Fake Discounts Drive Real Revenues in Retail: Donald Ngwe
Fake Discounts Drive Real Revenues in Retail: Donald Ngwe
in Retail
Donald Ngwe
Donald Ngwe∗
Abstract
Prices in a wide variety of contexts are often presented in three parts: an original or sug-
gested list price, a discount off that price, and the final selling price. Limited empirical evi-
dence is available that speaks to the relative impact of each component on purchase behavior,
even as theories abound. Measuring these impacts is critically important to sellers, to con-
sumers, and to regulators who are keen on enforcing deceptive advertising guidelines against
“fictitious pricing”—the practice of quoting list prices that do not truthfully reflect prior sell-
ing prices. This paper uses a large retail transaction data set that features wide variations in
these pricing components within a relatively homogeneous product space. The data set has
the unique feature of containing sales records from outlet stores wherein a subset of products
have verifiably fictitious list prices and discounts, allowing for measurement of their impact
on purchase incidence in actual retail settings. By measuring the impact of different pricing
components on purchase likelihood, we find that list prices have a strong influence on purchase
outcomes, with a $1 increase in the list price having the same positive effect on purchase like-
lihood on average as a $0.77 decrease in the actual selling price. This effect is larger for fake
list prices, but smaller in longer-lived stores and stores closer to regular retail channels. In a
∗
E-mail: [email protected]. This paper is based on a chapter of my dissertation. I am grateful to my dissertation
committee members Chris Conlon, Brett Gordon, Kate Ho, Michael Riordan, and Scott Shriver for their guidance and
advice. Daniel Schwam provided excellent research assistance.
1
complementary laboratory experiment, we find that fake list prices have no impact on purchase
intent when consumers have full knowledge of the true original price. These results imply that
fake prices enhance demand by misleading consumers about true original prices.
1 Introduction
Virtually all firms engage in some form of discount pricing. There are several reasons that firms
might drop the price of a good over time: when it seeks to discriminate between consumers accord-
ing to their willingness to pay; as a means of managing its inventory; or when it resolves demand
uncertainty after the good’s introductory phase. In many of these instances, consumers can be
thought of as having nearly full information and making rational responses to price incentives. In
contrast, this paper focuses on motivations for discount pricing that arise from consumers having
imperfect information or possibly exhibiting irrational behavior. These are motivations that might
encourage firms to post high “original” prices at which products were never actually available for
purchase.
This practice, termed “fictitious pricing” by the Federal Trade Commission (FTC), occasionally
results in litigation. In one recent case, a class-action lawsuit was filed against Kohl’s Department
Stores in California for allegedly misstating in advertising that items had been marked down (Dolan
2013). In reversing an earlier district court dismissal of the case, the US Ninth Circuit Court of
Appeals stated that California consumer laws permit such lawsuits if the consumer would not have
made the purchase but for the perceived bargain. In December 2016, JCPenney, Sears, Macy’s,
and Kohl’s were all hit with lawsuits over similar allegations. JCPenney and Kohl’s had previously
been sued over the same practice and settled for $50 million and $6.5 million, respectively (Popken
2016).
Fictitious pricing is one area in which enforcement differences between states and the FTC loom
2
large. Between 1950 and 1970, 30% of FTC challenges to advertising were related to “fictitious
price claims” (Pitofsky et al. 2003). Since 1979, however, the FTC has not brought a single
fictitious pricing case to court. Several FTC chairs have indicated that enforcement actions in
this area have done more harm than good. This is because of the difficulty and arbitrariness with
which a “genuine” discount might be differentiated from a deceptive one. In the current state of
enforcement, state attorneys general apply state statutes in bringing cases; however, enforcement
is highly uneven, with major cases attracting widespread attention every few years but with many
firms consistently ignoring fictitious pricing statutes with impunity.
The main obstacle in litigating individual cases, and therefore in policing this practice more broadly
in the marketplace, stems from the difficulty in assessing private damages from fictitious pricing
(Friedman 2015). Various recent cases in Illinois against large retailers such as Joseph A. Bank
(Camasta v. Jos. A Bank Clothiers, Inc., 2013), Carter’s (Kim v. Carter’s, Inc., 2010), and QVC
(Mulligan v. QVC, Inc., 2008) were found in favor of the defendants due to the plaintiffs’ failure
in establishing economic harm due to false price claims. The question of whether consumers are
deceived by fake price claims is also a factor, with one court decision seeing no fault because
consumers “see through the ruse” (B. Sanfield, Inc. v. Finlay Fine Jewelry Corp., 1999).
In this paper, we identify patterns in how consumers respond to false and genuine discount claims
by the same seller. We use data from a dominant fashion goods retailer that makes heavy use of
discounting in its outlet stores. This data set offers an unprecedented opportunity to study this
pricing strategy because it records both list and selling prices, as well as repeat purchases. A
portion of these list prices are observably genuine, while the remainder are fictitious. As with
many brands and retailers, the firm implements a degree of randomized discounting across both
time and products in-store, providing much variation in final selling prices for each product.
While a number of earlier studies have examined the question of reference prices using laboratory
experiments or survey methodologies, there have been, to our knowledge, no prior studies that
establish the effect of fake list prices on purchase behavior in actual retail settings. There are sev-
3
eral benefits from using transaction data to address this question. Whereas survey methodologies
must rely on a constructed notion of what constitutes the actual “market price” for a particular
good, in our empirical setting such a price is directly observed from the data. Using sales data also
sidesteps the issue of ensuring incentive compatibility in measuring changes to consumers’ intent
to purchase. Perhaps the most important advantage of using sales data is the ability to measure
the effects across hundreds of different products, rather than relying on a few focal products in the
typical survey questionnaire.
