Lifting the Veil: The Benefits of Cost
Transparency
Bhavya Mohan
Ryan W. Buell
Leslie K. John
Working Paper 15-017
Lifting the Veil: The Benefits of Cost
Transparency
Bhavya Mohan
University of San Francisco
Ryan W. Buell
Harvard Business School
Leslie K. John
Harvard Business School
Working Paper 15-017
Copyright © 2014, 2015, 2016 by Bhavya Mohan, Ryan W. Buell, and Leslie K. John
Working papers are in draft form. This working paper is distributed for purposes of comment and discussion only. It may
not be reproduced without permission of the copyright holder. Copies of working papers are available from the author.
Lifting the Veil: The Benefits of Cost Transparency
BHAVYA MOHAN
RYAN W. BUELL
LESLIE K. JOHN
Firms typically treat their costs as tightly-guarded secrets. In six studies, we test the effect
of firm disclosure of the costs to produce a given product (i.e., cost transparency) on purchase
interest. We begin with a natural field experiment conducted with an online retailer, in which
cost transparency increased sales (Study 1A). We subsequently replicate this field experiment in
a controlled lab setting (Study 1B), and show that cost transparency is particularly potent in
boosting purchase interest above other forms of transparency (Study 2). Guided by our
theoretical framework, Studies 3 and 4 show that the effect is mediated by consumers’ trust in
the firm, with Study 4 showing that this mediator explains variance above and beyond
perceptions of price fairness. Finally, Study 5 demonstrates the critical role of the voluntary
nature of the disclosure, by showing that cost transparency boosts purchase interest only when
voluntarily instated by the firm, as opposed to involuntarily (e.g., as required by law). These
results imply that the proactive revelation of costs can improve a firm’s bottom line.
2
INTRODUCTION
Cost transparency refers to the voluntary disclosure of the costs to produce a good or
provide a service. It has been studied in operations and marketing within the context of supplierfirm relationships, whereby the two-way sharing of cost information between these parties
facilitates collaboration on cost reduction measures (Lamming et al. 2002; Zhu 2004). In this
paper, we investigate cost transparency within a different context: customer-firm relationships.
Although information on the costs associated with making a good is not typically proactively
shared with customers, we provide evidence of when and why cost transparency can increase
consumers’ purchase interest.
Cost, Operational, and Price Transparency
Cost transparency refers to a firm’s disclosure of the costs that the firm incurs to provide
a given product or service. By contrast, operational transparency refers to a firm’s disclosure of
its operating processes (Buell, Tsay, & Kim, 2014); specifically, of the “behind-the-scenes” work
that the firm is undertaking through its operating processes (Buell & Norton, 2011). Research has
shown that consumers prefer service web sites that are operationally transparent relative to those
that are not (Buell & Norton, 2011). For example, the travel site Kayak.com is beloved in part
because of its operational transparency: the site discloses in real-time which airlines it is
searching. Operational transparency increases consumers’ perception of the effort required to
create the product (or, in the case of Kayak, to generate the quote), in turn heightening their
sense of gratitude and willingness to pay (Buell and Norton 2011; Chinander and Schweitzer
2003; Gershoff, Kivetz, and Keinan 2012; Morales 2005).
3
Cost transparency is also distinct from price transparency. Whereas the former entails
disclosing the firm-side costs inherent in a price, price transparency entails disclosing and
delineating the firm-side proceeds inherent in a price; for example, by dividing a price into gross
retail proceeds, royalties, and taxes (Carter and Curry 2010). Similarly, price partitioning refers
to revealing the price of the component parts of a product; for example, by dividing a product's
price into its base price and shipping and handling (Bertini and Wathieu 2008; Morwitz,
Greenleaf, and Johnson 1998). Price transparency and price partitioning have both been found to
increase purchase intentions, and to do so via a cognitive process (Morwitz et al., 1998).
Specifically, by dividing a price into several sub-components, each of which is necessarily
smaller than the total price, small prices are made salient. The result is that these tactics cause
consumers to perceive prices to be relatively low, in turn increasing purchase intentions.
