Framing effect
Claim assessed: The same choice, described as a gain or as a loss, gets different answers.
Published · how we work
In short
The framing effect means that the same choice, presented in different words, leads to different decisions. When outcomes are described as what you gain, people usually prefer the safe option. When the same outcomes are described as what you lose, they prefer the risk. Logically, the choice should be the same, because the options are identical.
What it says
The disease problem. Tversky and Kahneman (1981) gave students the following situation: an unusual disease is expected to kill 600 people, and there are two programmes to fight it. One group was told that programme A saves 200 people for sure, while programme B saves all 600 with a probability of one third and nobody with a probability of two thirds. 72% chose the sure programme. Another group was given the same programmes in terms of deaths: with programme C, 400 people die for sure, while with programme D nobody dies with a probability of one third and all 600 die with a probability of two thirds. Now 78% chose the risky option. Programmes A and C are identical, and so are B and D. The authors note that they saw the same reversal in university faculty and physicians.
In other situations too. In the same article, Tversky and Kahneman describe a question about the theatre. If you have lost a 10-dollar bill on the way to the theatre, 88% of respondents would still buy a ticket. If you have lost the 10-dollar ticket you already bought, only 46% would buy another. The loss is the same, but it is “booked” in a different mental account. The authors say the effects are large and systematic, though not universal, and that they don’t disappear even when the choices involve real money.
What the meta-analysis says. Kühberger (1998) gathered 136 papers with almost 30,000 participants. The conclusion: the effect is reliable, but small to moderate in size, and it varies a lot with how the experiment is built.
The huge replication. Rachev et al. (2021) gave the disease problem to 88,181 participants in 47 countries, in spring 2020, at the start of the pandemic. The effect appeared clearly and was even somewhat larger than in the Many Labs project, run in ordinary times. The share of risky choices and the size of the effect varied a lot from country to country, though.
What isn’t said matters. Kühberger and Tanner (2010) noticed that the risky option in the classic problem states both what happens in the good case and what happens in the bad case (“nobody will be saved”). Zhou et al. (2021) repeated their study with 1,119 participants in China. In the classic version, the effect appeared: 44% chose the risk with “saved”, 65% with “die”. When the risky option stated only the good case, the effect disappeared. When it stated only the bad case, the effect was largest. The authors conclude that part of the effect comes from how people simplify an incomplete description, not only from a different attitude towards gains and losses.
Example
One doctor tells a patient that, after a certain operation, 90% of patients are alive after five years. Another says that 10% die within five years. The information is the same, but the second wording may make the patient refuse the operation. The example is an illustration, not a study cited here.
How to apply it
The steps below are a practical way to apply it that we propose, based on the studies cited.
- Reframe the choice yourself. If an option is described as a gain, describe it as a loss too, and vice versa. If your preference changes, you were reacting to the words, not the options.
- Ask for the full description. “A 30% chance of success” tells only half the story. Ask what happens in the other 70% too.
- When you present options, give both versions. If you want the other person to decide for themselves, not just follow your wording, describe the outcomes both as a gain and as a loss.
- Ask who chose the wording. Someone who wants to persuade you can pick, out of two equivalent descriptions, the one that pushes towards the decision they want.
Limits and nuances
- The effect is small to moderate on average. In the classic experiments it looks huge, but the meta-analysis shows that its size depends a lot on how the experiment is built.
- It isn’t clear why it happens. The classic explanation rests on different attitudes towards the risk of gains and the risk of losses. The replication by Zhou et al. supports another explanation: people reduce the options to their gist (“some are saved” versus “maybe nobody is saved”) and decide on that.
- It isn’t necessarily irrational. Tversky and Kahneman note that, in some problems, it isn’t obvious which preference should be abandoned. Zhou et al. add that an incomplete description can legitimately be read in more than one way.
- Most of the evidence comes from hypothetical problems. Tversky and Kahneman say the effect doesn’t disappear with real money either, but the disease problem remains an imaginary situation.
- What we read. Tversky and Kahneman (1981), from the copy used in a North Carolina State University course, Zhou et al. (2021), from Judgment and Decision Making, and Rachev et al. (2021), from PLOS ONE, we read in full. For Kühberger (1998) we read the abstract. We haven’t read Kühberger and Tanner (2010); we describe it from Zhou et al. We mention the Many Labs project from Rachev et al.
Sources
- Amos Tversky, Daniel Kahneman (1981). The framing of decisions and the psychology of choice
- Anton Kühberger (1998). The influence of framing on risky decisions: A meta-analysis
- Anton Kühberger, Carmen Tanner (2010). Risky choice framing: Task versions and a comparison of prospect theory and fuzzy-trace theory
- Lei Zhou, Nan Liu, Ya-Qiong Liao, Aimei Li (2021). Risky choice framing with various problem descriptions: A replication and extension study
- Nikolay R. Rachev, Hyemin Han, David Lacko, Rebekah Gelpí, Yuki Yamada, Andreas Lieberoth (2021). Replicating the Disease framing problem during the 2020 COVID-19 pandemic: A study of stress, worry, trust, and choice under risk
See also: Decoy effect, Endowment effect and status quo bias, Loss aversion, Negativity bias