Retailer Pricing Strategy and Consumer Choice under Price Uncertainty

Abstract

This research examines how consumers choose retailers when they are uncertain
about store prices prior to shopping. Simulating everyday choice, participants made
successive retailer choices where on each occasion they chose a retailer and only
then learned product prices. The results of a series of studies demonstrated that
participants were more likely to choose a retailer that offered an everyday low pricing
strategy (EDLP) or that offered frequent small discounts over a retailer that offered
infrequent large discounts. This choice advantage for the retailer that was cheaper
more often manifested even when its average price was judged to be higher. The
same results were obtained when choices were made a day apart, when price feedback
was only given for the chosen retailer, and when price feedback was given for
both retailers. Participant’s expectations of future prices but not their judgments of
retailer’s past average prices predicted their subsequent retailer choice.

Full text

761
䉷2014 by JOURNAL OF CONSUMER RESEARCH, Inc. ●Vol. 41 ●October 2014
All rights reserved. 0093-5301/2014/4103-0012$10.00. DOI: 10.1086/677313
Retailer Pricing Strategy and Consumer
Choice under Price Uncertainty
SHAI DANZIGER
LIAT HADAR
VICKI G. MORWITZ
This research examines how consumers choose retailers when they are uncertain
about store prices prior to shopping. Simulating everyday choice, participants made
successive retailer choices where on each occasion they chose a retailer and only
then learned product prices. The results of a series of studies demonstrated that
participants were more likely to choose a retailer that offered an everyday low pricing
strategy (EDLP) or that offered frequent small discounts over a retailer that offered
infrequent large discounts. This choice advantage for the retailer that was cheaper
more often manifested even when its average price was judged to be higher. The
same results were obtained when choices were made a day apart, when price feed-
back was only given for the chosen retailer, and when price feedback was given for
both retailers. Participant’s expectations of future prices but not their judgments of
retailer’s past average prices predicted their subsequent retailer choice.
Imagine that each Friday you purchase fresh pasta from
retailer A or retailer B. You have observed that both re-
tailers offer the pasta at the same regular price, but that A
offers large though infrequent discounts on it, while B offers
small but frequent discounts on it. Unfortunately, you do not
know how much each retailer will charge on any given day
unless you visit that store. If your goal is to minimize spending
across shopping trips, where will you shop this Friday? Will
you choose the retailer for whom you believe the average
price across shopping occasions is lower? Will you choose
the retailer you believe offers larger discounts? Will you
Shai Danziger (shaid@tau.ac.il) is associate professor of marketing at
the Recanati Business School, Tel-Aviv University, TelAviv, 69978, Israel.
Liat Hadar (lhadar@idc.ac.il) is assistant professor of marketing at the
Arison School of Business, The Interdisciplinary Center (IDC) Herzliya,
P.O. Box 167, Herzliya 46150, Israel. Vicki G. Morwitz (vmorwitz@stern
.nyu.edu) is Harvey Golub Professor of Business and professor of mar-
keting at the Stern School of Business, New York University, 40 West 4th
Street, Room 807, New York, NY 10012. This work was supported in part
by a Marie Curie International Reintegration Grant PIRG06-GA-2009-
252592 to Liat Hadar and by the Henry Crown Institute of Business Re-
search in Israel. The authors acknowledge the helpful input of the editor,
the associate editor, and the three anonymous reviewers. The authors are
grateful to Ido Erev for insightful comments. Address correspondence to
Shai Danziger.
Mary Frances Luce served as editor and Rashmi Adaval served as associate
editor for this article.
Electronically published June 20, 2014
choose the retailer you think is cheaper more often? Or will
you choose the retailer you expect will be cheaper this Friday?
Much research has demonstrated the importance of price
in purchase decisions (Monroe 2003). A more fine-grained
analysis suggests that consumers’ purchase decisions are
driven by price perceptions rather than by actual prices. These
perceptions are highly subjective and susceptible to contextual
influences (Alba et al. 1999; Krishna 1991; Krishna et al.
2002; Zeithaml 1988). Retailers use various pricing strategies
to influence consumers’ price perceptions, assuming thatthey
will impact choice. Three prominent retail pricing strategies
are frequency discounting, where retailers offer frequent but
small discounts; depth discounting, where retailers offer in-
frequent large discounts; and everyday low pricing (EDLP),
where retailers offer products at a constant low regular price
(Hoch, Dre`ze, and Purk 1994). We study how these pricing
strategies influence consumers’ retailer choice decisions, their
perceptions of retailer prices, and the relation between them.
In contrast to existing research, we focus on settings in which
consumers decide where to shop without knowing the re-
tailers’ current prices (i.e., they choose under price uncer-
tainty) and where they only learn a retailer’s prices when they
visit it and see the prices in the store.
We show that consumers tend to choose the retailer that
is cheaper on the most shopping occasions and not the one
they believe to be cheapest on average. We discuss several
reasons why this choice pattern manifests and provide evi-
dence that consumers use past observed patterns to form
predictions of which store will be cheaper on each occasion

762 JOURNAL OF CONSUMER RESEARCH
and that these predictions, rather than average price judg-
ments, drive choice.
We next review and contrast two streams of literature to
develop our predictions: research on retailer and brand price
perceptions and research on experience-based and descrip-
tive-based choice. Then we present a series of studies that
test our hypotheses and elucidate the relations between re-
tailer pricing strategy, consumer price judgments, andchoice
under price uncertainty. Finally, we discuss the findings’
theoretical and practical implications.
RETAILER AND BRAND
PRICE PERCEPTION
Several studies have examined how competing retailers’
frequency, depth, and EDLP pricing strategies influence
consumers’ average or basket price perceptions when con-
sumers have full price information (i.e., under price cer-
tainty; Alba et al. 1994, 1999; Lalwani and Monroe 2005).
In the typical paradigm, participants make decisions over
multiple trials that simulate daily/weekly purchases of a
product offered by two competing retailers or brands. On
each trial they view each retailer’s or brand’s daily/weekly
price for a given product and then choose one based on
these prices. Notably, the choice task is not designed to
examine the impact of pricing strategy on choice, since all
prices are observed and participants presumably always
choose the cheapest price. Rather the purpose of the choice
task is to allow participants to sample from both pricing
distributions. The choice task includes at least nine trials
(but usually more than 24), and product prices vary within
retailer/brand. Once the choice phase is completed, consum-
ers retrospectively judge each retailer’s/brand’s average
price in a nonanticipated judgment task.
For example, in one study (Alba et al. 1999, study 5),
participants viewed the prices of one shampoo brand that
received frequent small discounts (frequency brand) and an-
other that received infrequent large discounts (depth brand)
over 36 simulated periods. The average price of the two sham-
poos was the same. Participants first observed prices and
chose a brand for each of the 36 trials. They then retrospec-
tively judged each brand’s average price. When each brand
was priced at only two levels, a regular or a discounted price
(e.g., 9.89 and 7.89; a dichotomous price distribution), the
depth brand was judged to have the lower average price (a
depth effect). Conversely, when each brand was priced at
many price points (e.g., 9.89, 9.39, 8.89, 8.39, and 7.89; a
nondichotomous price distribution), the frequency brand was
judged to have the lower average price (a frequency effect).
Using the same procedure, Lalwani and Monroe (2005)
found relative salience to be a key factor that determined
perceived average price (for a general discussion of saliency
effects see Taylor et al. [1979]). For both dichotomous and
nondichotomous price distributions, when discount frequency
was made more salient, a frequency effect was found, but
when discount magnitude was made more salient, a depth
effect was found. Krishna and Johar (1996) also showed that
perceived average prices for a single stream of prices will be
lower for depth versus frequency deals, and they explained
this related finding based on the salience of deeper discounts.
All these findings suggest that, under full price infor-
mation (i.e., price certainty), the relative salience of discount
frequency and depth influence consumers’ judgments of av-
erage price. However, what is not yet known from these
findings is whether these average price perceptions influence
retailer choice in the typical situation where a consumer
must decide which retailer to visit without knowing the
prices for that day (i.e., price uncertainty).
The only study that examined consumer choice under
price uncertainty was Alba et al.’s (1994) study 3. In that
study participants examined the prices of three sets of three
different products, for a frequency and a depth retailer, where
the total basket price for the retailers was the same. Partic-
ipants judged which store offered lower prices overall and
estimated the stores’ total basket price. They next chose
their preferred store given a goal to obtain good value. This
reflected a choice under price uncertainty because it was
prospective and made without knowledge of future price
information. There was a consistent frequency effect for the
two price judgments and for choice. Importantly, however,
because of the order in which the questions were asked
(price judgments first and then choice), it is possible that
participants’ prospective choices were influenced by the act
of having made prior price judgments, which may have
heightened the salience of these judgments. Given the pro-
tocol used, we cannot tell whether participants would have
chosen the frequency retailer if they had not previously been
asked to judge average prices, or whether consumers’ per-
ceptions of a retailer’s average price, under normal circum-
stances, influence their choices under price uncertainty.
Our goal is to systematically examine the effects of re-
tailer pricing strategies on consumers’ retailer choice when
consumers learn price information only after choosing a
retailer. Participants in our studies first choose their preferred
retailer (many times) and only then provide average price
judgments. This procedure eliminates the possibility that
asking questions about price perceptions artificially in-
creases their salience and their reliance in choice. Wepredict
that under price uncertainty, participants tend to choose the
retailer that is cheaper most often. We provide evidence that
consumers’ expectations regarding the prices they are likely
to encounter on the next shopping trip drive their choices
and not the average price perceptions they form based on
observing retailers’ past prices. We next describe the the-
oretical background for these hypotheses.
REASONS WHY CONSUMERS
CHOOSE RETAILERS THAT ARE
CHEAPER MORE OFTEN
Decision-making research has long been interested in how
people evaluate outcomes (Kahneman and Tversky 1979;
Payne 2005; Thaler 1985) and how people choose between
options for which the distributions of potential outcomes

