The Frame

The Reversal changed the way a preference was elicited; the frame can change the answer while leaving the offered consequences where they were. The earlier Collected Frame placed the difficulty in the words themselves. A choice appears to tell us which outcome the chooser values, yet the same outcome can be welcomed as a rescue and refused as a loss. What, then, has the act revealed? There is a temptation to answer that no preference existed until the sentence manufactured one. That conclusion runs ahead of the observation. A changed description may alter attention, interpretation or the point from which a consequence is judged; the resulting choice need not announce the creation of a whole new person. What the finding does defeat is the easy passage from an observed selection to a ranking assumed to stand independently of how the alternatives reached the mind.

The famous public-health problem makes the difference painfully spare. In Tversky and Kahneman's 1981 study, one group considered a disease expected to kill six hundred people. A program saving two hundred for certain was chosen by 72 percent; the alternative offered a one-third chance of saving everyone and a two-thirds chance of saving nobody. Another group received the problem in deaths: four hundred would die for certain under the first program, while the second offered a one-third chance of no deaths and a two-thirds chance of six hundred. Here 78 percent selected the gamble. The groups contained 152 and 155 respondents respectively. The result is a reversal of the majority pattern across descriptions, not an observed reversal by every individual. And the equivalence runs between descriptions of each program: the certain rescue remains the certain rescue, the gamble the same gamble. Equal expected lives do not make certainty and risk identical.¹

The reference point gives the result a possible mechanism. A gain can be felt as a diminishing increment above a baseline; a loss as a departure below it. In prospect theory's descriptive account, the value curve typically bends differently on its two sides, while probabilities receive decision weights that need not equal their numerical chances. A certain gain may consequently attract a chooser whom a certain loss sends toward a gamble. The familiar phrase loss aversion does not do all this work: greater steepness for losses is different from the curvature that changes attitudes toward risk. Nor is the baseline a dial held entirely by the experimenter. Tversky and Kahneman expressly give a part to the chooser's habits, norms and personal characteristics, and describe their patterns as substantial without making them universal. Wording can move the vantage point from which an outcome is evaluated. It does not command every eye.²

What the analyst calls the same option must therefore survive inspection. Imagine a merchant charging 100 in cash and 105 by card. Calling the five-unit difference a cash discount or a card surcharge leaves those two bills unchanged. This invented example preserves the earlier plate's question about who supplies the baseline: is 105 the normal price from which cash receives a favor, or 100 the normal price to which card users incur a penalty? But if the description also changes what the customer learns about a fee, a guarantee or the merchant's conduct, something relevant may have changed. An outcome described incompletely is not equivalent merely because an accountant can recover the same total. The invariance test is strongest when the consequences and the information needed to understand them are held fixed. It becomes evasive when the analyst dismisses a difference the chooser reasonably treats as part of the bargain.³

There is a second difference that a theory of final balances can miss: descriptions can affect what the person subsequently experiences. An anticipated disappointment, a sense of rescue or a recurring resentment may be part of living with the outcome, not only a transient error in selecting it. Tversky and Kahneman themselves consider the possibility that a frame changes experience as well as decision, and distinguish coherence of choice from correspondence with the experience one will actually have. That does not establish that every induced feeling deserves the same authority or that every seller's suggestion serves the buyer. It identifies the work a welfare claim must do. If only the presentation changes, ask whether the judgment persists when the equivalence is understood. If experience changes too, ask what changed and why it matters. The arithmetic of the invoice is indispensable; it is not a complete description of a life.⁴

The invitation to reconsider is valuable precisely because it is an inquiry rather than an automatic cure. Put the saved lives beside the deaths, disclose both prices, let the chooser explain what each account appeared to mean. Sometimes the comparison reveals an overlooked consequence or a dominated option. Sometimes two coherent attitudes remain, and the last answer given has no special claim merely because it came last. Bernheim and Rangel make the difficulty explicit in their discussion of choices between choices: a second decision is still made under conditions of its own. To privilege it over the first requires a reason, such as evidence that the first rested on misunderstood information. Framing can thus weaken a particular inference from behavior without silencing the person whose behavior is at issue. The discovery that presentation matters grants the investigator a question to ask. It grants no general warrant to substitute the investigator's preference.⁵

The earlier essay's political question survives these distinctions: who gets to describe the choice, and whose purposes does that description serve? A seller selecting the baseline is participating in the decision the market will later record as the customer's demand. That participation must be counted without imagining that the customer has become an empty surface. Choices can reveal durable commitments, contextual judgments, mistakes, or some mixture; finding which requires more than pointing at the sale. If wording decides the choice, the choice remains a sign of how the option was understood and evaluated in that setting. It ceases to be sufficient evidence of a ranking that would have endured every equivalent description. The frame belongs in the account of what happened. So does the chooser's capacity to see it, dispute it, and ask for the picture again.