We use a parsimonious discrete choice model of demand to estimate the effect of list prices on
purchase likelihood. Identification of this effect arises from variation in list prices within the same
style across different colors. The large number of observations, combined with the firm’s national
pricing policy, permits the estimation of style and color fixed effects, without the need to use
instrumental variables to control for endogeneity. We explore how responses to list prices vary
according to consumer subgroups by taking advantage of store openings and closures within the
sample duration.
We find that consumer responses to list prices are consistent with theories that treat prices as a
signal of product quality and that hypothesize reference-dependent behavior in how consumers
evaluate prices. Controlling for selling prices and other product characteristics, a higher list price
substantially increases a good’s purchase probability. Moreover, this effect is larger for products
with fake list prices than for those with genuine list prices. Importantly, this effect seems to be
moderated by the consumer’s level of information, with a lower estimated sensitivity to fake dis-
counts in longer-lived stores or outlet stores closer to regular stores.
We augment our analysis with findings from a laboratory experiment. We design the experiment
with the goal of measuring the effect of false original prices on purchase intent when subjects have
full knowledge of a product’s true original price. Our findings consistently indicate that keeping
selling price constant, a higher true original price results in higher purchase intent. However,
conditional on knowledge of the true original price, a higher false original price has zero effect on
4
purchase intent.
Taken together, evidence from our analysis of observational transaction data and our laboratory
experiment suggests that fake prices provide sellers with a powerful tool to enhance demand, but
one that may come at the expense of misleading consumers about products’ true initial selling
prices. Consumers take initial prices as signals of product quality and rate offers as being better
deals the higher these initial prices are with respect to present selling prices. Accordingly, fake
prices have the highest influence on purchase likelihood for less-informed consumers.
The paper proceeds as follows. Section 2 reviews the related literature on discount pricing and ref-
erence dependence. Section 3 describes the data used for the empirical analysis and provides some
descriptive statistics. Section 4 outlines a demand model and presents parameter estimates. Sec-
tion 5 reports findings from a laboratory experiment. Section 6 concludes and points to directions
for future work.
2 Related literature
This research relates to several literatures concerning price signals in retail settings. It takes its
basic premise from research that explores the function of price as a signal of quality. Because
our paper focuses on prices that are framed as discounts, it draws from research on the impact of
discount cues on shopper behavior and research on reference dependence more generally. Finally,
our findings relate to work on the welfare and regulatory implications of fake prices and misleading
sales. In this section, we review the most closely related papers in each of these literatures with an
eye toward highlighting our contributions.
There is a large body of research in marketing and economics that has studied how uninformed
consumers use prices as signals of product quality. High prices can serve as credible signals of
high product quality because the “consequent loss of sales volume is most damaging for lower-cost,
lower-quality products” (Bagwell & Riordan 1991). Within this paradigm, the profitability of using
5
price to signal quality depends on the relative proportions of informed and uninformed consumer
segments (Milgrom & Roberts 1986; Armstrong & Chen 2013). Our research does not rely on the
disciplining force of either informed consumers or marginal costs on prices; hence, it more closely
relates to research that treats high prices as strategic substitutes for advertising (e.g. Gerstner 1985;
Linnemer 2002). Unlike most signals available to firms as marketing-mix variables, however, fake
prices require no immediate costs to implement (Kirmani & Rao 2000). Our research thus aims
to determine which aspects of the extant thinking on the relationship between prices, perceived
quality, and perceived value extends to fake prices (Zeithaml 1988).
By definition, a fake price offers a fake discount—a discount that does not represent a decrease
from some previous selling price but, rather, the difference between the current selling price and a
fake introductory price. There is much existing literature on the impact of discounts on consumer
behavior beyond what may be ascribed to quality signaling (Blattberg, Briesch & Fox 1995).
Discounts have been shown to dampen further consumer search (Urbany, Bearden & Weilbaker
1988), cause consumers to weigh non-price product attributes more heavily (Bordalo, Gennaioli
& Shleifer 2013), and exploit bargain-hunting tendencies (Armstrong & Chen 2013). While these
effects have been explored where discounts are fake as well as where they are real, our research
focuses on the common yet unexplored setting in which discounts have substantial real and fake
components.
Efforts to contextualize the impact of discounts have often appealed to notions of reference depen-
dence (Tversky & Kahneman 1981; Thaler 1985). Researchers have identified several reference
points that sellers rely on: advertised prices (Biswas & Blair 1991), competitors’ prices (Della
Bitta, Monroe & McGinnis 1981), same-store offers (Rajendran & Tellis 1994), and fake prices
(Blair & Landon 1981). Previous work has also identified upper and lower bounds beyond which
discounts are less effective (Gupta & Cooper 1992). More recent work has posited a theoretical
explanation of reference dependence relying on opportunity cost considerations (Kim, Joo & Al-
lenby 2017). There has also been research that documents how certain pricing policies can have
6
large effects on consumer behavior due to local or cultural factors despite lacking a clear basis in
established theory, such as the choice of price-endings (Stiving 2000; Anderson & Simester 2003).
Our research differs from previous studies by focusing on the moderating role of what consumers
know about how truthfully a discount offer reflects prior selling prices.