DANZIGER, HADAR, AND MORWITZ 763
are uncertain and learned only from experience (Edwards
1962; Erev and Barron 2005; Estes 1961; Gonzalez and Dutt
2011). We propose that retailer choice under price uncer-
tainty is a form of experience-based choice. In the standard
experience-based choice paradigm, participants choose one
of two unlabeled buttons across a series of trials. On each
trial, once participants choose a button (an option), they are
shown a value (feedback) drawn at random with replacement
from that button’s payoff distribution. This value constitutes
their payoff for that trial. Participants typically receive no
prior information about the payoff distributions, and there-
fore they base their choices only on the feedback they re-
ceived in prior trials. The findings indicate that people tend
to give small probability outcomes less weight than is nor-
matively warranted (Barron and Erev 2003).
We contend that retailer choice under price uncertainty is
conceptually similar to experience-based choice. When con-
sumers choose a retailer under price uncertainty, they do not
know the available retailers’ price distributions and therefore
can rely only on prices they encountered on prior shopping
trips. We therefore posit thata similar pattern of results should
manifest as in experience-based choice, such that consumers
will tend to choose a retailer that is cheaper more often over
one that offers less frequent large discounts.
Research on experience-based choice indicates that small
probability events may be undersampled—participants ex-
perience these events less frequently than they actually occur
(when only partial feedback is provided; Fox and Hadar
2006; Hertwig et al. 2004)—and/or that participants may
underweight them in choice (Erev and Barron 2005; Hertwig
et al. 2004; Ungemach, Chater, and Stewart 2009). This
leads small probability outcomes to impact choice less than
is normatively warranted, and in the retailer choice context,
would lead consumers to be less likely to choose the depth
retailer, whose large discounts occur with small probabili-
ties. Prospect theory and theories regarding prediction of
future outcomes offer additional insights as to why consum-
ers should tend to choose a retailer that is cheaper more
over one that infrequently offers a much lower price.
Mental Accounting Principles
When consumers choose a retailer under price uncertainty,
they likely assess its price by comparing it to the observed
or expected price of the competing retailer. If the chosen
retailer offers a lower price than the forgone retailer, the out-
come is coded as a “gain,” and if it offers a higher price, the
outcome is coded as a “loss.” Previous research demonstrates
that losses loom larger than gains (Tversky and Kahneman
1991) and that consumers prefer options that offer a lower
overall chance of incurring a strict loss (Payne 2005). Because
the depth retailer will usually be more expensive, offering
more frequent losses and less frequent gains, a consumer will
deem the depth retailer less attractive than the frequency or
EDLP retailer. Furthermore, over shopping occasions, the
depth retailer offers few large gains while the frequency re-
tailer offers many small gains, and the depth retailer offers
many small losses, while the frequency retailer offers few
large gains. Given the shape of the value function (Kahneman
and Tversky 1979), consumers should obtain more positive
value from many small gains than a few large gains, and they
should receive less negative value from a few large losses
than many small losses (Thaler 1985; Tversky and Kahneman
1991). These mental accounting principles predict a general
preference for the retailer that is cheaper more often.
Predicting Future Outcomes Based on Previously
Encountered Outcomes
Humans tend to perceive patterns and correlations ev-
erywhere even if the patterns are not actually present and
even when perceiving them leads to erroneous inferences
(Shermer 2011). For example, in an investment setting,
Bloomfield and Hales (2002) found that MBA students used
the prevalence of past trend reversals in the value of a se-
curity as an indicator of the likelihood of future reversals
even when they were explicitly told that the sequences were
random and that past outcomes provided no information
about future outcomes. Whitson and Galinksy (2008) report
that, when faced with uncertainty, people seek to gain con-
trol over the situation perceptually, by identifying patterns
among stimuli and by making predictions based on these
perceived patterns. Likewise, in the experience-based choice
context, Gonzalez and Dutt (2011) offer that if people as-
sume that past sequences are representative of future se-
quences, they will likely look for patterns in previously
experienced outcomes (whether they exist or not) and use
them to predict future outcomes. Conceptually consistent
with this view, Danziger and Segev (2006) report that for
product prices purportedly sampled over time, participants’
reference price for evaluating a target price reflects the tem-
poral patterns of the price sequence (ascending and descend-
ing prices). In our retailer choice context, use of a prediction
strategy should produce a choice pattern whereby a retailer’s
choice share is proportional to the number of times the
retailer provides the best outcome (probability matching;
Erev and Barron 2005; Estes 1961; Humphreys 1939). This
leads to a larger choice share for the retailer that more fre-
quently offers lower prices. Although, when using this strat-
egy, in some shopping trips, consumers may predict that the
depth retailer will be cheaper than the frequency retailer and
will therefore choose it, the frequency retailer will more
often be predicted to offer the best outcome and therefore
will be chosen more often.
In summary:
H1: Consumers will choose a retailer that is slightly
cheaper on the majority of shopping occasions
more often than a retailer that is much cheaper on
a small number of shopping occasions.
Note that none of the explanations described above impli-
cates consumers’ perceptions of average retailer prices as a
factor that influences their choice under price uncertainty.
In fact, we contend that in this realistic shopping context
with price uncertainty, consumers do not naturally think

