Professional Documents
Culture Documents
IOANNIS EVANGELIDIS
JONATHAN LEVAV
ITAMAR SIMONSON
1
2 EVANGELIDIS, LEVAV, AND SIMONSON
classic consumer research shows that choice behavior can THEORETICAL DEVELOPMENT
be further influenced by the decision context, such as
changes in the configuration of the choice set (Huber, How Consumers Resolve Tradeoffs between
Payne, and Puto 1982, 1983; Simonson 1989). Desirability and Feasibility
Consumer research on context effects has primarily
Drawing on Montgomery’s (1983) seminal work, we
focused on two prominent findings: the attraction effect
Note that we do not argue that consumers always priori- a given alternative after accumulating supporting evidence
tize desirability over feasibility. Clearly, consumers often over time (Bhatia 2013; Noguchi and Stewart 2018; Roe,
choose HF options, which implies that feasibility is the key Busemeyer, and Townsend 2001; Trueblood et al. 2014;
driver of many decisions. Rather, we conjecture that, all Usher and McClelland 2004). However, while these mod-
else equal, consumers typically covet HD options and are els typically assume that evidence is accumulated ran-
motivated to assess whether their choice can be justified domly (e.g., all reasons are equally likely to be invoked),
FIGURE 1
Desirability
B B
A C A
Feasibility Feasibility
NOTES.—Illustration of the upscaling effect—introducing a symmetrically dominated decoy option (i.e., C) to a binary set that consists of an HF option (i.e., A) and an HD
option (i.e., B) increases the choice share of the HD option (i.e., B).
4 EVANGELIDIS, LEVAV, AND SIMONSON
choosing between A, a 1TB hard drive that costs $40, and likely to occur for A because A is superior to C on the fea-
B, a 2TB drive that costs $80. In this example, A is the HF sible dimension. In other words, the fact that A has the
option and B is the HD option. According to our frame- same capacity at a lower price than C is less likely to be
work, to reach a decision, consumers will first search for a invoked because B was prioritized over A when it was
promising alternative. As we argued above, an HD option identified as the promising alternative during the first stage
is more likely to be identified as a promising alternative of the decision process. Thus, A is less likely to draw
decoy to the HD option. In this research, we operationalize are systematic differences in the importance of the attrib-
ease of comparison in two ways. First, we facilitate the utes along which our stimuli are described that could possi-
comparison between the decoy and the HD alternative by bly explain the upscaling effect; we do not find such
manipulating whether the decoy is displayed next to the systematic differences. Second, while the finding reported
HD option versus the HF alternative. We propose that the in Evangelidis and Levav (2013) involves a decrease in the
impact of the decoy on the choice share of the HD option choice share of target options (i.e., options that are superior
TABLE 1
Overall results
Two-option condition (%) Three-option condition (%)
as whether decisions were hypothetical as opposed to con- S6 demonstrates that the upscaling effect can be extended
sequential. Across studies, we obtained robust evidence for to choices that involve a higher number of options. Study
the upscaling effect. Studies S1 and S2 provide evidence S7 provides evidence for the upscaling effect when prod-
that the upscaling effect can be observed when consumers ucts are represented perceptually using pictures (vs. text).
face tradeoffs on various attributes associated with desir-
ability and feasibility besides quality and price. Study S3
replicates our effect using consequential purchase deci- STUDIES 2 AND 3: PROCESS EVIDENCE
sions. Study S4 demonstrates that the effect replicates
when participants are forced (vs. not) to make a choice. Study 1 and studies S1–S7 provided robust evidence for
Study S5 shows that the upscaling effect replicates when the upscaling effect. In studies 2 and 3, we delve into the
the decoy option is not strictly dominated by the focal process underlying the effect. We have advanced a two-
alternatives but is a relatively inferior option instead. Study stage view of choice between options that present tradeoffs
8 EVANGELIDIS, LEVAV, AND SIMONSON
TABLE 2
between desirability and feasibility. Our account asserts support choice of HD options than they are to invoke rea-
that, in the first stage of the decision process, consumers sons supporting the choice of HF alternatives.
will identify the HD option as the promising alternative. Furthermore, we also find support for our hypothesis that
Furthermore, in the second stage of the decision process, consumers find it easier to justify choosing the HD option
consumers will search for reasons that support choice of when the decoy option is included in the set.
the promising HD option. When choosing from the binary
set, consumers may struggle to find strong enough evi- Study 2
dence that allows them to justify choosing the HD option.