In addition to legalistic debates on fake prices among policymakers, marketing scholars have also
conducted research on the impact of fake prices and comparative price advertising more generally
on consumer welfare (Compeau & Grewal 1998). A central question is whether fake prices are
deceptive, and if they are, how substantial the damage is on consumers (Grewal & Compeau 1992;
Compeau, Grewal & Chandrashekaran 2002; Rhodes & Wilson 2018). Our research demonstrates
the possibly large negative impacts of fake prices on consumer welfare resulting from deception.
Data are provided by a major fashion goods manufacturer and retailer in the United States. The
firm sells above 90% of its products by revenue through its own physical stores. The firm derives
the majority of its revenue from a single product category. This paper focuses on sales patterns
within this product category. The firm is the market leader in this category with a market share of
about 40%.
Channels. The firm operates two types of stores: regular and outlet. Because fake prices are
set only for products sold in outlet stores, only transaction data from outlet stores are used for
estimation; however, summary statistics for both channels are provided in this section in order to
provide context for the empirical setting. Regular stores are centrally located in cities and do not
typically offer discounts on products. Outlet stores are located about an hour’s drive away from
city centers and offer deep discounts (both real and fake). The firm offers two types of goods:
original and factory. Original goods are first sold at full price in regular stores and then sold at
a discount in outlet stores. Factory goods are only sold in outlet stores. Table 1 shows counts of
7
unique products sold by channel and product type. The firm implements a fictitious pricing policy
for its factory goods by indicating list prices that are never actual selling prices.
The data consist of transaction-level records over a five-year period. Each record contains the list
price of each item and any active discount. Also included are consumer observables, including
billing ZIP code, date of first purchase from the firm, and household ID.1
Consumers. A total of 16,019,140 unique consumers are observed to make purchases within the
sample. Repeat purchases by consumers are observable in the data. The proportion of purchases
that are made by return consumers is significant (see Figure 1). In the firm’s outlet channel, 24% of
purchases are made by return consumers. Of this group, 38% have made purchases in the regular
channel.
Products, prices, and channels. The firm produces different styles of its main product and offers
each style in different colors. A product is defined by its style and color. Each style-color combi-
nation has only one list price, but styles may vary in list price according to color. The list price is
set at the time of product introduction and never changes. The firm’s managers explain that there
are several reasons why list prices may vary between colors of the same style. One reason is that
some colors may be thought of as more desirable than others. Another reason is that colors may
be introduced at different times, with timing being a factor in determining a product’s list price.
Yet another reason is that some style-color combinations are regular or outlet channel-specific.
Variation in list prices within product styles but between colors provides our primary source of
identification of their effects. Overall, there are 4,610 styles observed in the data, and an average
of 2.93 colors in each style for a total of 13,522 unique products sold.
1
The firm is able to construct longitudinal data through the help of a third-party provider, which uses credit card
information, mailing list, and loyalty program membership, among other data sources, to link different purchases made
by the same consumer over time.
8
Each style-color combination is either an original good or a factory good based on the channel in
which the product is introduced. Table 2 describes the pricing differences between original and
factory goods in outlet stores.2 Original goods are more expensive than factory goods on average,
both in terms of list prices and selling prices. However, there is much variation in selling prices
over time, and factory goods occasionally carry higher prices than original goods (see Figure A2
in the Appendix).
Figure 2 graphs the average percent discount over time for original goods and factory goods in
the firm’s outlet channel. (Average overall discounts over time are presented in Figure A1 in the
appendix.) Recall that original goods are sold at full price in the regular channel, while factory
goods are sold exclusively in the outlet channel. The similar trend in discount increase over time
for these two product classes reflects the firm’s policy of trimming prices at even rates across all
products over time, regardless of sales performance.
The pattern of discounting implemented by the firm in its outlet stores has store, period, and
product-specific components. Across stores and time periods, the firm implements randomized
discounting within certain parameters. Occasionally these discounts are also affected by outlet
mall-wide events. The product-specific component of discounting is highly correlated with the
product’s design age, i.e. the time since its introduction. Older-lived products are discounted more
heavily. Figure 3 plots the list price and observed selling prices for a representative factory good
over time, displaying the resulting discount pattern for a typical product.
The red line on Figure 3 marks the product’s list price. Every single purchase instance over the
product’s lifetime is plotted on the graph; not a single unit was sold at or even near the list price.
2
Further details are in appendix Table A1.
9
Not observable from the graph is that 94% of units were sold by the end of 2008. This is consistent
with high consumer values for a product’s newness in this industry. The remaining purchase obser-
vations are consistent with statements by the firm’s executives that disavow the use of discounting
in order to clear inventory.
4 Demand
In this section we present a parsimonious discrete choice model of consumer purchase behavior
and estimate its parameters using the data set described above. The objective of estimation is to
determine whether fake discounts affect purchase behavior, keeping all other product attributes,
including actual selling prices and real discounts, constant. We are particularly interested in how
the impact of fake discounts compares to that of real discounts. In addition, we seek to measure
how this effect varies according to proxies for consumer familiarity with the brand. We find that
fake prices and their implied fake discounts have a large positive effect on purchase likelihood, and
we find evidence that this effect is smaller for better-informed consumers.
4.1 Model
Let product j in store m and month t be defined by observable characteristics Xjt , unobservable
quality ξj , list price LPj , and selling price pjmt . The list price may be “true” in the sense that
LPj = maxm,t pjmt or “fake” if LPj > maxm,t pjmt . In general, the difference between the list
price and the selling price is composed of a real and a fake discount, i.e.
