The Framing Effect of Price Format

06-055


The Framing Effect of
Price Format

Marco Bertini and Luc Wathieu






















Copyright © 2006 Marco Bertini and Luc Wathieu
Working papers are in draft form. This working paper is distributed for purposes of comment and
discussion only. It may not be reproduced without permission of the copyright holder. Copies of working
papers are available from the author.

---

---

May 16, 2006

The Framing Effect of Price Format

Marco Bertini
London Business School, Regent’s Park, London, NW1 4SA, United Kingdom, mbertini@london.edu

Luc Wathieu1
Harvard Business School, Soldiers Field, Boston, Massachusetts 02163, lwathieu@hbs.edu

Existing evidence suggests that preferences are affected by whether a price is presented as one all-
inclusive expense or partitioned into a series of charges. To explain this phenomenon, we propose
a simple psychological mechanism whereby price format determines how many product attributes
are actively processed at the time of valuation. Three studies support the hypothesis that price
partitioning acts as an incentive to process multiple product dimensions. This process sometimes
leads to the paradoxical overweighting of minor (but easy to evaluate) attributes that would be
overlooked under an all-inclusive price format. The effect of price partitioning on demand can be
detrimental or beneficial, consistent with existing conflicting findings in the literature and with
variance in practice. Beyond its predictive and prescriptive implications, this theory contributes to
the general notion that pricing might affect as much as capture perceived value.

Keywords: Price Format, Framing Effects, Information Processing, Attention



A common approach to evaluating consumer preferences is to assume that individuals have a
utility function defined on multiple underlying product attributes or dimensions (Keeney and
Raiffa 1993; Lancaster 1966). Price information then enters the consumer choice process
indirectly through the budget constraint or, as is customary in conjoint analysis, directly as a
separate observable attribute in the utility function (Green and Rao 1971; Srinivasan 1982;
Winer 2005). Either way, the convention is that the role of price is to index the cost of making a
purchase.
Although this framework has been usefully applied to a variety of marketing problems,
recent research on the psychological aspects of pricing suggests that the role of price might be
more complex than anticipated by standard economic principles. In particular, a number of
studies have shown that the way price information is presented, termed price framing, often
significantly influences perceptions of value (e.g., Anderson and Simester 2003; Gourville 1998;
Heath, Chatterjee, and France 1995; Russo 1977).
The present research investigates the effects of framing prices alternatively as one all-
inclusive expense or partitioned into a series of mandatory charges. We posit that price format
determines the “depth” to which consumers analyze the various dimensions of an offer.
Consumers presented with an all-inclusive price are expected to concentrate their evaluation on

1 This paper is based on the first author’s dissertation and he would like to thank the other members of his
dissertation committee – John Deighton, John T. Gourville, and Elie Ofek – for their support and assistance through
the evolution of the research.

---

the focal attribute of the transaction (a book, movie tickets, groceries, etc.). Consumers presented
with a partitioned price, however, are expected to base their preferences on attribute-specific
evaluations that lead to an increase in the amount of attention paid to secondary attributes
(shipping and handling, booking service, delivery scheduling, etc.). This mechanism reflects the
straightforward intuition that not every product dimension is equally salient at the time of
purchase, and that the presence of multiple prices can re-sensitize consumers to elements of an
offer they might otherwise overlook.
The predicted effect of price partitioning on preferences raises the question of how attribute
evaluations aggregate to form overall judgments. Normatively, perceived gains and losses on
individual attributes should compensate each other such that only the total price is relevant.
However, existing research on information integration suggests that attribute evaluations can
receive differential weighting depending on the confidence with which these judgments are held,
with less weight assigned to more ambiguous evaluations. If true, this argument implies that
firms may benefit simply by partitioning an expense such that the attributes easiest to evaluate
are priced attractively.
The observation that preferences might be sensitive to price format has already generated
some interest in the literature. Previous research argues consistently that partitioning an expense
changes the way consumers process price information, but relies on different psychological
mechanisms to predict whether the outcome influences demand positively (Ayres and Nalebuff
2003; Hossain and Morgan 2006; Morwitz, Greenleaf, and Johnson 1998; Xia and Monroe 2004)
or negatively (Lee and Han 2002; Schindler, Morrin, and Bechwati 2005; Thaler 1985; Yadav
and Monroe 1993). In the present research, we make the claim that price format affects the way
consumers process the benefit side of market offerings (product dimensions, attributes, etc.).
Adopting this alternative line of attack allows us to propose a more flexible framework that
distinguishes circumstances under which price partitioning induces or suppresses demand.
Our theory can be summarized in two simple hypotheses. One, price components presented
separately to a consumer activate a matching number of attribute evaluations. Two, integration of
separate attribute evaluations to reach an overall assessment will be naturally biased in favor of
the attributes whose prices are easier to evaluate. These hypotheses are justified in the next two
sections and subsequently tested in three studies.