Consequently, while consumers may covet the HD option, Method. All participants (N ¼ 506: 56.3% male; Mage
they may find it difficult to justify choosing it. In turn, con- ¼ 38.3 years, MTurk) made an initial choice between two
sumers may end up choosing the HF alternative or defer hard drives. Specifically, participants were provided with
choice. However, when the decoy is added to the set, con- brand A (price: $39.99; capacity: 2TB) and brand B (price:
sumers who previously might have had a difficult time jus- $79.99; capacity: 4TB). Participants could choose one of
tifying the choice of the HD option may now find it easier the alternatives or defer choice (worded as “search for
to justify choosing that option. Consequently, when the other options”). After making an initial choice, participants
decoy is added to the set, there should be more consumers rated their liking of each option on two independent 20-
that are willing to switch to the HD option than there are point rating scales (“How much do you like each brand?”
consumers willing to switch to the HF option because con- (1 ¼ not at all, 20 ¼ very much; slider scale). In the next
sumers will primarily invoke reasons that support choice of page, participants completed a 10-item self-construal scale
the HD option. (Singelis 1994) that served as a filler task. After complet-
We provide supporting evidence for our hypotheses in ing the filler task, participants were shown the same two
studies 2–3. Study 2 uses a repeated-measures design to hard drives, brand A and brand B, as well as an additional
demonstrate that the HD option is liked more by choosers decoy option, brand C (price: $79.99; capacity: 2TB).
of the HF option than the HF option is liked by choosers of Participants were asked to make a choice and indicate
the HD option in an initial choice between the two options. which option they believed was clearly better than C (i.e.,
As a result, when the symmetrically dominated decoy A, B, or both). Both measures were jointly presented on
option is subsequently added to the set, there are more par- the same page in a random order.
ticipants that switch from the HF option to the HD option Results. Results were consistent with our hypotheses
than there are participants switching from the HD option to (table 3). In the initial choice between brand A and brand
the HF option. Furthermore, study 2 shows that participants B, 38.7% of the participants selected the HD option (brand
are more likely to express the belief that the HD (vs. HF) B), while 47% selected the HF option (brand A) and 14.2%
option is better than the decoy. Study 3 directly tests our deferred choice. Furthermore, we found that the HD option
theoretical claim that consumers will be more likely to was liked more by choosers of the HF option (M ¼ 12.13,
invoke reasons that support choice of HD options than they SD ¼ 4.16) than the HF option was liked by choosers of
are to invoke reasons supporting the choice of HF alterna- the HD option (M ¼ 11.41, SD ¼ 4.08); t(432) ¼ 1.81, p ¼
tives. Specifically, in study 3, we use an open-ended ques- .072). This finding suggests that, as we have argued, some
tion to ask participants to describe the reasons that individuals may covet the HD option when choosing from
motivate their choices. We find support for our hypothesis the binary set but may find it difficult to justify choosing it
that consumers are more likely to invoke reasons that given its high price. As a result, there are more consumers
JOURNAL OF CONSUMER RESEARCH 9
TABLE 3 95% LLCI ¼ 0.34, ULCI ¼ 0.64) and the number of partic-
RESULTS OF STUDY 2 (N ¼ 506) ipants who indicated that both A and B dominated C
(41.9% vs. 35.4%, v2 ¼ 4.54, p ¼ .033; logit d ¼ 0.15,
Choice
95% LLCI ¼ 0.01, ULCI ¼ 0.29). Thus, participants in
study 2 were more likely to express the belief that the HD
Two-option set Three-option
(vs. HF) option is better than the decoy. Interestingly,
Before making a choice, participants were asked to in that participant’s experimental condition]. B and C are
explain what were the reasons motivating their preference the same price and have different ratings. My conclusion
(“Before making a choice, we would like to ask you to will simply be that hotel C is unnecessarily expensive. It
please explain in a few sentences what motivates your should probably be $89 dollars or less.” A third participant
decision. Specifically, what are the reasons or factors moti- stated, “Hotel B [the HD option in that participant’s experi-
vating your preference for one (or none) of these mental condition]—I want something a bit nicer if its for a
Results. We replicated our basic effect. Participants STUDIES 4–7: MODERATORS OF THE
were significantly more likely to select the HD option UPSCALING EFFECT
when the symmetrically dominated decoy was added to the
choice set compared to the control two-option condition Study 4
(59.2% vs. 48.9%; Wald v2 ¼ 8.62, p ¼ .003; logit d ¼
Study 3 demonstrated that consumers more frequently
0.23, 95% LLCI ¼ 0.08, ULCI ¼ 0.38).