LPj − pjmt ≡ (LPj − max pjmt ) + (max pjmt − pjmt ) ≡ fake discj + real discjmt (1)
m,t m,t
The indirect utility of consumer i from purchasing product j in store m at time t is denoted as
10
fd rd
uijmt = αmt pjmt + Xjt β + γmt fake discj + γmt real discjmt + ξj + ijmt (2)
fd rd
where αmt , β, γmt , and γmt are taste parameters, and ijmt are idiosyncratic demand shocks. We
incorporate consumer heterogeneity by allowing parameters to vary by store and period in order
to streamline estimation and exposition. This form of utility is similar to previously considered
specifications in which other price components or reference prices are included as predictors of
demand (e.g. Greenleaf 1995).
Letting ijmt be i.i.d. Type-I extreme value, and inverting the resulting system of market share
equations (Berry 1994), mean utilities δjmt can be written as
fd rd
log(sjmt ) − log(s0mt ) = αmt pjmt + Xjt β + γmt fake discj LPj + γmt real discjmt + ξj (3)
where sjmt are market shares (the number of units of product j sold divided by the market size for
store m and time t) and s0mt is the share of the outside good. A consumer is considered to have
chosen the outside good if she visited a store but did not make a purchase. In the estimation of this
model, availability of per-period foot traffic counts for each store enables direct measurement of
these outside shares.
4.2 Identification
In this subsection, we describe how variation in list prices, discounts, and the assignment of real
and fake discounts allows us to estimate demand parameters. In general, using transaction data can
raise possible issues of selection relating to both products and consumers. On the product side,
if certain kinds of products are systematically more likely to be assigned fake prices, or if they
have higher fake prices, then this may influence the estimated effects of fake prices on purchase
outcomes. The estimates may also be skewed if specific types of consumers are more attracted to
11
the outlet channel and the fake pricing policy accounts for this market composition.
The ideal setting for measuring the effect of list prices on purchase likelihood is one in which list
prices are randomly assigned to products, or in which list prices vary exogenously within products
over time or location. The current setting falls short of this ideal, but arguably comes close. The key
identifying assumption is that list prices are uncorrelated with unobservable product characteristics
after controlling for style and color. We argue that this is a weak assumption to make given the
industry. For this assumption not to hold true, two conditions need to be met: (i) there are particular
colors and styles that are particularly good “matches” such that style and color fixed effects do not
adequately control for product desirability; and (ii) the seller is able to identify these matches and
sets list prices accordingly with consistent accuracy. While the first condition may be true, the
second condition is highly unlikely to be the case in an industry characterized by high demand
uncertainty.
Style fixed effects address much of the selection issue where products are concerned, as their
inclusion results in fake price effects being identified from within-style variation in list prices.
This within-style variation arises from between-color variation in fake prices. Table 3 illustrates
the source of variation in list prices by style and color. The frequency of specific list prices is
listed in Table A1 in the appendix. Color fixed effects are included to account for the relative
attractiveness of each color.
Factory products, which all feature fake prices, are only sold in the outlet channel—which is the
only channel for which data is used in the analysis. Hence, a factory product indicator adequately
controls for selection in which products are assigned fake prices; the coefficient on this indicator
reflects demand variation for styles that are found in both channels. In a general retail environment
there may be selection in the relative magnitudes of fake and real discounts assigned to products.
For instance, products that have been found to sell poorly may be assigned higher real discounts,
thus biasing the effect of real discounts toward zero. As discussed in Section 3, however, the
12
firm implements a national pricing scheme that does not allow for discounting based on the sales
performance of individual products.
It is not possible to exactly verify the absence of a systematic relationship between products’ sales
performance and their selling prices, as in practice prices can be influenced by a combination of a
large number of non-product-specific factors, e.g. outlet mall-wide sales and national promotions.
In order to gain some measure of the likelihood of such a systematic relationship, we look to
the most important distinction between products in the firm’s portfolio: that between factory and
original goods.
We run a series of hedonic regressions in order to examine systematic differences in list and selling
prices between factory and original goods. The results from these regressions are reported in Table
4. Each observation in these regressions is a unique product (style-color pair). The estimates in
column 1 show that on average, a factory product has a list price that is $43.38 less than an original
product. Column 2 includes style fixed effects, and shows that, for styles that the firm sells in both
channels, factory versions do not have systematically lower or higher list prices. Columns 3–5
contain results from regressions of average selling prices in the outlet channel. Factory products
sell for $42.46 less than original products on average. While there is a statistically significant
difference in selling prices between factory and original products of the same style, this difference
is minimal at $7.15. When controlling for the duration in which each product was available for
sale, we find no evidence of a difference in pricing strategies for factory and original goods on
average. These results suggest that including style fixed effects and product age in our estimation
of demand coefficients can adequately control for selection in pricing strategies between factory
and original goods.
13
4.3 Estimation and Results
Table 6 contains the baseline results of demand estimation. Estimates from an OLS regression of
mean utility are reported in the first column. Dummies for product color and style, as well as for
stores and months, are included as explanatory variables. As anticipated, purchase probability is
positively correlated with both real and fake components of discounts, negatively correlated with
selling price, and negatively correlated with product age.
The second column in Table 6 includes an interaction between real discount and a factory dummy.
We include this interaction to control for possible discrepancies in how real discounts for original
products (where the real discount is always the whole discount) and real discounts for factory
products (for which the discount always includes real and fake parts) impact purchase likelihood.
