applications span medicine (informed consent), tax compliance, pensions, marketing, and climate communication.
- The open problem: which frame reveals 'true' preferences? Working ethics: balanced presentation, transparency, reversibility.
Common misconceptions
✓ Truth: They mean preferences are constructed, not pre-stored — even experts and physicians show them. It's a fact about cognition, not intelligence.
✓ Truth: Every presentation frames. 'Neutral' usually means the familiar frame. The honest goal is balanced presentation of multiple frames, not frame-free communication.
✓ Truth: It's also the mechanism of informed consent, public-health messaging, and democratic communication. The ethics depend on intent and transparency, not the technique.
Practice questions
Q1. The Asian disease problem demonstrates framing effects because:
- People can't understand probabilities
- Identical outcomes ('200 saved' vs. '400 die' of 600) produce reversed majority choices — violating description invariance
- The math was wrong
- Doctors are irrational
Same outcomes, opposite choices: the gains frame triggers risk aversion (certainty), the losses frame triggers risk-seeking (gamble). Description, not outcomes, drove the choice.
Q2. Attribute framing is best exemplified by:
- '75% lean' vs. '25% fat' beef — the same product evaluated differently
- Stock market crashes
- Interest rate changes
- Election campaigns
One attribute described positively or negatively — the most robust framing type. Positive frames evoke positive associations, shifting evaluation of the identical product.
Q3. The 'true preferences' problem in framing asks:
- Whether preferences exist at all
- Which of the frame-dependent choices reflects what the person really wants — the central puzzle for behavioral welfare economics
- How to ban framing
- Whether economics is a science
If choices flip with frames, welfare analysis needs a principled way to identify the 'true' preference — via reflection, robustness across frames, or other criteria. Still debated.
Click a question to reveal the answer and explanation.
Frequently asked questions
What's the difference between framing and anchoring?
Framing changes choices via how options are described (gain vs. loss language); anchoring pulls numerical judgments toward an initial number. Related but distinct mechanisms.
Can framing be used ethically?
Yes — balanced presentation of both frames, frames that aid true comprehension, and transparent persuasion are ethical. Manipulation hides the alternative frame or exploits misunderstanding.
Do framing effects replicate?
Attribute framing replicates very robustly; risky-choice framing (Asian disease) replicates well; some complex goal-framing results are more variable. The core phenomenon is solid.
How should doctors present treatment statistics?
Present both frames (survival AND mortality), absolute risks (not just relative), and visual aids. Balanced framing is an ethical duty of informed consent.