invoke reasons that support choice of the HD option than
Furthermore, we examined participants’ explanations of
they invoke reasons supporting choice of the HF alterna-
their choices. As predicted, we found that participants in
tive. Building on the results of this study, in study 4, we
the three-option condition more frequently invoked the rea-
attempt to moderate the upscaling effect by prompting con-
son that the preferred option provides higher quality at the
sumers to rate the justifiability and invoke reasons support-
same price as the decoy than the reason that the preferred
ing the choice of both options before (vs. after) choosing.
option provides the same quality as the decoy at a lower
By doing so, we anticipated that the upscaling effect would
price (10.2% vs. 3.5%; Pearson v2 ¼ 14.23, p < .001; logit
be attenuated because it would become salient that the
d ¼ 0.63, 95% LLCI ¼ 0.29, ULCI ¼ 0.98). For example,
symmetrically dominated decoy provides strong enough
one participant stated that “I would choose Hotel A [the
reasons for choosing either of the dominating alternatives.
HD option in that participant’s experimental condition] as
it is 4-star while being the same price as Hotel C, a 3 star Method. Participants (N ¼ 1,609: 50.4% female; Mage
hotel.” Another stated that, “I would take B [the HD option ¼ 39.5 years, MTurk) were randomly assigned to one of
JOURNAL OF CONSUMER RESEARCH 11
attenuated when the price of the HD option was low (vs. decoy. Consequently, an HD option is less likely to benefit
high). from the addition of a decoy when its price is only slightly
higher than that of the HF alternative.
Results. Consistent with our hypotheses, the upscaling
effect was moderated by the price of the HD option (Wald
v2 ¼ 22.53, p < .001; table 5). We observed an upscaling Study 6
alternative B. For the other half (decoy next to HF), the are adjacent to each other, presumably because it is rela-
HD option was alternative A, while the HF option was tively easier for consumers to compare the decoy to the
alternative B. The decoy option was always alternative C. HD option. In study 7, we vary the ease of comparing the
For example, when the decoy was next to the HD option, decoy to the HD option by manipulating the mode by
in the two-option condition of the Bluetooth speaker prob- which alternatives are presented. We reasoned that the
lem, participants were provided with two options: brand A upscaling effect is larger in magnitude when the alterna-
TABLE 6
Overall results
Two-option condition (%) Three-option condition (%)
TABLE 6 (CONTINUED)
Overall results
Two-option condition (%) Three-option condition (%)
Hotel A (HF) 53 33
Hotel B (HD) 34 62
the data of study 2, there are more consumers that switch literature. Specifically, in a meta-analysis of the attraction
from an HF option to an HD alternative than there are con- effect, Heath and Chatterjee (1995) found that asymmetri-
sumers that switch from an HD option to an HF alternative cally dominated decoys commonly increase choice shares
when the decoy is added to the set. Our data also show that of high-quality high-price (i.e., HD) options, while they
consumers find it easier to justify choice of the HD option often failed to increase shares of low-price low-quality
when the decoy is added to the set. Finally, studies 4–7 (i.e., HF) options. According to Heath and Chatterjee, these
promising alternative (i.e., the HD option). In contrast, symmetrically dominated decoy option is displayed right
when an HF alternative is discounted, consumers may be next to the high-quality high-price product (e.g., in the
less likely to invoke a similar reason for choice because same page and within the same table in an online retail set-
HF options are typically not qualified as promising alterna- ting) and when the price of the higher-quality product is
tives during the first stage of the decision process. substantially higher than that of the lower-quality
Third, our research suggests a potential modification to competitor(s).
Prominence Effect,” Journal of Consumer Psychology, 7 (4), Liu, Wendy (2008), “Focusing on Desirability: The Effect of
323–43. Decision Interruption and Suspension on Preferences,”
Dumbalska, Tsvetomira, Vickie Li, Konstantinos Tsetsos, and Journal of Consumer Research, 35 (4), 640–52.
Christopher Summerfield (2020), “A Map of Decoy Influence Luce, R. Duncan (1977), “The Choice Axiom after Twenty
in Human Multialternative Choice,” Proceedings of the Years,” Journal of Mathematical Psychology, 15 (3),
National Academy of Sciences of the United States of 215–33.