We find that real discounts have a stronger effect for factory products than original products, while
the effect of fake discounts dominates that of real discounts.
The third column in Table 6 contains results from an alternative specification in which the list price
is included as a regressor instead of discount variables. Comparing coefficients, we see that a $1
increase in a product’s list price has the same effect on purchase probabilities as a $0.77 decrease
in selling price, all else held constant. Considering that a firm can increase a firm’s list price at
virtually no cost, this has potentially huge consequences for producer and consumer welfare.
14
The fourth column presents estimates from a model with an interaction term between list price and
the factory dummy. Recall that all factory products have fake prices while no original products
have fake prices; hence the factory dummy is equivalent to a fake price dummy. The estimates
show that fake list prices have an effect size on purchase probabilities that is almost twice that
of real list prices. This comparison needs to be interpreted with caution, however, since there
may be unobservable differences between original and factory products that influence the relative
importance of list prices.
These results show that fake list prices and discounts have effects on purchase behavior that rival
those of real ones in magnitude. These findings raise interesting questions on the reasons behind
this demand response, as well as its implications on firm decisions. In the succeeding analyses, we
explore how consumer familiarity with the brand moderates these effects.
Consumers who lack full information on product desirability may take price as a signal of quality
(Gerstner 1985). The availability of demand data that include both list and selling prices presents
an opportunity to cleanly measure this signaling effect separately from price sensitivity. If less-
informed consumers are more reliant on price as a signal of quality, then they should demonstrate
more sensitivity to list prices than better-informed consumers (Armstrong & Chen 2013).
The current scenario differs from the typical environment considered in theoretical models that
characterize price as a signal of quality. First, here we have a single firm selling multiple goods,
whereas the usual model has two or more firms of different “quality” with one representative
product each. Second, a standard assumption is for quality to be increasing in marginal cost, with
the higher-quality firm having a lower cost of quality, and hence a more credible commitment
to keep prices high. In the current example, higher-quality products are not necessarily more
expensive to produce, as desirability is greatly influenced by aesthetic components that do not
affect production costs.
15
While the current scenario has not specifically been explored in prior research, there are similar
and additional reasons why one might expect price to signal quality here. Even as a monopolist,
a firm may have an incentive to guide new, uninformed consumers toward better options within
its own product assortment so as to increase customer lifetime value. Fake list prices may play a
special role given the cost structure: because quality is not strongly related to marginal cost, the
ability to post fake list prices may relieve the tension of pricing against marginal cost to maximize
profits (downward pressure) while depending on price to signal quality (upward pressure).
To explore how sensitivity to fake discounts relates to consumer familiarity with the brand, we
estimate interactions between discount variables and proxies for consumers’ experience with the
brand. The first proxy is the length of time, in months, that the store has been operating. This
presumes that the average consumer in a store is better-informed about the store’s pricing policies,
and the existence of fake prices, the longer the store has been open. The second proxy is the
distance between a focal store and the brand’s nearest regular retail store. (Recall that each store
in the estimation data set is an outlet store.) The operating assumption for this variable is that
the closer a regular retail store, the greater the awareness of the average consumer in identifying
differences in product assortment between the two channels and, consequently, the presence of
fake prices in the outlet channel. Given that store fixed effects are included in all regressions, these
interactions are identified from store openings and closures in the data.
Table 7 contains the estimation results including store-level interactions. The first column includes
results from the model accounting for store age. Precise store ages are available only for stores
that opened during the five-year sample duration; the variable In sample is an indicator for these
stores. The estimates indicate that consumers in stores that have been operating for a longer period
are less sensitive to fake discounts. There is also some evidence that consumers in older stores
are more sensitive to real discounts, although this difference is statistically insignificant between
stores that opened within the sample duration.
16
The second column of results in Table 7 accounts for the distance of each store from the nearest
regular retail store. This distance changes over time due to the opening or closing of regular stores
by the brand. The average distance between an outlet store and the nearest regular store, taken over
all observations in the data, is 39 miles. There are 53 cases in which an outlet store’s closest regular
store changes as a result of a store opening, and seven cases where the change is the result of a store
closure. For these cases, the average change in the distance variable is -23 miles when a regular
store opens and 12 miles when a regular store closes. The estimates imply that the farther the
nearest regular store is, the more sensitive consumers are to both real discounts and fake discounts.
Both proxies are included in the model for which estimates are presented in the third column.
These estimates provide suggestive evidence that the more experience consumers have with the
brand, either through a store’s age or its proximity from a second channel, the less sensitive they
are to fake discounts. We motivate the laboratory experiment in the following section by observing
that any proxy of consumer information would be an imperfect indicator of consumer beliefs about
the firm’s pricing practices. Indeed, our proxies are possibly correlated with other factors that may
also influence sensitivity to fake prices. We complement our findings from observational data with
those from experimental data in which subjects are provided with specific information about true
and displayed original prices.
5 Laboratory experiment
This section contains results from a laboratory experiment designed to measure the effect of fake
prices on purchase intent. This study serves as a useful complement to the preceding analysis of
observational data as it allows for direct manipulation and verification of subjects’ understanding
of the fakeness or realness of list prices. The findings strongly reject the hypothesis that fake prices
influence purchase intent when subjects understand that they are false.