Background

In economics price reflects the monetary sacrifice a consumer makes to acquire a product or
service (Stigler 1987). From the standpoint of the firm, price is supposed to capture rather than
shape value. Consistent with this view, marketing tools for estimating consumer preferences
generally treat price as if it has no influence on how a product’s benefits are perceived. In
conjoint analysis (Green and Rao 1971; Green and Srinivasan 1990), for example, rank order
preferences are computed from a multivariate utility function in which price is a separate profile
attribute (Bradlow 2005) that has only a main (negative) effect on preferences.
Recent behavioral research, however, suggests that shifts in preferences could be determined
by the way prices are framed (Krishna et al. 2002; Winer 1988, 2005). Russo (1977), for
example, demonstrated that consumer expenditure is affected by whether unit prices are shown

2

---

as separate tags or ordered lists. Other researchers have found that firms can benefit from
communicating prices in percentages rather than absolute terms (Heath et al. 1995), reframing an
expense into a series of ongoing charges (Gourville 1998), adding plausible comparative price
information to an advertising message (Urbany, Bearden, and Weilbaker 1988), and even setting
prices one cent below the nearest dollar value (Anderson and Simester 2003; Thomas and
Morwitz 2005).
When considering the related question of whether firms should set an all-inclusive price for
their products or partition the expense into a series of mandatory charges, a useful starting point
is the theory of rational choice, which clearly predicts that the way prices are presented is
irrelevant as long as the overall terms of exchange remain the same (Arrow 1982). In practice,
however, the incidence of price partitioning is increasingly common.2 Whereas in the past firms
favored the use of single prices, today we see component charges not only in predictable settings
such as Internet sites and catalogs (product plus shipping and handling fees, convenience
charges, etc.), but also in unexpected circumstances, as when furniture stores break out the cost
of sofa pillows, hotels assess fees for room keys, airlines itemize landing and refueling expenses,
Christmas tree sellers separate the price of netting from that of trees, and so on.
Existing research on this topic lacks consensus on the likely impact of alternative price
formats. One argument is that price partitioning increases demand. Morwitz et al. (1998), for
example, suggest that consumers underestimate partitioned prices because they anchor on the
larger expense (the base price) and adjust insufficiently for the remainder (the surcharge). The
basis for this prediction is the notion that individuals trade off decision accuracy and cognitive
effort before deciding how to process price information (Johnson and Payne 1985; Shugan
1980). The authors demonstrate in an auction task that participants charged a buyer’s premium
on their bids consistently paid more for the same item than those in a control group who were not
assessed the premium. Further evidence of this processing heuristic and of the positive effect of
price partitioning are provided by Hossain and Morgan (2006) in a field experiment on eBay and
by Ayres and Nalebuff (2003) in the context of services. Finally, Xia and Monroe (2004)
demonstrated that the size, nature, and number of surcharges might reduce, but will not reverse,
this effect.3
Conversely, other papers have posited that price partitioning decreases demand. Most of this
research is grounded in prospect theory (Kahneman and Tversky 1979) and mental accounting
(Thaler 1985). Individuals are assumed to evaluate outcomes according to prospect theory’s
value function and, in line with mental accounting principles, perceive multiple losses as more
punishing than a single loss of equal monetary value, which implies, in turn, that an all-inclusive
price should be viewed more favorably than a partitioned one. Initial support for this hypothesis
was provided in the context of gambles (Thaler and Johnson 1990). The behavioral literature on
price bundling (Drumwright 1992; Gaeth et al. 1990; Johnson, Herrmann, and Bauer 1999;
Yadav and Monroe 1993) similarly finds that listing the price of each bundle component