764 JOURNAL OF CONSUMER RESEARCH
about average prices. Rather we contend that they only gen-
erate and think about average price estimates if they are
explicitly asked to do so, such as was done in the previous
literature. We therefore predict that, even in situations where
consumers later judge the frequency retailer to have a higher
average price than the depth retailer, they will still choose
it more often. Recall that the previous research has dem-
onstrated that average price perceptions for the frequency
retailer are higher than for the depth retailer when retailer
price distributions are dichotomous, yet we predict that in
the same setting consumers will tend to choose the frequency
retailer. To help us demonstrate this dissociation between
average price perceptions and choice, we primarily focus
on dichotomous price distributions in our studies. However,
we also generalize our choice results to situations where
price distributions are nondichotomous, where prior work
has shown that price perceptions for the frequency and depth
retailers depend on price discriminability.
H2a: When retailer price distributions are dichoto-
mous, consumers choose the frequency retailer
more often, yet judge it to have a higher average
price.
H2b: When retailer price distributions are nondicho-
tomous, price discriminability affects average
price judgments but it does not affect choice.
Consumers choose the frequency retailer more
often whether price discriminability is high or
low. In contrast, judgments of average price are
lower for the frequency retailer when discount
frequency is made salient (low price discrimin-
ability) and are lower for the depth retailer when
discount magnitude is made salient (high price
discriminability).
Next, in studies 1A, 1B, and 1C, we demonstrate that par-
ticipants tend to choose a retailer that is cheaper more often,
and we demonstrate the dissociation between average price
perceptions and choice. Studies 2–4 show that these effects
generalize to a wider range of settings, and they offer initial
explanations for why they occur and evidence for what in-
ternal price representation instead drives choices if average
prices do not. Next we describe the studies that test these
predictions.
STUDY 1A: FREQUENT VERSUS
DEEP DISCOUNT RETAILER
The main aim of the first set of studies was to test hy-
potheses 1 and 2a. Study 1A examined consumers’ retailer
choice and average price perceptions for retailers using fre-
quent versus deep discounts under price uncertainty. A sec-
ond aim was to compare these findings to those obtained
under price certainty (Alba et al. 1999; Lalwani and Monroe
2005). A third aim was to test whether participants only
use average price judgments in choice decisions when
these judgments are made salient prior to choice.
Method
Eighty-seven students at Ben-Gurion University partici-
pated in return for course credit. The study used a 2 #2
#2 mixed design with a within-subject factor (pricing pat-
tern: frequency vs. depth) and two between-subjects factors,
to which participants were randomly assigned: price cer-
tainty (price certainty vs. price uncertainty) and task order
(average price judgments made first vs. future choice task
completed first). In this study and in all of our studies,
retailer location on the computer screen (left vs. right) was
counterbalanced. Since location never influenced our results
(all F!1), we do not discuss this further.
Participants were told they would purchase a pack of
portobello mushrooms from one of two competing retailers
over 100 weeks and that their choice should be based only
on price because the retailers sold mushrooms of the same
quality. Participants were also asked to minimize their total
spending over the 100-week period. To incentivize partic-
ipants, they were informed that the two with the lowest
overall spending would receive additional compensation of
40 Israeli New Shekels (NIS; ≈$10 at the time of the study).
Participants made 100 successive choices between a fre-
quency and a depth retailer, simulating 100 weekly pur-
chases. The mushrooms were priced by the frequency re-
tailer at 9.89 NIS for 50 weeks and at a sale price of 7.89
NIS for the remaining 50 weeks. They were priced by the
depth retailer at 9.89 NIS for 88 weeks and at a sale price
of 1.56 NIS for the remaining 12 weeks. The average price
for both retailers was 8.89 NIS. Table 1 provides a summary
of the study characteristics and findings for this and all the
remaining studies. All participants saw the same price se-
quence, which was constructed such that the frequency re-
tailer had five sales randomly distributed in each 10-trial
interval and the depth brand had one sale randomly distrib-
uted every 10 trials and an additional two sales randomly
distributed across the 100 trials. Only one retailer could offer
a discount on a given trial. To avoid a recency effect on
average price judgments, the depth retailer’s last discount
did not occur in the final five periods.
In each trial, participants in the price certainty condition
first saw the retailers’ prices and then chose between them.
Participants in the price uncertainty condition first chose a
retailer and afterward saw both retailers’ prices for that week
(these were revealed so that participants’ price judgments
made after the choice phase would be based on the same
price information as in the price certainty condition).
Next, participants either judged the average price of each
retailer and then chose which retailer they would visit for
10 future purchases, or they completed these two tasks in
the reverse order. As in previous research, participants did
not know in advance that they would be asked to make
average price judgments (Alba et al. 1994, 1999; Lalwani
and Monroe 2005). Finally, to better understand what drives
choices under price uncertainty, we asked participants in the
price uncertainty condition to describe in writing the strategy
they used to make their 100 choices (see Alba et al. [1994]
for use of a similar procedure).

DANZIGER, HADAR, AND MORWITZ 765
TABLE 1
SUMMARY OF STUDIES
NFrequency/EDLP
distribution Depth
distribution Feedback
Frequency/EDLP
retailer choice
share
(%)
Study 1A*36 Regular: 9.89 (.5) Regular: 9.89 (.88) Complete 72
Discount: 7.89 (.5) Discount: 1.56 (.12)
Study 1B 47 EDLP: 8.39 Regular: 9.89 (.7) Partial 66
Discount: 4.89 (.3)
48 EDLP: 8.39 Regular: 9.89 (.7) Complete 64
Discount: 4.89 (.3)
Study 1C 31 EDLP: 8.39
Partial 58
Frequency:
Regular: 8.89 (.3)
Discount: 7.75 (.7)
Study 2 32 Regular: 9.89 (.5) Regular: 9.89 (.8) Partial 60
Discount: 6.79–9 (.5) Discount: 4.19–5.60 (.2)
27 Regular: 989 (.5) Regular: 989 (.8) Partial 56
Discount: 679–900 (.5) Discount: 419–560 (.2)
Study 3 55 EDLP: 8.39 Regular: 9.89 (.666) Partial 73
Discount: 4.89 (.333)
Study 4 56 Regular: 9.89 (.2) Regular: 9.89 (.8) Complete 75
Discount: 8.95 (.8) Discount: 6.14 (.2)
N
OTE
.—The table presents number of participants (N), distribution characteristics (frequency, EDLP, and depth), type of feedback (partial
and full), and choice share of the frequency/EDLP distribution for each of the studies.
*An additional 51 participants completed study 1 under price certainty. The choice share of the frequency retailer was 76% under price
certainty.
Results
100 Successive Choices. One participant whose choice
share for the frequency retailer was more than three standard
deviations below the mean was removed from this analysis.
The results showed the same pattern with and without this
participant.
The frequency retailer was chosen 76% of the time under
price certainty (SD p15%, median p85%) and 72% of
the time under price uncertainty (SD p20%, median p
67%). Both values differ significantly from 50% (t(50) p
12.5, p!.0001, and t(34) p6.6, p!.0001, respectively),
but they do not differ significantly from each other (t(84) p
⫺1.2, NS). Although the choices shares under certainty and
uncertainty were similar, participants chose very differently
in the two conditions. Under price certainty, because par-
ticipants knew each retailer’s price before choosing, they
almost always chose the retailer that was cheaper on a given
trial (P
depth
p95%, P
frequency
p97%). In contrast, under
price uncertainty, participants chose the depth retailer on
only 21.4% of the occasions when it turned out to be cheaper
and the frequency retailer on 73.7% of the occasions when
it turned out to be cheaper.
Perceived Average Price. We excluded 17 participants
whose average price judgments fell outside the range of
prices offered by each retailer from this and the following
analyses.
Participants’ average price judgments were submitted to
an ANOVA with price (certain vs. uncertain) and task order
(perceived average price first vs. allocation first) as inde-
pendent variables and pricing pattern (frequency vs. depth)
as a repeated variable. Consistent with our prediction (hy-
pothesis 2a), there was a marginally significant depth effect
(M
frequency
p8.26 vs. M
depth
p7.91; F(1, 66) p3.2, pp
.08). No other effects were significant.
We next examined, in the price uncertainty condition,
where prices were seen only after choice, the relation be-
tween average price perceptions and choices across the 100
trials. A regression analysis with the frequency retailer’s
choice share as the dependent variable and the perceived
average prices of the two retailers as the independent var-
iables revealed that neither the perceived average price of
the frequency retailer (bp.23, t(30) p1.28, NS) nor that
of the depth retailer (bp.13, t(30) p0.73, NS) predicted
the choice share of the frequency retailer.
Future Choices. Future choice allocations to the fre-
quency retailer (out of 10 total) were submitted to an ANOVA
with price (certain vs. uncertain) and task order (perceived
average price first vs. purchase allocation first) as independent
variables. Participants allocated significantly more choices to