As discussed in our review of the literature, several authors have previously established a positive
17
relationship between reference prices and purchase intention. In multiple studies, the plausibility of
reference prices—manipulated by comparing the effects for unfamiliar versus familiar brands—has
been found to moderate this relationship (e.g. Biswas & Blair 1991). However, to our knowledge,
there is no prior work published that tests for this effect when consumers have full information
about the truthfulness of posted list prices. Existing research on uninformative anchors provides
reason to expect a positive relationship on purchase intent even under these conditions (e.g. Wilson,
et al. 1996).
Participants (N = 1,085: 53% female, Mage = 35 years, participants from MTurk) were presented
with the image of a bag together with price information. The bag was chosen to roughly correspond
with the relevant product category in the preceding analysis of observational data. All participants
were told that the bag had a sale price of $40. Each participant was also presented with a potentially
fake displayed original price, as well as information about the true original price. Figure 4 is one
instance of the stimulus.
Participants were randomly assigned to conditions with varying levels of true and displayed origi-
nal prices, ranging from $40 to $90 in $10 intervals (where the true original price is at most as high
as the displayed original price). Table 8 shows the number of subjects that were assigned to each
treatment group. To ensure that participants understood the scenario correctly, they were asked to
type in what they understood to be the product’s true original price. We drop observations in which
participants enter an incorrect price. This shrinks the number of observations from 1,085 to 877.3
Each participant was asked to rate their purchase likelihood on a seven-point scale.
To measure the impact of each piece of price information on purchase likelihood, we run the
following model on the data:
3
Results from analysis using all 1,085 observations are nearly identical to those using the smaller set of 877
observations.
18
P urchase intenti = β0 + β1 Displayed pricei + β2 T rue original pricei + (4)
Table 9 contains the results of this regression. Strikingly, although the true original price has a
significantly positive effect on purchase intent, the displayed original price does not have a statis-
tically significant effect. These results are particularly remarkable when considering that, in the
survey design, the true original price was provided only in the initial prompt, whereas only the
displayed original price was presented when asking for purchase intent (and subsequent questions
about sentiment).
These results lend support to the contention that fake prices increase purchase likelihoods by de-
ception. To further investigate how consumer inferences derive from price signals, subjects were
asked to rate the offer on specific dimensions after they are asked about purchase intent. These
dimensions, all rated on a seven-point scale, include: (i) the subject’s own rating of product qual-
ity, (ii) the subject’s prediction of other people’s rating of product quality, (iii) whether the subject
thinks the offer is a good deal, (iv) whether the subject thinks the product is being offered at a fair
price, and lastly (v) whether the subject feels that the seller is being dishonest.
Table 10 summarizes the results of this survey by regressing each rating on the pricing components.
We find that high true original prices lead subjects to rate the product as higher quality, as well as to
infer other subjects’ higher quality ratings. We also find that they perceive the offer to be a better
deal, and the price to be fairer the higher the true original price. Ratings of deal attractiveness
are roughly three times more sensitive to the true original price than those of product quality.
Strikingly, the displayed original price has no influence on these ratings. We do find, however, that
consumers consider sellers more dishonest the farther the displayed original price is from the true
original price.
19
[Table 11 about here.]
We complete a mediation analysis by regressing purchase intent on price signals and sentiment rat-
ings jointly. Results are presented in Table 11. We find that sentiment ratings completely mediate
the effect of price signals on purchase intent, with the subject’s own quality rating being slightly
more important than their ratings for deal attractiveness. Subject’s perceptions of others’ quality
ratings do not seem to impact their own purchase intent. Interestingly, ratings for whether the seller
is dishonest also do not correlate with purchase intent. Analysis of these different dimensions of
consumer sentiment are remarkably consistent in that they verify previously theorized or estab-
lished relationships between reference prices and consumer beliefs (e.g. Dodds, Kent & Grewal
1991).
These results imply that positive demand effects from fake prices detected from observational data
arise largely from the false understanding by consumers that fake prices represent past selling
prices. The higher the product’s previous selling price, the higher subjects rate product quality
and the attractiveness of the offer, and the higher subjects rate their likelihood of purchase. Given
knowledge of the true original price, however, a higher reference price in the form of a displayed
original price has no effect on demand or consumer sentiment apart from the inference that the
seller is dishonest.
6 Conclusion
Price comparisons of practically every shape and form have been heavily used by retailers in their
communications from the very earliest examples of retail advertising up to the current shift to
e-commerce. The question of how these signals affect purchase behavior is relevant to firms,
regulators, and consumers themselves. Firms, in posting list prices different from selling prices,
possess a potentially powerful driver of demand that is virtually costless to produce and adjust.
Regulators face the challenge of assessing whether list prices inform or deceive, and ultimately
20
whether they enhance or damage consumer welfare. Consumers may be surprised to find out how
list prices are determined, and by the extent to which their own decisions are reliant on them.
The results show that list prices have significant effects on purchase decisions. On average, con-
sumers may be thought of as assigning a monetary value to list prices at over 70 “selling price
cents” to a “list price dollar.” This rate is moderated by the length of time at which stores have
been in operation, as well as their distance from alternative channels under the same brand. This
implies that better-informed customers may place less stock on list prices than newer customers
do.
Findings from a laboratory experiment confirm that knowledge of a product’s true original price
attenuates the effect on purchase intent of a displayed original price. Strikingly, when subjects are
informed of a product’s true original price, their purchase intent is found to be completely invariant
to displayed original prices. We find that subjects rely on products’ prior selling prices in making
inferences of product quality and that they regard sellers that display original prices higher than
true original prices as dishonest.