2 A number of popular press articles discuss this trend. Examples include: Jennifer Bayot, “Fees Hidden in Plain
Sight, Companies Add to Bottom Line,” New York Times, 28 December 2002; Ellen Neuborne, “The Shipping
Charge: Break It to Them Quickly,” BusinessWeek, 29 October 2001; Gene Sloan, “Hotel Guests Hit with a
Surfeit of Surcharges,” USA Today, 15 August 2003; and Emily Thornton, “Fees! Fees! Fees!” BusinessWeek, 29
September 2003.
3 But see Lee and Han (2002) for an interesting discussion of the role of affect in this framework and Chakravarti et
al. (2002) for an alternative explanation of the positive impact of price partitioning.

3

---

increased the negative impact of the loss of money associated with a transaction.4 Lastly,
Schindler et al. (2005) combined prospect theory with perceptions of fairness to conclude that
online retailers should avoid pricing the cost of delivery separately.
In the following section we advance a theoretical argument that can explain both a positive
and a negative impact of price partitioning on demand. Whereas past research emphasized the
relationship between price format and numerical processing of the price, the present argument
explores how price format affects the way consumers process a product’s multiple dimensions.

Price Format and Product Evaluation

To explain how price format influences perceptions of value we propose a simple psychological
mechanism that links prices to the salience of product benefits. Consistent with the
characterization of consumer decision making as a goal-directed, problem solving process
(Bettman, Luce, and Payne 1998) that places primacy on goal-relevant criteria in order to
economize on cognitive load without sacrificing accuracy (Gigerenzer and Goldstein 1996;
Johnson and Payne 1985; Shugan 1980), we assume that consumers (1) hold a subjective
ordering of product dimensions based on their relevance to the task at hand (Fishburn, 1974) –
for example, they readily distinguish between focal and secondary attributes, and (2) employ a
heuristic by which they form as many attribute evaluations as the number of prices presented to
them. Specifically, whereas an all-inclusive price is expected to yield a single evaluative
judgment based on the focal attribute of the transaction, a partitioned price is expected to
highlight the presence of multiple benefits and raise the salience of secondary attributes that
would otherwise be overlooked in order to reduce cognitive effort.5
Intuitively, this theory suggests that an important function of price is to “spell out” product
benefits. If multiple prices alert consumers to the presence of multiple attributes, then all-
inclusive and partitioned prices should differ in terms of the impact of secondary attributes in
evaluation. For a product with clear focal and secondary attributes, variations in the perceived
value of the latter should exert greater influence on preferences when the price is partitioned.6
Formally:

H1: The perceived value of a secondary attribute has a greater influence
(positive or negative) on the overall assessment of a product when price is
partitioned than when it is all-inclusive.

4 A point of contention in the literature is whether price partitioning and bundling are different phenomena.
According to Morwitz et al. (1998, p. 453), the former involves the division of prices of single products, while the
latter the collective pricing of distinct products. But others find this distinction irrelevant (e.g., Stremersch and
Tellis 2002). We emphasize the framing effect of price presentation, whereas bundles are usually offered at some
“special” price and therefore viewed as a vehicle for price discrimination (Guiltinam 1987; Schmalensee 1982).
5 Note that this process is also compatible with three of the most common mental “shortcuts’ found in the literature:
adopting overly myopic decision frames (Payne, Bettman, and Johnson 1992); eliminating information even when
it is readily available (Kahneman and Tversky 1979; Russo and Dosher 1983); and sequential/lexicographical
processing (Fishburn 1974; Yadav 1994).
6 Consistent with previous literature (Zeithaml 1988, p. 14), we define perceived value broadly as “the consumer’s
overall assessment of the utility of a product based on perceptions of what is received and what is given.”