766 JOURNAL OF CONSUMER RESEARCH
the frequency retailer when the choice allocation task pre-
ceded the average price judgments than when it followed it
(Mp7.24 vs. 5.57, respectively; F(1, 66) p4.72; p!.05).
No other effects were significant.
To further test the relation between perceived average
price and decisions under uncertainty in this choice allo-
cation task, we ran a regression that examined the effect of
a judged depth advantage variable (computed by subtracting
the judged average price for the depth retailer from that of
the frequency retailer) on choice allocations as a function
of task order. The interaction was significant (bp⫺.27,
t(68) p⫺2.31, p!.05). Judged depth advantage predicted
choice allocations when the perceived average price judg-
ments preceded the choice allocations (bp⫺.61, t(29) p
⫺4.17, p!.001) but not when they followed it (bp⫺.22,
t(29) p–1.4, pp.16).
Jointly the ANOVA and regression results suggest that
participants did not spontaneously rely on judged average
price for the choice allocation task, which, like the 100
successive choices task, reflects a decision under price un-
certainty. Had they done so, choice allocations should have
been similar for the two different task orderings. These find-
ings suggest that only when average price perceptions are
made salient by the measurement task do they influence
choice allocations such that the lower a retailer’s average
price the more often it is chosen.
Decision Rule. If average price perceptions do not seem
to naturally influence choice, a natural question then is, what
led participants to be more likely to select the frequency
retailer? To obtain some insights into this, we examined
participants’ descriptions of the strategy they used to make
their 100 choices. These descriptions were classified by two
independent coders to one of the following four categories,
involving different internal price representations: a predic-
tion strategy (e.g. , “I tried to identify a pattern and topredict
which retailer would be cheaper on the next trial”), an av-
eraging strategy (e.g., “I estimated the average price of each
retailer and chose the lower one”), a frequency strategy (e.g.,
“I chose the retailer that offered cheaper prices moreoften”),
or a depth strategy (e.g., “I chose the retailer that offered
fewer discounts, but the discounts were more substantial”).
Finally, we also coded for other strategies (e.g., “I chose by
the color of the retailer boxes”). Of those in the price un-
certainty condition (Np35), 10 reported using a prediction
strategy, 1 an average price strategy, 11 a frequency strategy,
2 a depth strategy, and 11 were classified as “other.”
Discussion
The results of study 1A support our predictions by dem-
onstrating a clear tendency to choose the retailer that is
cheaper more often in the initial successive choice task and
in the later choice allocation task (hypothesis 1). Further,
we show that a frequency effect in choice occurs concur-
rently with a depth effect for judged average price (hy-
pothesis 2a).
Past research assumed that consumers’ perceptions of re-
tailers’ average price influences their retailer choice. Four
findings of study 1A suggest this is not the case under price
uncertainty. First, although the frequency retailer was chosen
more often, it was judged to have a higher average price.
Second, the correlation between average price judgments
and choice share in the 100 successive choices task was not
significant. Third, perceived average price judgments only
influenced future choice when the judgment task preceded
the choice task; and fourth, only one participant reported
using a decision rule that involved relying on average prices.
Since average price perceptions did not drive choice, what
did? Since prices for both retailers were fully revealed in
both conditions, the sampled and the actual price distribu-
tions did not differ in this study, and therefore undersam-
pling of deep discounts cannot explain the findings. Partic-
ipants’ reported decision rules provide evidence that some
are making predictions about which store will be cheaper
on a trial-by-trial basis and others simply stated they went
with the retailer that they perceived to be cheaper more
often. Both of these strategies would result in a frequency
effect in choice. Note that while in both cases participants
discuss selecting a store they expect to be cheaper, we cannot
tell whether they are driven by more frequent lower prices
or more frequent lower deals since they co-occurred here.
We examine this next.
STUDY 1B: EDLP (CONSTANT PRICE)
VERSUS DEPTH PRICING
Study 1B builds from 1A by examining whether the re-
sults replicate with a different retailing pricing strategy, ev-
eryday low pricing (EDLP), and in the more realistic case
when there is partial feedback and consumers only see the
prices of their chosen retailer but not of the forgone retailer.
Many of today’s most successful retailers, such as Walmart
and Costco, profess to use an EDLP strategy where they
offer low prices consistently. Are these retailers so attractive
to consumers because consumers believe they offer lower
prices on average than other retailers? Or, as the results of
study 1A suggest, are they attractive because consumers
believe their prices are cheaper more often than those of the
competitors?
In the real world it may be difficult to determine the
relative contribution of these two factors since many EDLP
retailers are both cheaper on average and more often than
their competitors. To disentangle these factors, we examine
choice under uncertainty between an EDLP retailer (that
offers a constant low price) and a depth retailer, while keep-
ing the average price of the retailers the same. Importantly,
this study also allows us to further understand why con-
sumers choose frequency retailers over depth retailers. Con-
sumers may choose frequency retailers because they are
cheaper more often or because they discount more often. If
consumers choose frequency retailers because they are
cheaper more often, they should choose an EDLP retailer
(that does not discount at all but frequently offers lower
prices than the depth retailer) over a depth retailer. However,

DANZIGER, HADAR, AND MORWITZ 767
if consumers choose frequency retailers because they dis-
count more often, they should choose the depth retailer be-
cause only it offers discounts.
Method
Ninety-five students at Tel-Aviv University participated
in exchange for 10 NIS (≈$3). The study used a 2 #2
mixed design with pricing pattern as a within-subject factor
(EDLP vs. depth) and feedback as a between-subjects factor
(partial vs. complete), to which participants were randomly
assigned.
The EDLP retailer priced a pack of portobello mushrooms
at 8.39 NIS for all 100 weeks. The depth retailer priced this
product regularly at 9.89 NIS for 70 weeks and discounted
it at 4.89 NIS on the remaining 30 weeks. The average price
offered by both retailers was 8.39 NIS. The same price
sequence was presented to all participants. The discounts
were distributed uniformly throughout the 100 trials. Par-
ticipants made all 100 successive choices (trials) under price
uncertainty. Participants were provided either with complete
or partial feedback. Finally, participants judged the average
price of both retailers.
Results
100 Successive Choices. Participants chose the EDLP
retailer on 66% of the trials in the partial feedback condition
(SD p19%, median p63%) and on 64% of the trials in
the complete feedback condition (SD p14%, median p
65%). These values, which do not significantly differ from
each other (t(93) p⫺0.67, NS), are both significantly
greater than 50% (partial feedback: t(46) p5.7, p!.0001,
complete feedback: t(47) p6.9, p!.0001).
Sampling Error and Perceived Average Price. The data
of 11 participants whose average price judgment of the
EDLP and/or depth retailers fell outside the price range
offered by the retailers were excluded from subsequent anal-
yses. Because some participants were provided with partial
feedback in this study, we tested for sampling error. In real
life, where people typically receive feedback from only the
chosen retailer (partial feedback), the price distributions con-
sumers experience may differ markedly from the retailers’
actual price distributions because of sampling error. In par-
ticular, it follows from the binomial distribution that low-
frequency prices tend to be undersampled, whereas high-
frequency prices tend to be oversampled, and that this
sampling error is more pronounced the smaller the sample
size and the lower the frequency of a price point (Yechiam
and Busemeyer 2006). This characteristic of the binomial
distribution implies that consumers may undersample dis-
counts offered by depth retailers. In other words, due to
sampling error, fewer visits to a store decrease thelikelihood
that consumers will experience a discount, especially when
the discount frequency is low, as is the case with a depth
retailer. This notion puts depth retailers in a far more inferior
position than previous research has observed. Our findings
confirm this expectation. Because sampling error could oc-
cur only for the depth retailer in the partial feedback con-
dition, we tested for it in this condition only. The average
sampled price for the depth retailer was significantly higher
than the actual average price (M
depth_sampled
p8.65, M
depth_actual
p8.39; t(1, 43) p2.86, p!.01).
Next, we submitted participants’ judged average prices
to an ANOVA with feedback (partial vs. complete) as an
independent variable and pricing pattern (EDLP vs. depth)
as a repeated variable. The analysis revealed a significant
depth effect (M
depth
p7.87, M
EDLP
p8.56; F(1, 81) p24.9,
p!.0001). The main effects of feedback and the interaction
were not significant. Thus, although the average of the sam-
pled prices was higher for the depth retailer than for the
EDLP retailer, participants perceived the depth retailer to
have a lower average price.
Discussion
These results suggest that the frequency effect in choice
observed in study 1A was driven by participants’ tendency
to choose the retailer they believed was cheaper more often
and not by their tendency to choose the retailer they believed
discounted more frequently. If retailer choice was driven by
discount frequency, then the depth retailer should have been
chosen most often since only it offered discounts. We also
show that the tendency to choose the retailer that is cheaper
more often occurs for both partial and complete feedback.
Consistent with the results of study 1A, we find that par-
ticipants chose the EDLP retailer more often yet judged its
average price to be higher.
STUDY 1C: EDLP (CONSTANT PRICE)
VERSUS FREQUENCY PRICING
Participants in study 1B tended to choose an EDLP re-
tailer over a depth retailer. Although we proposed that this
is because the EDLP retailer was cheaper more often, it may
also be because its prices did not vary. To rule this out we
next pitted an EDLP retailer that offered a constant price
against a frequency retailer (whose frequent discounted price
was lower than the price of the constant-price retailer). The
average price of both retailers was the same. If participants
in study 1B tended to choose the EDLP retailer because it
was cheaper more often, they should now tend to choose
the frequency retailer. If however, they chose the EDLP
retailer because it offered a constant price, they should now
choose the EDLP retailer, even though the frequency retailer
was cheaper more often.
Method
Thirty-one students at IDC Herzliya participated for
course credit. The design and procedure were slightly mod-
ified from those used in study 1B. First, the depth retailer
was replaced by a frequency retailer that offered the product
regularly at 9.89 NIS for 30 weeks and discounted it to 7.75
NIS on 70 weeks. Second, because we did not find any