The empirical settings in this paper introduce some limitations. The single-firm data source used in
this paper precludes studying the competitive aspects of fake list pricing, such as possible impacts
on consumers’ likelihood of further search (e.g. Grewal & Compeau 1992). Although, to our
knowledge, this paper is the first to exploit real transactions data to investigate fake list prices,
identification of the effect of fake pricing on purchase likelihood also falls short of that which may
be obtained from field experimentation.
Given the effect of list prices on purchase behavior, a potentially worthwhile area for future re-
search lies in empirically modeling the seller’s problem when setting fake list prices. The persis-
tence of fake pricing despite frequent high-profile lawsuits against the practice suggests that it is
indeed profitable for sellers to pursue. On the other hand, the practice does seem to be disciplined
by natural constraints that result in familiar patterns within industries (e.g. Anderson & Simester
1998). It would be of particular interest for regulators to identify these constraints in their efforts
21
to curtail this practice and protect consumer interests.
Differences between industries may also merit future investigation. The existing scholarly litera-
ture on price as a signal of quality and regulatory guidelines for advertised prices frequently cast
fictitious list prices as a means of deceiving uninformed consumers. This literature relies on the
strictly monotonic relationship of quality and marginal cost as providing credibility to actual sell-
ing price as a signal of quality. Some industries may be better represented by production functions
in which quality is generated through a stochastic process only weakly correlated with marginal
cost, with fake list prices used to reduce asymmetric information about quality between the firm
and consumers. Such a viewpoint may better suit settings in which quality can only imperfectly be
set by firms, such as in fashion and design-related industries.
References
[1] Anderson, Eric T., and Duncan I. Simester. “The Role of sale signs.” Marketing Science 17.2
(1998): 139–155.
[2] Anderson, Eric T., and Duncan I. Simester. “Effects of $9 price endings on retail sales: Evi-
dence from field experiments.” Quantitative Marketing and Economics 1.1 (2003): 93–110.
[3] Armstrong, Mark, and Yongmin Chen. “Discount pricing.” CEPR Discussion Paper No.
DP9327. (2013).
[4] B. Sanfield, Inc. v. Finlay Fine Jewelry Corp., 999 F. Supp. 1102, 1103-04 (N.D. Ill. 1999)
[5] Berry, Steven T. “Estimating discrete-choice models of product differentiation.” RAND Jour-
nal of Economics (1994): 242–262.
[6] Bagwell, Kyle, and Michael H. Riordan. “High and declining prices signal product quality.”
American Economic Review (1991): 224–239.
22
[7] Biswas, Abhijit, and Edward A. Blair. “Contextual effects of reference prices in retail adver-
tisements.” Journal of Marketing (1991): 1–12.
[8] Blair, Edward A., and E. Laird Landon Jr. “The effects of reference prices in retail advertise-
ments.” Journal of Marketing (1981): 61–69.
[9] Blattberg, Robert C., Richard Briesch, and Edward J. Fox. “How promotions work.” Marketing
Science 14.3 (1995): 122–132.
[10] Bordalo, Pedro, Nicola Gennaioli, and Andrei Shleifer. “Salience and consumer choice.”
Journal of Political Economy (2013): 803–843.
[11] Camasta v. Jos. A Bank Clothiers Inc., No. 12-C-7782 (N.D. Ill. 2013).
[12] Compeau, Larry D., and Dhruv Grewal. “Comparative price advertising: an integrative re-
view.” Journal of Public Policy & Marketing (1998): 257–273.
[13] Compeau, Larry D., Dhruv Grewal, and Rajesh Chandrashekaran. “Comparative price adver-
tising: Believe it or not.” Journal of Consumer Affairs 36.2 (2002): 284–294.
[14] Della Bitta, Albert J., Kent B. Monroe, and John M. McGinnis. “Consumer perceptions of
comparative price advertisements.” Journal of Marketing Research (1981): 416–427.
[15] Dodds, William B., Kent B. Monroe, and Dhruv Grewal. “Effects of price, brand, and store
information on buyers’ product evaluations.” Journal of Marketing Research (1991): 307–319.
[16] Dolan, Maura. “Kohl’s can be sued over sale ads, court says.” Los Angeles Times. May 22,
2013.
[17] Friedman, David Adam. “Reconsidering fictitious pricing.” Minnesota Law Review 100
(2015): 921–981.
[18] Gerstner, Eitan. “Do higher prices signal higher quality?” Journal of Marketing Research
(1985): 209–215.
23
[19] Greenleaf, Eric A. “The impact of reference price effects on the profitability of price promo-
tions.” Marketing Science 14.1 (1995): 82–104.
[20] Grewal, Dhruv, and Larry D. Compeau. “Comparative price advertising: Informative or de-
ceptive?” Journal of Public Policy & Marketing (1992): 52–62.
[21] Gupta, Sunil, and Lee G. Cooper. “The discounting of discounts and promotion thresholds.”
Journal of Consumer Research 19.3 (1992): 401–411.
[22] Hastings, Justine, and Jesse M. Shapiro. “Mental accounting and consumer choice: Evidence
from commodity price shocks.” NBER Working Paper No. 18248. (2012).
[23] Hendel, Igal, and Aviv Nevo. “Intertemporal price discrimination in storable goods markets.”
NBER Working Paper No. 16988. (2011).
[24] Kim, Dong Soo, Mingyu Joo, and Greg M. Allenby. “An Economic Model Explaining Ref-
erence Price Effects.” Mimeo. (2017).