4

---

Two points need to be highlighted in association with H1. If, in fact, preferences are more
sensitive to the perceived value of secondary attributes when price is partitioned, it should be
possible to design an experiment whereby a change from all-inclusive to partitioned prices
benefits a product’s perception when the secondary attribute is attractive and damage it when a
secondary attribute is not attractive. Also, if preferences are less sensitive to secondary attributes
when price is all-inclusive, changes in these attributes could end up having little or no effect on
product evaluation. We test H1 and these related implications in study 1.
If price partitioning leads to separate evaluations for each product attribute, the question
naturally arises how these judgments are subsequently integrated to form an overall impression
of the offer. Normatively, a straightforward addition of the gains and losses perceived on each
dimension implies that aggregate judgments should be insensitive to price partitioning. However,
existing research suggests that judgment integration is often subject to bias (Anderson 1971;
Gaeth et al. 1990; Kahn and Meyer 1991). Borrowing from the literature on joint versus separate
evaluation (e.g., Hsee et al. 1999), we propose that the weight of each product attribute will be
determined by the evaluability of the price assigned to it. A price is considered to be more
“evaluable” if the consumer is able to judge its desirability with greater confidence. This
judgment depends on the precision (or ambiguity) of reference prices (Kalyanaram and Winer
1995; Lichtenstein, Bloch, and Black 1988; Monroe 1971). A price is more evaluable when a
consumer’s perception of what represents an attractive price is more precise, for example, when
the range of market prices deemed acceptable is narrow. Conversely, a price is less evaluable
when the range of market prices deemed acceptable is broad. Preference is expected to be biased
towards attributes whose prices are easier to evaluate. This prediction is captured in our second
hypothesis:

H2: Under price partitioning, the weight of an attribute in the overall assessment
of a product is conditional on the evaluability of its price. In particular, the
weight of an attribute is inversely related to the width of the range of
acceptable prices for that feature.

We test H2 in study 2. The concept of evaluability is closely related to the notion of price
uncertainty in marketing (Mazumdar and Jun 1993; Urbany and Dickson 1991) and cue validity
in social psychology (Mellers, Richards, and Birnbaum 1992). Consistent with recent empirical
evidence that suggests that consumers are overly responsive to peripheral expenses (Brynjolfsson
and Smith 2003; Lewis, Singh, and Fay 2006), we believe that in many commercial settings (e.g.
Internet sites, catalogs) the prices of secondary attributes that are more frequently encountered or
more homogeneous across purchases than focal attributes are often easier to evaluate as a result.
When such is the case, H2 yields the paradoxical effect of secondary attributes being overlooked
under all-inclusive pricing and over-emphasized under price partitioning.
Whereas studies 1 and 2 examine the impact of price formats in terms of changes in
participants’ overall evaluation of products, study 3 tests H1 and H2 more directly by eliciting
evaluations of attributes and verifying that the influence of focal and secondary attributes under
each price format conforms to the foregoing theory.

5

---

Study 1: All-Inclusive versus Partitioned Prices

Our first study involved a series of experiments devised to test H1. The main analysis is
presented in experiment 1a. Experiment 1b tested the robustness of these results by reminding
participants that the two price formats are equivalent. Experiment 1c tested the assumption that
the effect of price format is conditional on the relative salience of secondary attributes.