768 JOURNAL OF CONSUMER RESEARCH
effects of feedback (partial vs. complete) on choice or av-
erage price judgments in study 1B, we provided only partial
feedback in this study. Finally, as in study 1A, we asked
participants to describe the choice strategy they used to make
their 100 choices.
Results
100 Successive Choices. Participants were more likely
to choose the frequency retailer over the EDLP (constant-
price) retailer. On average, the frequency retailer waschosen
on 58% of the trials (SD p17%, median p62%), which
is significantly greater than 50% (t(30) p2.6, p!.02).
Sampling Error and Perceived Average Price. Because
sampling (choices) could result in sampling error only for
the frequency retailer, we tested for sampling error in this
condition only. Consistent with properties of the binomial
distribution, where sampling error is not expected for fre-
quently experienced discounts, this difference was not sig-
nificant (t!1).
The judged average price of the frequency retailer was
not significantly different from the judged average price of
the constant-price retailer (M
frequency
p8.46, M
EDLP
p8.43;
F(29) p.28, NS).
Decision Rule. Participants’ descriptions of their choice
rule in the successive choice task were classified by two
independent coders to one of the same categories as in study
1A. Eighteen participants reported using a prediction strat-
egy, 1 used an average price strategy, 1 used a frequency
strategy, and 11 were classified as “other.”
Discussion
The results of study 1C provide further evidence that
consumers tend to choose the retailer that is cheaper more
often. The findings of study 1C together with those of study
1B indicate that retailers that offer constant prices are chosen
more often only when they are cheaper more often than their
competitors. As in studies 1A and 1B, the retailer with the
greater choice share was not judged to have a lower average
price. Interestingly, although the frequency retailer’s price
distribution was dichotomous and this retailer offered the
lowest price, its judged average price was not significantly
lower than that of the constant-price retailer. This may be
because the discount price offered by the frequency retailer
was not a sufficiently low anchor to sway average price judg-
ments downward, consistent with prior work showing that
the salience of deep discounts may be a key factor influencing
average price judgments (Krishna and Johar 1996). Finally,
as in study 1A, many participants reported using a prediction
strategy, which should, and did, result in a higherchoice share
for the frequency retailer, and only one participant based the
choice on average prices. Jointly, the results of studies 1A–
1C indicate that the likelihood of being cheaper drives choice
share, while discount magnitude drives judgments of average
price.
These studies also provide preliminary evidence that while
average price may not be the internal reference price that
drives retailer choice under price uncertainty, expected price
on a given occasion may be. These studies’ results are also
consistent with the notion that consumers compare the chosen
retailer’s price with that of the forgone retailer and code lower
prices as gains, and because segregating multiple small gains
provides more value than fewer larger gains, they tend to
choose the retailer that they perceive offers more frequent
lower prices. Next, we examine whether the observed effects
replicate over a broader set of conditions, and we attempt to
obtain stronger evidence for what drives the observed effects.
STUDY 2: NONDICHOTOMOUS PRICE
DISTRIBUTIONS AND PRICE
DISCRIMINABILITY
In studies 1A–1C, retailers’ prices were either constant
or dichotomous. However, in real life consumers may also
encounter more complex, nondichotomous price distribu-
tions. In study 2, we generalize our findings to nondicho-
tomous price distributions. We expect a frequency effect for
choice under price uncertainty. In line with Lalwani and
Monroe (2005), we expect judgments of average price to
be lower for the frequency retailer when discount frequency
is made salient (low price discriminability) and to be lower
for the depth retailer when discount magnitude is made sa-
lient (high price discriminability). Importantly,showing that,
under uncertainty, price discriminability does not influence
choice but does influence perceived average prices would
further support our claim that perceived average prices do
not underlie choice under uncertainty.
Method
Fifty-nine students at Ben-Gurion University participated
in return for 10 NIS (≈$3). The study used a 2 #2 mixed
design with a two-level, within-subject factor (pricing pat-
tern: frequency vs. depth) and an additional between-sub-
jects factor (price discriminability: low vs. high), to which
participants were randomly assigned. The procedure was
similar to the choice under price uncertainty condition of
the previous studies.
The product was 1 kilogram of avocados. Price distributions
were nondichotomous, and only partial feedback was given.
The price distributions in the low discriminability condition
were composed of decimal numbers, as in the previous studies.
In the high discriminability condition, these prices were mul-
tiplied by 100, as in Lalwani and Monroe (2005, study 2).
Participants in the high discriminability condition were asked
to imagine purchasing avocados in the Dori currency worth
1% of a NIS (1 NIS p100 Dori).
In the low discriminability condition, the frequency re-
tailer’s price was 9.89 NIS for 50 weeks and was on sale
for the remaining 50 weeks at prices that ranged between
6.79 NIS and 9 NIS. The depth retailer’s regular price of
9.89 NIS was offered for 80 weeks, and it was offered for
the remaining 20 weeks at a sale price that ranged between

DANZIGER, HADAR, AND MORWITZ 769
4.19 and 5.60 NIS. Both retailers’ average price was 8.89
NIS. Within each condition the same price sequence was
presented to all participants. The discounts were distributed
uniformly throughout the 100 trials. In the high discrimin-
ability condition, the comparable prices in the low discri-
minability price distributions were multiplied by 100. In this
study, retailers could simultaneously offer a sale. Conse-
quently, the frequency retailer’s price was cheaper on 39%
of the trials, the depth retailer’s price was cheaper on 20%
of the trials, and both retailers offered the same price on
41% of the trials.
Results
100 Successive Choices. Two participants whose pro-
portion of choosing the frequency retailer was more than
three standard deviations above the mean were removed
from the analysis. As in study 1A there was a significant
frequency effect: the frequency retailer was chosen on 57%
of the trials (SD p9%, median p55%). This value is
significantly greater than 50% (t(56) p5.6, p!.0001).
Choice of the frequency retailer did not differ across the
two price discriminability conditions (low discriminability:
M
frequency
p58%; t(29) p4.1, p!.001; high discrimina-
bility: M
frequency
p56%; t(26) p3.9; p!.001), which did
not differ from each other (t(57) p1.33, NS). Thus, the
frequency effect is robust to the discriminability of price
distributions.
Sampling Error and Perceived Average Price. The data
of one participant whose average price judgment of the fre-
quency retailer fell below its lowest price offered was ex-
cluded from further analyses. Sampled and judged average
prices in the high discriminability condition were divided
by 100 to allow comparison with those in the low discri-
minability condition. We submitted the average of the sam-
pled prices to an ANOVA with price discriminability (low
vs. high) as an independent variable and pricing pattern
(frequency vs. depth) as a within-subjects variable. None of
the effects were significant.
Next, we submitted participants’ average price judgments
to the same ANOVA. Perceived average price was signif-
icantly lower in the high than in the low discriminability
condition (M
high
p8.11, M
low
p8.43; F(1, 56) p5.3, p
!.03). This may be because high discriminability increases
discount salience and therefore “lowers” perceived average
price. Although the price discriminability #pricing pattern
interaction was not significant (F(1, 56) p1.0, NS), ad-
ditional analyses revealed a marginally significant frequency
effect in the low discriminability condition consistent with
hypothesis 2b (M
frequency
p8.27, M
depth
p8.60; F(1, 56) p
3.8, pp.056) and similar perceived average prices for the
frequency and depth retailer in the high discriminability con-
dition (M
frequency
p8.07, M
depth
p8.15; F(1, 56) p0.2,
NS). The main effect of pricing pattern was not significant
(F(1, 56) p2.6, pp.11).
Discussion
These results demonstrate that the frequency effect in
choice under uncertainty is robust to the complexity and
discriminability of price distributions. Consistent with pre-
vious research conducted under price certainty, we found a
frequency effect for perceived average price in the low dis-
criminability condition under price uncertainty and partial
feedback. While we did not find a depth effect for perceived
average prices in the high discriminability condition (Lal-
wani and Monroe 2005), there was also no frequency effect.
As in Lalwani and Monroe (2005), the depth retailer’s dis-
counts swayed average price judgments more in the high
discriminality condition than in the low discriminability con-
dition. Finally, the fact that discriminability influenced re-
tailer choice and price judgments differently further indi-
cates that different factors drive these two and that average
price judgments do not seem to drive choice.
STUDY 3: A 1-DAY GAP
BETWEEN CHOICES
In our studies, in prior studies on price perception (Alba
et al. 1994, 1999; Lalwani and Monroe 2005), and in studies
on experience-based choice (Erev and Barron 2005; Erev
and Haruvey 2008; Gonzalez and Dutt 2011), participants’
experience was operationalized as a series of choices made
in (relatively) rapid succession. Yet in real life the con-
sumer’s decision of which store to visit is typically made
on a daily, weekly, or monthly basis. There are several rea-
sons why temporally separated experience-based decisions
may differ from rapid successive decisions done in a lab.
First, in the lab, prices associated with the most recent
choices may more strongly influence successive choices be-
cause they may be more accessible in working memory.
Second, the encoding of prices in long-term memory may
be more likely when choices are separated due to the more
deliberate and less speedy nature of these choices. However,
it is also possible that, because of interfering information
from other sources, consumers find it more difficult to
remember prior prices as the time lag between decisions
increases, causing them to rely more heavily on heuristics
to govern future choices, such as the frequency heuristic
(Alba et al. 1994, 1999; Pansky and Algom 2002). Finally,
respondents may choose more automatically when choices
are made in rapid succession. Under a more deliberate
mind-set, consumers making a single choice a day may
call upon more complex internal price representations to
govern their choices. Due to these potential differences
between successive and separated choices, and to provide
external validity to the findings of the study 1 and study
2, in study 3 we asked participants to make a series of 15
daily choices between an EDLP retailer and a depth retailer.
In addition, to better examine which internal price repre-
sentations drive choice, participants in this study provided
three price judgments: judging average price, as in previous
studies; judging which retailer offers cheaper prices more
often, which we propose may be the representation that