[25] Kim v. Carter’s, Inc., 598 F.3d 362 (7th Cir. 2010)
[26] Kirmani, Amna, and Akshay R. Rao. “No pain, no gain: A critical review of the literature on
signaling unobservable product quality.” Journal of Marketing 64.2 (2000): 66–79.
[27] Kumar, V., et al. “The impact of internal and external reference prices on brand choice: the
moderating role of contextual variables.” Journal of Retailing 74.3 (1998): 401–426.
[28] Lattin, James M., and Randolph E. Bucklin. “Reference effects of price and promotion on
brand choice behavior.” Journal of Marketing Research (1989): 299–310.
[29] Linnemer, Laurent. “Price and advertising as signals of quality when some consumers are
informed.” International Journal of Industrial Organization 20.7 (2002): 931–947.
[30] Marshall, Roger, and Seow Bee Leng. “Price threshold and discount saturation point in Sin-
gapore.” Journal of Product & Brand Management 11.3 (2002): 147–159.
24
[31] Mayhew, Glenn E., and Russell S. Winer. “An empirical analysis of internal and external
reference prices using scanner data.” Journal of Consumer Research 19.1 (1992): 62–70.
[32] Milgrom, Paul, and John Roberts. “Price and advertising signals of product quality.” Journal
of Political Economy 94.4 (1986): 796–821.
[33] Monroe, Kent B., and Joseph D. Chapman. “Framing effects on buyers’ subjective product
evaluations.” Advances in Consumer Research 14 (1987): 193–197.
[34] Mulligan v. QVC, Inc., 888 N.E.2d 1190 (Ill. App. Ct. 2008).
[36] Pitofsky, Robert, Randal Shaheen, and A Mudge. “Pricing laws are no bargain for con-
sumers.” Antitrust 18 (2003): 62–65.
[37] Popken, Ben. “JCPenney, Sears, Macy’s and Kohl’s sued for ‘fake’ sale pricing.” NBC
News (2016), https://www.nbcnews.com/business/consumer/jcpenney-sears-macy-s-kohl-s-
sued-fake-sale-pricing-n694101. Accessed 29 September 2017.
[38] Rajendran, Kharan N., and Gerard J. Tellis. “Contextual and temporal components of refer-
ence price.” The Journal of Marketing (1994): 22–34.
[39] Rhodes, Andrew, and Chris M. Wilson. “False advertising.” RAND Journal of Economics
49.2 (2018): 348–369.
[40] Thaler, Richard. “Mental accounting and consumer choice.” Marketing Science 4.3 (1985):
199–214.
[41] Tversky, Amos, and Daniel Kahneman. “The framing of decisions and the psychology
of choice.” Environmental Impact Assessment, Technology Assessment, and Risk Analysis.
Springer, Berlin, Heidelberg, 1985. 107–129.
25
[42] Urbany, Joel E., William O. Bearden, and Dan C. Weilbaker. “The effect of plausible and
exaggerated reference prices on consumer perceptions and price search.” Journal of Consumer
Research 15.1 (1988): 95–110.
[43] Wilson, Timothy D., Christopher E. Houston, Kathryn M. Etling, and Nancy Brekke. “A
new look at anchoring effects: basic anchoring and its antecedents.” Journal of Experimental
Psychology: General 125.4 (1996): 387–402.
[44] Zeithaml, Valarie A. “Consumer perceptions of price, quality, and value: a means-end model
and synthesis of evidence.” Journal of Marketing (1988): 2–22.
26
Figure 1: Consumers by number of within-sample purchase instances
27
Figure 2: Discounting pattern in outlet channel
28
Figure 3: Observed selling prices of a typical factory good in the outlet channel
29
Figure 4: Lab study stimulus
Note: In this instance, the “true original price” is $60 and the “displayed original price” is $80.
30
Table 1: Unique product counts by channel and product type
31
Table 2: Average prices in outlet format
32
Table 3: Illustration of variation in list prices
33
Table 4: Hedonic regressions
34
Table 5: Descriptive statistics for estimation data
35
Table 6: Baseline demand estimates
36
Table 7: Heterogeneity
37
Table 8: Subject counts by treatment group
38
Table 9: Main results from lab study
Dependent variable:
Variables Purchase intent
Observations 877
R-squared 0.032
Standard errors in brackets
*** p<0.01, ** p<0.05, * p<0.1
39
Table 10: Sentiment ratings
40
Table 11: Mediation
Dependent variable:
Variables Purchase intent
Observations 877
R-squared 0.477
Standard errors in brackets
*** p<0.01, ** p<0.05, * p<0.1
41
Appendix
42
Figure A1: Average discount percent in outlet stores
43
Figure A2: Prices over time
44
Table A1: Frequency of list price values
List price Percent of products Average transacted price Discount Percent original styles
298 15.84 114.63 60.68 40.80
398 10.87 149.30 61.21 45.65
358 9.14 141.59 59.93 37.93
348 6.54 123.45 63.82 61.45
328 6.38 104.72 67.30 74.07
198 5.67 96.00 50.52 9.72
278 4.33 113.80 58.13 12.73
498 4.02 190.59 61.08 17.65
268 3.62 102.72 59.97 32.61
598 3.23 249.80 57.55 9.76
248 2.60 91.30 62.32 57.58
258 2.52 91.38 61.35 6.25
458 2.29 150.79 66.45 27.59
428 2.29 149.41 64.37 72.41
378 2.21 129.50 64.31 60.71
45