Experiment 1a: Air Travel
This initial experiment used a 2 (price format: all-inclusive, partitioned) × 2 (perceived value of
secondary attribute: bad deal, good deal) between-subjects design. The stimulus described a
purchase situation in which participants were asked to choose between two air travel alternatives
(appendix A). The focal attribute was one-way transit from Boston to San Juan, the secondary
attribute in-flight entertainment and meal service.7 The first option, Airline A, was held constant
across conditions and described simply by its long travel time (two flight segments for a total of
7 hours and 49 minutes) and price of $165. The second option, Airline B, varied in line with the
experimental design offering a shorter, non-stop flight (4 hours and 15 minutes) plus in-flight
entertainment and meal service. The total expense of this alternative was framed either as one
aggregate price of $215 or individual charges of $205 for the focal attribute and $10 for the
secondary attribute, which was manipulated by varying the type of benefit provided: one episode
of a sitcom and refreshments (bad deal) versus six movie channels and a full-service meal (good
deal).
After reading this scenario, participants were asked to indicate their choice intentions (1 =
definitely Airline A, to 8 = definitely Airline B) and rate the overall attractiveness of each
alternative (1 = very unattractive, to 7 = very attractive). To determine whether the perceived
value of the secondary attribute was manipulated as intended we collected an attractiveness
rating specifically for the in-flight entertainment and meal service (-3 = very unattractive, to 3 =
very attractive).
Participants (n = 210) were registered members of a subject pool managed by the research
center of a large U.S. business school. Because the same resource was used for all three studies,
we describe logistical details only once. At the time of the experiment the general population of
5,447 members was, on average, 39% male and 31 years of age. Eighty-seven percent of the
members had completed undergraduate education or higher. The participants, selected at random
and recruited via e-mail, were informed that the poll involved hypothetical purchase decisions,
that there were no right or wrong answers, and that they should consider only their own
preferences. Participation was voluntary, with a $5 payment upon completion. The experiment
was carried out online.
A preliminary analysis of variance (ANOVA) indicated that the secondary attribute was
manipulated as intended (Mgood deal = .91, Mbad deal = -.17, F(1, 206) = 20.59, p < .001). Neither the

7 A separate pre-test (n = 46) provided evidence that these attributes were not equally significant: 74% (χ2(1) =
10.52, p < .001) of participants asked to rank the airfare and in-flight entertainment and meal service in order of
importance listed the former first.

6

---

main effect of price format nor the interaction between the two independent factors was
significant.
An ANOVA on the choice measure revealed a significant main effect of the perceived value
of the secondary attribute (F(1, 208) = 6.67, p = .010). More important, this effect was qualified
by the expected interaction with price format (F(1, 206) = 9.52, p = .002). For these data to
support H1, two effects needed to be shown, (1) that preference for option B when the secondary
attribute is attractive is significantly higher under a partitioned than under an all-inclusive price,
and (2) that preference for option B when the secondary attribute is unattractive is significantly
lower under a partitioned than under an all-inclusive price. Individual contrasts revealed
precisely this pattern of results: participants offered a “good deal” on the secondary attribute
were more inclined to choose option B when the price was partitioned (M = 6.04) than when it
was all-inclusive (M = 5.16, t(101) = 1.99, p = .049), but the effect reversed when the secondary
attribute was perceived to be a “bad deal” (M = 4.20 vs. M = 5.33, t(105) = -2.37, p = .020).
The outcome of the individual evaluation of option B was similar. The ANOVA showed a
main effect of the perceived value of the secondary attribute (F(1, 208) = 4.19, p = .042),
qualified by the interaction with price format (F(1, 206) = 5.88, p = .016). Consistent with H1,
specific comparisons revealed the effectiveness of each price format to be contingent on the
perceived value of the secondary attribute. For an attractive secondary attribute, participants
evaluated option B more favorably when the price was partitioned (M = 5.79) than when it was
all-inclusive (M = 5.29, t(101) = 1.79, p = .077), but for an unattractive secondary attribute
participants evaluated option B less favorably when the price was partitioned (M = 4.91 vs. M =
5.37, t(105) = -1.68, p = .096).
An interesting empirical finding that corroborates our initial intuition involves the null effect
of secondary attributes on preferences under an all-inclusive price. We believe our theory to be
strengthened by the apparent insensitivity of both choice intention (t(105) = -.32, ns) and
individual evaluation (t(105) = -.25, ns) to shifts in the perceived value of the secondary
attribute. It turns out that participants responded “as if” receiving a better deal on this dimension
did not influence their assessment of the overall offer.
That participants might have been surprised by the separate charge for the secondary attribute
could potentially confound these results. Had this been the case because the practice was viewed
as unusual, an expectations-based rather than information processing-based explanation would
be supported. We addressed this eventuality by asking participants given partitioned prices
whether they believed the format to be atypical (1 = not at all typical, to 7 = very typical) and
whether they believed that firms should always present their prices in this manner (1 = strongly
disagree, to 7 = strongly agree). Apparently, this alternative account is not responsible for the
effects we observed: participants rated the format slightly atypical (M = 3.82, t(102) = -1.12, ns),
but the values were not statistically different from the mid point of the scale, nor did participants
seem to have strong convictions about how airlines ought to price their products (M = 3.74,
t(102) = -1.31, ns).