770 JOURNAL OF CONSUMER RESEARCH
FIGURE 1
PROPORTION OF EDLP CHOICES (STUDY 3)
N
OTE
.—The choice share for the EDLP retailer was 46% on day
1 and increased to 73% on day 15. This pattern resembles thatfound
in previous experience-based choice studies where the choice
proportion on trial 1 is about 50% and quickly stabilizes on a higher
choice proportion for one of the available options (e.g., Barron and
Erev 2003).
drives retailer choice under price uncertainty; and judging
each retailer’s cheapest price, which we used as a proxy
for perceived discount depth.
Method
We recruited 60 US Amazon Mechanical Turkers to com-
plete 15 daily surveys in return for a payment of $8. Fifty-
five participants completed the 15 days of the survey (62%
female; M
age
p34.7, SD p11.9). Pricing pattern (EDLP
vs. depth) was manipulated within subjects.
Participants were asked to make 15 daily purchases of a
fresh 12-inch cheese pizza made by a particular brand sold
by two competing grocers. Participants were instructed that
they should base their choices solely on price and that the
price charged by each grocer may change on a daily basis.
They were incentivized to choose what they believed to be
the cheaper grocer by being informed that the four partic-
ipants with the lowest overall spending would receive an
additional $5 in compensation.
The EDLP retailer priced the product at $8.39 every day.
The depth retailer priced the product at $9.89 for 10 days
and at a discounted price of $4.89 for the remaining 5 days.
The average prices for the EDLP and the depth retailer were
$8.39 and $8.22, respectively. A single price sequence was
formed, with the depth retailer’s discounts randomly dis-
tributed across the 15 days. All choices were made under
price uncertainty, and only partial feedback was provided.
After participants had completed their purchase on the
fifteenth day, they judged each retailer’s average price and
indicated the cheapest price offered by each retailer, as well
as which retailer they thought was cheaper more often, on
a 10-point scale (1 pretailer A [denoting the depth retailer],
5pboth retailers were similar, 10 pretailer B [denoting
the EDLP retailer]).
Results
Given 15 choices, the participants revealed a clear ten-
dency to choose the EDLP retailer. Specifically, the EDLP
retailer was chosen on 73% of the trials (SD p21%, median
p85%). This value is significantly greater than 50% (t(54)
p8.2, p!.0001). Figure 1 depicts the EDLP retailer’s
choice share along the 15 days of the survey. This result
provides strong external validity for our earlier findings.
Sampling Error and Perceived Average Price. Because
sampling (choices) could result in sampling error only for
the depth retailer, we tested for sampling error in this con-
dition only. While the actual average price for the depth
retailer was $8.22 (and lower than the EDLP price), the
average experienced price was significantly higher, $9.16
(t(54) p6.64, p!.0001).
Because the average experienced price for the two re-
tailers differed markedly in this study (EDLP, $8.39, vs.
depth, $9.16), to reveal possible biases in participants’ av-
erage price judgments, we compared the experienced and
the judged average prices for each retailer. Participants’ per-
ceived average price for the EDLP retailer was $8.40, which
did not significantly differ from the experienced price of
$8.39 (t(54) p0.35, NS). For the depth retailer, the per-
ceived average price was $9.11, which also did not signif-
icantly differ from the experienced average price of $9.16
(t(54) p⫺0.38, NS). Thus, in this study with dichotomous
price distributions, the depth discount was not judged to
have lower average prices. This may be because in this study
the depth discount was rarely experienced (it was experi-
enced once on average) and possibly also because the av-
erage number of days since it was last experienced before
participants provided their judgment of the average price
was 4 days, leading to a weak memory trace.
To examine which internal price representations most
strongly impact choice, we regressed participant’s choice
share for the EDLP retailer on the difference between the
actually experienced average prices of the two retailers (ac-
tual average price), the difference between the perceived
average prices of the two retailers (perceived average price),
the difference between the two retailers in the perceived
cheapest price (discount magnitude), and participants’ judg-
ment of which retailer was cheaper more often (cheaper
more often). Only the difference between the actually ex-
perienced average prices of the two retailers (bp⫺.32,
t(50) p⫺2.7, p!.02) and the judgment of which retailer
was cheaper more often (bp.37, t(50) p2.6, p!.02)
were significant predictors of choice share. Specifically, the
lower the experienced average price of the EDLP retailer
compared to the depth retailer, the more likely consumers
were to choose the EDLP retailer. Also, the more the EDLP

DANZIGER, HADAR, AND MORWITZ 771
retailer was perceived as cheaper more often, the more likely
participants were to choose it over the depth retailer.
Given the high correlation between the actually experi-
enced average prices and the perceived average prices of
the two retailers, it is not surprising that in the presence of
the difference between actually experienced average prices,
the difference in judged average prices did not predict choice
(bp–.18, t(50) p–1.1, NS). To examine the extent to
which multicollinearity affects our ability to reliably inter-
pret the regression results, we calculated VIF (variance in-
flation factor) and tolerance values. The VIF and tolerance
values for the difference between the experienced average
prices of the constant and the depth retailer were 1.97 and
0.51, respectively; the VIF and tolerance values for the dif-
ference between the perceived average prices of the constant
and the depth retailer were 3.81 and 0.26, respectively. These
values suggest that although multicollinearity exists in our
data, it is not extremely high, and we thus consider the
above-described results as sufficiently reliable.
Finally, the finding that the difference between the EDLP
and the depth retailers in the perceived cheapest price did not
predict choice (bp⫺.06, t(50) p⫺0.5, NS) indicates that
the magnitude of the discount offered by the depth retailer
did not predict choice.
Discussion. The findings of study 3 replicate and provide
external validity to those of studies 1 and 2 by showing that
when a day-long gap separates choices, consumers still are
more likely to choose the retailer that is cheaper more often
and by showing that average price judgments do notunderlie
choice. Rather, the results of the regression analysis show
that consumers are more likely to choose a retailer that they
perceive to be cheaper more often.
The results of this study also demonstrate more strongly than
those of the previous studies an important disadvantage of using
a depth discounting strategy. When consumers’ choices are
spaced out over time, they are made under price uncertainty,
and the sample size is small, the prices consumers experience
for the depth retailer may be far less attractive than its actual
prices. Moreover, the results again demonstrate a clear advan-
tage of being cheaper more often and suggest that choice may
be driven by participants’ belief regarding the relative frequency
with which a retailer is cheaper.
STUDY 4: PREDICTED PRICES
The results of the previous studies have consistently shown
that consumers are more likely to choose frequency (or EDLP)
over depth retailers; at the same time, their judgments suggest
that their average price perceptions do not underlie their
choices. Which price perceptions do? Consistent with the results
of study 3, when asked in studies 1A and 1C, most participants
indicated that they predicted which retailer would be cheaper
on the following trial and chose the one they thought would
be cheaper on the next trial (prediction strategy) or most of the
time (frequency strategy). In this study, we directly examine
whether predicted prices drive choice by asking the participant
every 20 trials either to predict each retailer’s next price or to
report each retailer’s average price, and we then correlate these
judgments with the participant’s following choice. Note that a
prediction strategy entails that in shopping trips where con-
sumers expect the frequency retailer to be cheaper, they will
be more likely to choose it, and when they expect the depth
retailer to be cheaper, they will be more likely to choose it. We
predict that overall this will lead to an aggregate tendency to
choose the frequency retailer since it is expected to be cheaper
than the depth retailer on more of the trials.
Method
Fifty-six students from IDC Herzliya participated for
course credit. The study used a 3 #2 mixed design with
a two-level, within-subject factor (pricing pattern: frequency
vs. depth) and a between-subjects factor (interchoice price
judgments: none, predicted retailer’s prices, or average re-
tailer’s prices), to which subjects were randomly assigned.
The frequency retailer priced a pack of portobello mush-
rooms at 9.89 NIS for 40 weeks and at the sale price of 8.95
NIS for the remaining 160 weeks. The depth retailer priced
it at 9.89 NIS for 160 weeks and at the sale price of 6.14
NIS for the remaining 40 weeks. The two retailers had the
same average price of 9.14 NIS. All participants saw the same
price sequence, which was constructed such that the depth
retailer offered a discount price every fourth, fifth, or sixth
trial (randomly determined) and the frequency retailer offered
a discount price in all other trials. Thus, a discount was offered
by one of the two retailers on every trial.
Before making every twentieth choice (i.e., 20, 40, 60,
. . . , 200), one randomly selected set of participants was
asked to predict the next price offered by each retailer; an-
other was asked to estimate the average price (up to that
point) offered by each retailer; and members of the last set
of participants, the control set, were not asked to make any
judgments throughout the series of 200 choices.
The procedure was similar to that used in the previous
studies. Participants were provided with complete feedback
in every condition. After completing the choice task, par-
ticipants judged each retailer’s average product price.
Results
200 Successive Choices. There was a significant fre-
quency effect. The frequency retailer was chosen on 75%
of the trials (SD p5%, median p75%). This value is
significantly greater than 50% (t(55) p34.4, p!.0001).
Importantly, choice of the frequency retailer did not differ
across the three interchoice price judgments conditions (con-
trol: M
frequency
p74%; t(19) p22, p!.0001; predicted
retailer prices: M
frequency
p75%; t(17) p25.6, p!.0001;
average retailer prices: M
frequency
p76%; t(17) p15.3, p!
.0001), which were not significantly different from each
other (F(2, 53) p.3, NS).
Perceived Average Price. The data of seven participants
whose average price judgments fell outside the price range
offered by at least one of the retailers were excluded from

