Decoy Effect: How a Third Option Changes Which of Two You Pick

People rarely judge an option against an internal standard of what it is worth. They judge it against the other options in front of them, and some comparisons are much easier than others. Two options that each win on a different attribute, one cheaper and one better, leave the chooser with a trade-off and no procedure for resolving it. Two options where one is as good or better on every attribute, and better on at least one, leave nothing to resolve.

A decoy is an option built to lose that second kind of comparison. In the classic arrangement, called asymmetric dominance, the new option is no better than one of the originals on any attribute and worse on at least one, while the other original does not beat it that cleanly. The original it loses to is the target. The chooser now has one clean comparison available in a set that otherwise offers only a trade-off, and taking the clean one feels like reasoning rather than guessing.

Part of the pull is the account you can give afterward. A decision between two good options cannot be justified by the options alone, and an unjustifiable decision is uncomfortable. The target, by contrast, comes with a sentence attached: it beats the third option outright, so picking it was not arbitrary. That sentence does not bear on whether the target was the right choice for you, and it does not need to.

The effect is not universal. Across a decade of work it has been easiest to produce with options described by two or three numbers, on screen, with no prior preference, in the kind of simple set the 1982 experiments used. Attempts to reproduce it with richer, more realistic choices, real products, images, more attributes, or choosers who already know what they want, have often found it small or absent. The honest summary is that the structure is real and robustly demonstrable in the laboratory form it was discovered in, and that its size outside that form is contested.

Joel Huber, John Payne and Christopher Puto published "Adding asymmetrically dominated alternatives" in the Journal of Consumer Research in 1982. They gave respondents choices between two options described on two numeric attributes, across six product categories including beer, cars, restaurants and televisions. One option was better on one attribute, the other on the second, so neither dominated.

Each respondent then saw either the plain two-option set or a version with a third alternative added that was no better than one of the originals on either attribute and worse on at least one, while the other original did not beat it that way. Adding that alternative raised the choice share of the option that dominated it. The finding mattered to decision theory because it violates regularity, the principle that adding an option to a set cannot increase the number of people choosing an option already in it.

The best-known illustration is not from a journal. In Predictably Irrational, Dan Ariely described an Economist subscription page offering web only at 59 dollars, print only at 125 dollars and print plus web at 125 dollars, and put it to 100 students: 84 took the bundle, 16 took web only and nobody took print only. He then removed the print-only line and put the two remaining offers to another 100 students: 68 took web only and 32 took the bundle. Deleting an option nobody had chosen reversed the majority.

The effect has a genuine replication problem, and it surfaced in a 2014 issue of the Journal of Marketing Research. Shane Frederick, Leonard Lee and Ernest Baskin reported a long series of attempts in which the effect vanished once the options were described with real products and perceptually rich information rather than numeric attributes. Sybil Yang and Michael Lynn reported similar failures across many settings in the same issue. Huber, Payne and Puto replied that the effect was always expected to depend on comparable numeric attributes and a chooser without a prior preference, which is a narrower claim than the one in most popular retellings.

Decoys are most common where a seller controls the whole menu and benefits from one particular choice.

The decoy works through the set of options, so the counter is to change the set before deciding.

• Delete the options you would never take, then choose. If the answer changes once the unpicked option is gone, the unpicked option was doing the work.
• Write down what you want before you look at the menu. A prior preference is the condition under which the effect is weakest in the research.
• Price the attributes separately. What is a print edition worth to you in dollars a year? Answer that and the three-line page becomes arithmetic.
• Be suspicious of an option that is strictly worse than another. Nothing is priced by accident, and an option that nobody should buy is on the page for a reason.
• Ask what justification you are reaching for. If the sentence in your head is "it is obviously better than that one," check whether that one was ever a candidate.
• Keep the comparison to the things you will actually use. Comparability makes a comparison easy, not relevant.

Several effects describe how the surrounding options change a judgment. They differ in what the extra option does.

• It is not the contrast effect, where something looks better or worse because of what sits next to it. The decoy effect is a specific structural case, requiring the added option to be dominated by one of the originals and not by the other.
• It is not anchoring, where a number seen first pulls an estimate toward it. Anchoring moves how much something seems to be worth; the decoy effect moves which option gets chosen, with the prices unchanged.
• It is not the less-is-better effect, where a smaller offering is preferred when the two are judged separately and the preference reverses when they are judged side by side. That turns on whether the options are seen together; the decoy effect happens with everything on the same page.