Framing Effect: How the Same Choice Flips When the Wording Changes

The two versions of the problem above describe the same two programs. Saving 200 of 600 people is the same outcome as losing 400 of them, and a one-third chance of saving everyone is the same gamble as a one-third chance that nobody dies. A decision maker who weighed outcomes alone would choose the same way in both versions. Most people do not, and the direction of the switch is predictable.

Prospect theory, which Kahneman and Tversky published in 1979, explains why. People evaluate options relative to a reference point rather than in absolute terms, and the value curve is shaped differently on each side of that point. Above it, in the domain of gains, the curve is concave: the first 200 lives saved feel worth more than the next 200, so a sure gain beats a gamble of equal expected value. Below it, in the domain of losses, the curve is convex: the first 200 deaths feel worse than the next 200, so a gamble that might avoid all loss beats a certain loss of the same expected size.

The wording of the problem sets the reference point. "200 will be saved" makes the 600 expected deaths the starting point and counts upward from it; "400 will die" makes the present moment the starting point and counts down. Nothing about the disease changes. What changes is which side of the value curve the options land on, and with it the appetite for risk.

Framing effects come in more than one kind. Risky choice framing, the one demonstrated here, changes risk preference. Attribute framing changes how good a single option looks: beef labeled 75 percent lean is rated better than beef labeled 25 percent fat. Goal framing changes how persuasive a message is: telling people what they stand to lose by not doing a breast self-exam was more effective than telling them what they stand to gain. Levin, Schneider and Gaeth, who set out this typology in 1998, argued that the three kinds work through different processes rather than one; the prospect theory account above is the account of risky choice framing, the kind demonstrated on this page.

Amos Tversky and Daniel Kahneman published "The framing of decisions and the psychology of choice" in Science in January 1981. The disease problem was the first of several demonstrations in the paper. One group of 152 participants saw the programs described in terms of lives saved; a separate group of 155 saw them described in terms of deaths. The two groups were drawn from the same pool of university students and did not know a second version existed.

In the gain frame, 72 percent chose Program A, the certain saving of 200 lives, and 28 percent chose the gamble. In the loss frame the preference reversed: 22 percent chose Program C, the certain loss of 400 lives, and 78 percent chose the gamble. The majority on each side was choosing the same policy that the majority on the other side had rejected.

The paper went on to show the same reversal with money, with a hypothetical lottery, and with a question about whether to buy a ticket after losing one, and it introduced the idea that people keep separate mental accounts for separate parts of a decision. The authors' conclusion was that the standard economic assumption that preferences do not depend on description, which they called invariance, fails in a systematic way that the theory of rational choice cannot repair.

The disease problem has been repeated many times in the years since, and the reversal is usually found, with a size that varies with the population and with how the two options are worded. The 2018 Many Labs 2 project replicated a different framing problem from the same 1981 paper, one about a discount on a purchase, and found the effect in its pooled sample across many sites; the disease problem itself was not part of that project.

Any decision described with numbers can be framed, and most are framed by whoever writes the description.

The framing effect lives in the description, so the counter is to change the description yourself.

• Restate every option in the other frame. If something is 90 percent likely to work, say out loud that it is 10 percent likely to fail, and notice whether the decision moves.
• Convert to absolute numbers. "200 of 600 saved" and "400 of 600 die" are both the number 200 and the number 400; writing them side by side makes it easier to check that the two descriptions are the same outcome.
• Decide the reference point on purpose. Ask what the baseline is and whether it is the right one; a loss relative to last year may be a gain relative to the plan.
• Separate the description from the decision in time. A second look later, especially in writing, makes it easier to check whether two descriptions are the same outcome.
• When you write for others, frame consistently, or give both frames. Choosing the frame that produces the answer you want is persuasion; stating both is information.

The framing effect is often used as a loose term for any influence of wording. Several nearby biases are more specific, and the differences matter when you are trying to counter one.

• It is not loss aversion, the underlying asymmetry that losses weigh more than equal gains. Loss aversion is one of the reasons the frame matters; the framing effect is the switch in choice that results when the wording moves an outcome from one side of the reference point to the other.
• It is not anchoring, where a number seen first pulls a later estimate toward it. Anchoring changes the size of a judgment; framing changes which option is preferred.
• It is not the pseudocertainty effect, in which an outcome that is only conditionally certain is treated as certain. That is a cousin from the same 1981 paper, driven by how stages of a decision are described rather than by gain and loss wording.