Experiment 1b: Groceries
Would the results of the first experiment have been different had the participants been reminded
that there is no real difference between the all-inclusive and partitioned prices? Although we do

7

---

not state a formal hypothesis for this question, the proposed theory suggests that making
participants aware of a potential numerical processing mistake should not matter because
different price formats influence the way consumers process a product’s dimensions, not its
price. To address this issue, we conducted a follow-up experiment in which participants (n = 85)
were asked to compare two ways an online grocer could present its prices to customers
(appendix B). The opening paragraph in the scenario alerted participants to the fact that these
formats were equivalent in terms of total expenditure. It was then explained via an example that
the price of a typical shopping basket (focal attribute) plus the price of the scheduling service
(secondary attribute) could be shown either as one lump sum of $95 (format 1) or as separate
charges of $86 and $9 (format 2), respectively. A pre-test similar to the one conducted for
experiment 1a confirmed that individuals (n = 43) perceived the shopping basket to be more
important than the scheduling service (74%, χ2(1) = 10.34, p < .001). The perceived value of the
secondary attribute was manipulated between-subjects by offering either a lengthy (8 hour),
firm-selected time slot for delivery during working hours (bad deal) or a brief (1 hour),
customer-selected time slot for delivery any time of the week (good deal).
We collected two preference measures. Participants were asked to indicate which format
made the offer more appealing (1 = definitely format 1, to 7 = definitely format 2) and to rate the
probability of purchase in each case (1 = very low, to 7 = very high). We also had participants
evaluate the attractiveness of the scheduling service (-3 = very unattractive, to 3 = very
attractive) and judge whether it was unusual for a grocer to price the scheduling service
separately (1 = highly usual, to 7 = highly unusual). As expected, the manipulation check
revealed the scheduling service to be viewed more favorably by participants when it represented
a “good deal” (M = .85) than when it represented a “bad deal” (M = -.66, t(83) = 4.39, p < .001).
We also confirmed that pricing the scheduling service separately was not perceived by
participants to be unusual (M = 3.81 vs. 4, t(84) = -.87, ns), which ensured that any effect of
price format could not be attributed to the “surprise” factor referenced earlier.
The results of the first experiment appear to generalize to situations in which individuals are
conscious of the numerical equivalence of the two price formats. Instead of providing answers
close to the indifference point on the scale, participants indicated the partitioned price to be more
effective when the secondary attribute was attractive (M = 4.66 vs. 4, t(40) = 2.35, p = .024) and
the all-inclusive price to be more effective when the secondary attribute was unattractive (M =
3.25 vs. 4, t(43) = -2.68, p = .010). We further observed that increasing the perceived value of
the secondary attribute influenced the likelihood of purchase under a partitioned price (Mgood deal
= 4.63 vs. Mbad deal = 4.05, t(83) = 2.00, p = .049), but not under an all-inclusive price (Mgood deal =
4.05 vs. Mbad deal = 4.16, t(83) = -.31, ns). Both results are consistent with H1.

Experiment 1c: Movie Tickets
The final experiment of study 1 was meant to test whether increasing the salience of secondary
attributes could moderate the effect of price format on preferences. This possibility derives from
our earlier assumption that product dimensions are ordered on the basis of their relevance to
consumers and evaluated sequentially, with a propensity for secondary attributes to be ignored
when the price is all-inclusive. Given two equally salient attributes, showing both price formats
to lead to the same valuation would reinforce the notions that (1) considering the processing of
product dimensions is critical to predicting the framing effect of price format, and (2) the

8

---