772 JOURNAL OF CONSUMER RESEARCH
subsequent analyses. Judged average prices were submitted
to an ANOVA with interchoice price judgments (none, pre-
dicted retailer prices, or average retailer prices) as an in-
dependent variable and pricing pattern (frequency vs. depth)
as a repeated variable. As in the prior studies, there was a
depth effect (M
frequency
p9.06 vs. M
depth
p8.79; F(1, 44) p
4.9, pp.03). No other effects were significant.
Interchoice Judgments and Subsequent Choice. We ex-
amined whether each of the 20 interchoice judgments suc-
cessfully predicted subsequent choice. A successful predic-
tion occurred when participants chose the retailer they
predicted would offer the cheaper price on the next trial (in
the predicted price condition) or the one they thought offered
a lower average price up to that point (in the average price
condition). Ties between retailer judgments (predicted or
average price) were excluded from this analysis. We cal-
culated the proportion of successful predictions per partic-
ipant (out of 10 trials, excluding ties). Predicted retailer price
judgments were more predictive of subsequent choice than
average retailer price judgments (M
prediction
p81.3%, M
average
p55.3%; t(34) p⫺2.58, p!.02).
Discussion
As in the previous studies, these results show that under
price uncertainty, participants are more likely to choose the
frequency retailer over the depth retailer. Importantly, these
results contrast two price perceptions that may underlie
choice: average price perception, implicated in past research,
and predicted prices, which participants described as their
choice strategy in the previous studies and also as implicated
by the literatures on prediction and experience-based choice.
We find evidence that predicted prices seem to underlie
choice under uncertainty and average price perceptions do
not.
GENERAL DISCUSSION
We examined how retailer pricing strategy influences con-
sumers’ retailer choice and price perceptions under price
uncertainty. Simulating everyday choice, in a series of stud-
ies participants made retailer choices where on each occa-
sion they chose a retailer and only then learned product
prices, with either partial or complete feedback. Participants
consistently tended to choose the retailer that was cheaper
more often, even when they judged this retailer to be more
expensive on average. Most participants indicated using a
prediction strategy where they chose the retailer they be-
lieved would be cheaper on each shopping occasion or was
cheaper more often.
A tendency to choose the retailer that was cheaper more
often was observed under conditions of dichotomous price
distributions with complete feedback (study 1A) and was
found for low and high discount discriminability conditions
for more complex nondichotomous distributions (study 2).
Studies 1A and 1B illuminate why consumers tend to choose
the frequency retailer. These studies show that being cheaper
more often, and not offering frequent discounts, drives choice.
Study 3 extended our findings from a setting in which par-
ticipants made multiple successive choices to a more realistic
environment where they made one choice per day, and this
replicated the earlier findings. Finally, study 4 demonstrated
that predicted prices seem to underlie choice under price un-
certainty, while average price judgments do not.
The superiority of offering frequently lower prices was
demonstrated with different price distributions, different
product categories, and different time lags between choices,
and also across different participant populations (American
adults and Israeli undergraduate students) and different sur-
veying methods (laboratory experiments and an online sur-
vey). The managerial implications are clear. Under price
uncertainty, when the consumers’ goal is to maximize sav-
ings, they will tend to choose the retailer they believe is
cheaper more often.
Our results also extend previous price perception research.
We show that average price judgments following successive
choices made under price uncertainty are generally quite
similar to those that follow successive choices made under
price certainty. More importantly, however, our findings
demonstrate that consumer beliefs regarding the propensity
with which each retailer is cheaper, in general or on a trial-
by-trial basis, and not average price perceptions, drive re-
tailer choice under price uncertainty. Several findings sup-
port this conclusion. First, in every study the retailer that
was cheaper more often was chosen more often. Second,
participants reported using choice strategies linked to the
propensity of being cheaper and not average prices (studies
1A and 1C). Third, the results of study 3 indicate that con-
sumer choice is driven by perceptions of which retailer offers
cheaper prices more often. Finally, the results of study 4
demonstrate the contribution of trial-by-trial price expec-
tations.
We offered several explanations for why consumers
choose the retailer offering cheaper prices more often. Our
findings are consistent with the mental accounting expla-
nation at the aggregate group level. In all studies, partici-
pants’ tendency to choose the retailer that was cheaper more
often suggests that they preferred the retailer that was more
likely to yield a positive outcome and that they preferred
to accrue many small gains over a few large ones. Future
research could seek stronger evidence for these accounts
through individual-level analyses. For example, if the val-
uation of outcomes contributes to the tendency to choose
the frequency retailer, participants with a more concave/
convex value function should reveal a stronger frequency
effect. Similarly, individual measures of loss aversion should
positively correlate with the tendency to choose the fre-
quency retailer. Importantly, consumers who more linearly
value outcomes or who are less loss averse may not reveal
any strong preference for either frequency or depth retailers.
Our findings also provide direct evidence for the predic-
tion account at both the individual and aggregate levels. The
results of study 4 show that participants were quite accurate
in their price predictions and that those explained subsequent

DANZIGER, HADAR, AND MORWITZ 773
choice. In addition, participants’ descriptions of their de-
cision rule in studies 1A and 1C show that many participants
explicitly mentioned using a prediction strategy. Finally, we
show a significant effect of sampling error, which, as we
have mentioned, has important practical implications. In
studies 1B (one condition), 1C, and 3, in which participants
were given partial feedback, the results indicate that partic-
ipants undersampled the discounts of the depth retailer,
something that may have reduced this retailer’s attractive-
ness. However, the existence of a frequency effect in the
studies in which participants were given full feedback (stud-
ies 1A, one condition in study 1B, and study 4) suggests
that sampling error alone does not drive consumers’ ten-
dency to select the retailer that is cheaper more often. While
sampling error does not seem to drive the frequency effect,
underweighting of small probability outcomes may. Future
research may estimate individuals’ weighting function (Ab-
dellaoui, L’Haridon, and Paraschiv 2011) and examine its
correlation with the magnitude of the frequency effect.
Our research also contributes to the experience-based
choice literature by providing external validity to one of its
main findings. To our knowledge, all previous experience-
based choice studies have examined repeated choice under
uncertainty by having participants make multiple consecu-
tive choices in a single sitting. In study 3, we find, with a
day’s lag between choices and a few choices to make, a
tendency for participants to select an option that offers many
small gains (relative to the competing option) over a com-
peting option that offers infrequent large gains.
Although we tried to simulate everyday experience-based
choice, the current paradigm still clearly represents a sim-
plification of real-life purchasing. For example, while we
asked participants to purchase a single item on each shop-
ping trip, consumers typically purchase a basket of goods
on each trip. When shopping for a basket of goods, factors
that influence price perceptions and choice are more com-
plex (Bu¨yu¨kkurt 1986). For example, to determine a re-
tailer’s overall price level, a consumer may rely on the total
basket price, or on a representative sample of product prices,
or even on the price of a single product. Tracking prices of
a number of items is a far more challenging task than track-
ing the price of a single item, and consumers may thus use
different strategies and rely on different representations to
simplify this task. For example, instead of choosing the
retailer that offers cheaper prices on a single item, more
often they may choose a retailer that offers discounts on
more items in the basket. Alternatively, consumers may de-
cide to simplify their task by relying only on their percep-
tions of the total basket price. Past modeling papers have
assumed that consumers have rational expectations of future
prices for a basket of goods (Lal and Rao 1997) and have
shown that price expectations for a basket of goods can
influence choice (Bell and Lattin 1998). More behavioral
research is needed to understand how consumers form these
price perceptions for a basket of goods, how that varies with
basket size (Bell and Lattin 1998), and whether those per-
ceptions vary across segments of consumers who differ in
their proneness for different types of deals (Pechtl 2004).
Future research should also examine how robust our find-
ings are to other characteristics of shopping behavior. In all
of our studies, participants had a purchase goal of mini-
mizing spending, participants were required to make a pur-
chase on each visit, and they were not allowed to stockpile.
If, for example, we had changed the participants’ goal to
having fun, had allowed them to defer purchase on each
occasion, or had allowed them to stockpile, we may have
found a shift in their tendency to choose the depth retailer.
It seems likely that any manipulation that would increase
the utility of finding a product at an extremely low price
without incurring the cost of having to buy the product at
a relatively expensive price would bias choice in favor of
the depth retailer. Future research could examine this pre-
diction.
DATA COLLECTION INFORMATION
The first two authors supervised the collection of data for
studies 1–4 by research assistants at Tel-Aviv University
(study 1B during winter 2012), IDC Herzliya (studies 1C
and 4, during winter and spring of 2013), and Ben-Gurion
University (studies 1A and 2 during autumn 2009 through
spring 2010). The first and second authors jointly analyzed
these data. The second author managed the collection of
data for study 3 using the Amazon Mechanical Turk panel
described in the methods section in the spring of 2012. These
data were analyzed jointly by the first and second authors.
REFERENCES
Abdellaoui, Mohammed, Olivier L’Haridon, and Corina Paraschiv
(2011), “Experienced versus Described Uncertainty: Do We
Need Two Prospect Theory Specifications?” Management Sci-
ence, 57 (October), 1879–95.
Alba, Joseph W., Susan M. Broniarczyk, Terence A. Shimp, and
Joel E. Urbany (1994), “The Influence of Prior Beliefs, Fre-
quency Cues, and Magnitude Cues on Consumers’ Percep-
tions of Comparative Price Data,” Journal of Consumer Re-
search, 21 (September), 219–35.
Alba, Joseph W., Carl F. Mela, Terence A. Shimp, and Joel E.
Urbany (1999), “The Effect of Discount Frequency and Depth
on Consumer Price Judgments,” Journal of Consumer Re-
search, 26 (2), 99–114.
Barron, Greg, and Ido Erev (2003), “Small Feedback-Based De-
cisions and Their Limited Correspondence to Description-
Based Decisions,” Journal of Behavioral Decision Making,
16 (3), 215–33.
Bell, David R., and James M. Lattin (1998), “Shopping Behavior
and Consumer Preference for Store Price Format: Why ‘Large
Basket’ Shoppers Prefer EDLP,” Marketing Science,17(1),
66–88.
Bloomfield, Robert, and Jeffrey Hales (2002), “Predicting the Next
Step of a Random Walk: Experimental Evidence of Regime-
Shifting Beliefs,” Journal of Financial Economics,65(3),
397–414.
Bu¨yu¨kkurt, B. Kemal (1986), “Integration of Serially Sampled

