The framing effect
The framing effect is the change in a decision produced by how a claim is worded rather than by what the claim says, so that identical evidence, described two ways, draws two different answers.
What the framing effect is
The framing effect is a violation of a principle that any account of rational choice takes for granted: two descriptions of the same situation should lead to the same decision. A flight that arrives on time 90 percent of the time and a flight that is late 10 percent of the time are the same flight. A ground beef that is 75 percent lean and a ground beef that is 25 percent fat are the same package. People reliably prefer one wording to the other, and prefer it strongly.
What separates the framing effect from the rest of the cognitive bias family is that nothing has been hidden. In cherry picking, evidence is left out. In anchoring, an irrelevant number is planted. Here the information is complete, correct and fully available in both versions, and the decision still moves. That is why the framing effect is the most direct evidence that people do not evaluate options; they evaluate descriptions of options.
The one sentence version, and the three kinds of frame
In one sentence, the framing effect means that a gain frame and a loss frame produce different choices from identical facts. Researchers usually distinguish three kinds, and they behave differently enough that the distinction is worth keeping.
- Risky choice framing: a certain outcome is set against a gamble, and the whole set is described in terms of gains or in terms of losses. This is where preferences reverse most dramatically.
- Attribute framing: a single characteristic is described positively or negatively, such as 75 percent lean against 25 percent fat, or a 95 percent on time record against a 1 in 20 delay rate.
- Goal framing: the same action is presented in terms of what you gain by doing it or what you lose by not doing it, such as savings earned against savings forfeited.
Attribute framing is the version you meet every day on packaging and in reporting. Risky choice framing is the version that produced the experiment the whole field rests on.
The Tversky and Kahneman experiment: survival against mortality
Amos Tversky and Daniel Kahneman demonstrated the risky choice framing effect in a 1981 paper in Science, and the design is simple enough to describe in full. Participants were told that an outbreak was expected to kill 600 people and were asked to choose between two programs. In the first version the outcomes were described in terms of lives saved: one program would save 200 people for certain, the other gave a one in three chance of saving all 600 and a two in three chance of saving nobody. A clear majority chose the certain option.
A second group of participants received the same two programs described in terms of lives lost: 400 people would die for certain, or there was a one in three chance that nobody would die and a two in three chance that 600 would die. A clear majority now chose the gamble. The two versions are arithmetically identical. Saving 200 of 600 and losing 400 of 600 are the same outcome, and the preference reversed anyway.
The explanation Tversky and Kahneman offered comes from their earlier work on prospect theory: people tend to be risk averse when they are thinking about gains and risk seeking when they are thinking about losses. The wording decides which of those two rooms the reader is standing in. Nothing about the frame changes the numbers; it changes the reference point the numbers are measured from.
Attribute framing has its own foundational study. Irwin Levin and Gary Gaeth reported in 1988 that people rated ground beef labeled as 75 percent lean more favorably than the identical product labeled as 25 percent fat, including after tasting it. The label moved the judgment of something the participants had in their mouths.
Where the framing effect shows up in ordinary reading
The framing effect appears wherever a proportion, a price or a risk is reported, such as product labels, pass rates, warranty terms, survey write ups, performance dashboards and any figure that has a complement nobody printed.
- The pass rate. A training course reports a 92 percent pass rate. The same course has a 1 in 12 failure rate. Both are true, and only one appears in the brochure.
- The price. A shop that offers a 3 percent cash discount and a shop that adds a 3 percent card surcharge charge exactly the same two prices. The first framing produces far less complaint, which is why it is the one used.
- The relative change. A defect rate that moves from 2 in 1,000 to 3 in 1,000 can be reported as a 50 percent increase or as one extra defect per thousand units. Neither is false. They will not produce the same meeting.
The third example is where framing does the most damage in practice, because a relative change with no absolute base rate attached is uninterpretable, and it is the standard way that percentage changes get reported.
What the framing effect is confused with: anchoring, loss aversion and missing facts
The framing effect is confused most often with anchoring bias, and one pair of terms in this area is not a distinction at all. Framing bias and framing effect are two names for the same phenomenon, with "effect" preferred in the research literature and "bias" more common in business writing. There is no technical distinction between those two, and a source that draws one is inventing it.
| Term | What moves the decision | How to tell it apart |
|---|---|---|
| Framing effect | The wording of identical information | Both versions contain the same facts |
| Anchoring bias | A number seen earlier, often irrelevant | The anchor is extra information, not a rewording |
| Loss aversion | Losses weighing more than equivalent gains | The mechanism behind risky choice framing, not a synonym for it |
| Cherry picking | Facts that were left out | Information is genuinely missing |
| Loaded question | A presupposition smuggled into the question | Answering at all concedes something |
Framing is also not the same as lying, and this is where readers get stuck. Every fact must be stated in some words, and every choice of words is a frame. There is no neutral description available, only a set of frames of which you can see some and not others. That is a real limit on how far this bias can be eliminated, and it is why the countermeasure below is symmetrical rather than corrective. It also explains why framing pairs so easily with confirmation bias: a reader who already holds a view will find one of the two frames obviously the natural way to put it.
One question separates framing from anchoring and from a missing fact: was anything added or taken away, or was the same information only reworded? Only reworded, and the mover is the framing effect. An extra number in the room that had no business being there, such as a list price before a negotiation, and the mover is anchoring. Facts you were not shown at all, and the problem is not framing but selection, which is cherry picking.
What actually reduces the framing effect
What reduces the framing effect is restating the claim in its complementary frame before deciding, which takes about five seconds and is the only reliably effective move known.
- Convert every percentage into its complement and read both aloud.
- Convert relative changes into absolute numbers over a stated base.
- Convert a gain frame into a loss frame, and check whether your preference survives.
- Present both frames to anyone you are asking to decide, in the same document.
- Ask what reference point the wording assumes, and whether you chose that reference point.
Be honest about how well this works. Presenting both frames reduces the gap between decisions but does not close it, and framing effects have been demonstrated in people who study them for a living. Expertise in the subject matter does not confer immunity, which distinguishes this bias from the pattern described on the Dunning Kruger effect page, where the problem is concentrated among those with the least skill. Nobody is standing outside the frames.
The test to run on your next number
Ask one question of any statistic before you react to it:
- What is the complement of this number, and would I feel the same about the claim if that were the figure printed?
Then a second, for decisions rather than statistics:
- Is this option being described by what I get or by what I avoid, and who chose which?
If reversing the frame reverses your answer, you have not yet made a decision about the world. You have made one about a sentence.
Where the framing effect sits in cognitive psychology
In cognitive psychology the framing effect is filed under judgment and decision making, and its theoretical importance is larger than its everyday examples suggest. Formal models of rational choice assume description invariance: an agent's preferences should not depend on how an option is labeled. The 1981 result showed that human preferences violate that assumption systematically and predictably, not randomly, which is what made it useful. A random error can be averaged away. A systematic one cannot.
That is also why the framing effect belongs on a site about testing claims rather than only in a psychology syllabus. Logical validity, the subject of the deductive reasoning page, is entirely indifferent to wording: two logically equivalent statements support exactly the same conclusions. Human judgment is not indifferent to wording at all. The gap between those two sentences is where most persuasion happens.