774 JOURNAL OF CONSUMER RESEARCH
Price Information: Modeling and Some Findings,” Journal of
Consumer Research, 13 (December), 357–73.
Danziger, Shai, and Ruthie Segev (2006), “The Effects of Infor-
mative and Non-informative Price Patterns on Consumer Price
Judgments,” Psychology and Marketing, 23 (June), 535–53.
Edwards, Ward (1962), “Dynamic Decision Theory and Probabi-
listic Information,” Human Factors,4(2),59–73.
Erev, Ido, and Greg Barron (2005), “Adaptation, Maximization,
and Reinforcement Learning among Cognitive Strategies,”
Psychological Review, 112 (4), 912–31.
Erev, Ido, and Ernan Haruvy (2008), “Learning and the Economics
of Small Decisions,” in The Handbook of Experimental Eco-
nomics, Vol. 2, ed. John H. Kagel and Alvin E. Roth, Prince-
ton, NJ: Princeton University Press.
Estes, William K. (1961), “A Descriptive Approach to the Dy-
namics of Choice Behavior,” Behavioral Science, 6 (3),
177–84.
Fox, Craig, and Liat Hadar (2006), “Decisions from Experience p
Sampling Error ⫹Prospect Theory: Reconsidering Hertwig,
Barron, Weber, and Erev (2004),” Judgment and Decision
Making, 1 (2), 159–61.
Gonzalez, Cleotilde, and Varun Dutt (2011), “Instance-Based
Learning: Integrating Decision from Experience in Sampling
and Repeated Choice Paradigms,” Psychological Review, 118
(4), 523–51.
Hertwig, Ralph, Greg Barron, Elke U. Weber, and Ido Erev (2004),
“Decision from Experience and the Effect of Rare Events in
Risky Choice,” Psychological Science, 15 (8), 534–39.
Hoch, Stephen J., Xavier Dre`ze, and Mary E. Purk (1994), “EDLP,
Hi-Lo, and Margin Arithmetic,” Journal of Marketing Re-
search, 58 (October), 16–27.
Humphreys, Lloyd G. (1939), “Acquisition and Extinction of Ver-
bal Expectations in a Situation Analogous to Conditioning,”
Journal of Experimental Psychology, 25 (3), 294–301.
Kahneman, Daniel, and Amos Tversky (1979), “Prospect Theory:
An Analysis of Decision under Risk,” Econometrica,47(2),
263–91.
Krishna, Aradhna (1991), “Effect of Dealing Patterns on Consumer
Perceptions of Deal Frequency and Willingness to Pay,” Jour-
nal of Marketing Research, 28 (November), 441–51.
Krishna, Aradhna, Richard Briesch, Donald R. Lehmann, and Hong
Yuan (2002), “A Meta-analysis of the Impact of Price Pre-
sentation on Perceived Savings,” Journal of Retailing,78(2),
101–18.
Krishna, Aradhna, and Gita V. Johar (1996), “Consumer Percep-
tions of Deals: Biasing Effects of Varying Deal Prices,” Jour-
nal of Experimental Psychology: Applied, 2 (3), 187–206.
Lal, Rajiv, and Ram Rao (1997), “Supermarket Competition: The
Case of Every Day Low Pricing,” Marketing Science,16(1),
60–80.
Lalwani, Ashok K., and Kent B. Monroe (2005), “A Reexamination
of Frequency-Depth Effects in Consumer Price Judgments,”
Journal of Consumer Research, 32 (December), 480–85.
Lattin, James L., and Randolph E. Bucklin (1989), “Reference
Effects of Price and Promotion on Brand Choice Behavior,”
Journal of Marketing Research, 26 (3), 299–310.
Monroe, Kent B. (2003), Pricing: Making Profitable Decisions,
3rd ed., Boston: McGraw-Hill/Irwin.
Pansky, Ainat, and Daniel Algom (2002), “Comparative Judgment
of Numerosity and Numerical Magnitude: Attention Preempts
Automaticity,” Journal of Experimental Psychology: Learn-
ing, Memory and Cognition, 28 (2), 259–74.
Payne, John (2005), “It Is Whether You Win or Lose: The Im-
portance of the Overall Probabilities of Winning or Losing
in Risky Choice,” Journal of Risk and Uncertainty,30(1),
5–19.
Pechtl, Hans (2004), “Profiling Intrinsic Deal Proneness for HILO
and EDLP Price Promotion Strategies,” Journal of Retailing
and Consumer Services, 11 (4), 223–33.
Shermer, Michael (2011), The Believing Brain, New York: Henry
Holt.
Taylor, Shelley E., Jennifer Crocker, Susan T. Fiske, Merle Sprin-
zen, and Joachim D. Winkler (1979), “The Generalizability
of Salience Effects,” Journal of Personality and Social Psy-
chology, 37 (3), 357–68.
Thaler, Richard H. (1985), “Mental Accounting and Consumer
Choice,” Marketing Science, 4 (Summer), 199–214.
Tversky, Amos, and Daniel Kahneman (1991), “Loss Aversion in
Riskless Choice: A Reference-Dependent Model,” Quarterly
Journal of Economics, 106 (4), 1039–61.
Ungemach, Christoph, Nick Chater, and Neil Stewart (2009), “Are
Probabilities Overweighted or Underweighted When Rare
Outcomes Are Experienced (Rarely)?” Psychological Science,
20 (4), 473–79.
Whitson, Jennifer A., and Adam D. Galinsky (2008), “Lacking
Control Increases Illusory Pattern Perception,” Science, 322
(October), 115–17.
Yechiam, Eldad, and Jerome R. Busemeyer (2006), “The Effect
of Forgone Payoffs on Underweighting of Small Probability
Events,” Journal of Behavioral Decision Making,19(1),
1–16.
Zeithaml, Valerie (1988), “Consumer Perceptionsof Price, Quality,
and Value: A Means-End Model and Synthesis of Evidence,”
Journal of Marketing,52(3),2–22.

CopyrightofJournalofConsumerResearchisthepropertyofJournalofConsumerResearch,
Inc.anditscontentmaynotbecopiedoremailedtomultiplesitesorpostedtoalistserv
withoutthecopyrightholder'sexpresswrittenpermission.However,usersmayprint,
download,oremailarticlesforindividualuse.