In addition to being conceptually distinct from operational and price transparency as
noted above, we posit that cost transparency’s effect on purchase intentions operates via a
distinct psychological process. We elucidate the theoretical underpinning for this proposition in
the next sections.
Mere Cost Salience versus Cost Transparency
Cost transparency makes costs salient. It is therefore related to previous work on how
cueing consumers to think about costs affects purchase interest (Bolton and Alba 2006; Bolton,
Warlop, and Alba 2003). This work is rooted in the principle of dual entitlement (Kahneman,
Knetsch, and Thaler 1986), which posits that consumers believe that although firms are entitled
to make a profit, they also believe that they themselves are entitled to a reasonable price. Hence,
when they perceive a firm to be making unreasonably large profits, they tend to view its prices as
4
unfair, and in turn are less willing to buy from the “offending” firm (Bazerman 1985; Kőszegi
and Rabin 2015).
However, consumer perceptions of the magnitude of a firm’s profits do not always match
reality. Since consumers do not routinely think about the (often considerable) costs a firm incurs
to offer a given product, they are prone to overestimating the firm’s profits, and hence to
erroneously conclude that it is making an unreasonably large profit (Bolton, Warlop, and Alba
2003). Consistent with this theorizing, Bolton et al. (2003) have shown that prompting
consumers to think about firms’ costs increases their perceptions of price fairness relative to
when they are given no such prompt. Although cost transparency draws consumers’ attention to
costs, it also entails a firm’s voluntary revelation of such costs. As we next describe, this act of
firm disclosure makes novel predictions about the effect of cost transparency on consumer
purchase interest.
Cost Transparency, Firm Disclosure, and Trust
A substantial body of work in social psychology and allied fields suggests that disclosure,
especially of information typically kept private, is associated with heightened relationship quality
(Laurenceau, Barrett, and Pietromonaco 1998). Clever experimental studies have shown that this
relationship can be causal: inducing people to self-disclose causes others to like them (Aron et al.
1997; Sedikides et al. 1999). A necessary condition for this causal relationship appears to be that
the disclosed information be sensitive in nature. Unlike merely disclosing information on
production processes (operational transparency) or say, taxes (price transparency), since costs are
generally viewed as confidential, we posit that costs represent the type of sensitive information
that previous research has found to lend itself to fostering liking in person-to-person disclosure.
5
What mechanism drives the capacity for self-disclosure to increase liking? Previous
theorizing has invoked trust: self-disclosure has been argued to foster trust, which in turn is
thought to be an ingredient that produces increased liking after self-disclosure. Consistent with
this account, recent research has documented causal evidence of the inverse: abstaining from
disclosure (for example, by opting out of answering survey questions), makes a person seem
untrustworthy, in turn reducing others’ liking of them (John, Barasz, and Norton 2016). This
research also pinpoints that it is the volitional – as opposed to inadvertent – withholding of
information that makes a person come across as shady. Previous research therefore leads us to
predict that cost transparency will increase purchase interest. Specifically, we predict that when a
firm voluntarily discloses its costs, consumers will view that firm as relatively trustworthy, and
in turn will be more interested in buying its products.
Importantly, cost transparency and its supporting account make predictions not covered
by the principle of dual entitlement. Although both perspectives predict that cost transparency
can increase purchase interest, they invoke different mechanisms for the process underlying such
an effect. Dual entitlement predicts that the capacity for cost salience (which can be achieved
through cost transparency) to increase purchase interest is driven by perceptions of price fairness.
Thus if cost transparency’s effect on purchasing arises simply because it makes costs salient,
then the effect should be driven solely by price fairness. If, as we argue, there is something
special about a firm’s voluntary disclosure of such information, then additional variance should
be explained via consumer trust in the firm. We test this in Studies 3 and 4.
Secondly, there is a situation, which we exploit in Study 4, in which these accounts make
directionally different predictions of the effect of cost transparency on purchase interest.
Specifically, dual entitlement predicts that when prices are revealed to be lower than consumers’
6
expectations, revealing costs will reinforce that firm profit is relatively low, and hence purchase
interest will increase. Critically however, dual entitlement would also predict that when prices
are higher than consumers’ expectations, revealing costs will reinforce that the firm’s profit is
relatively high, and hence in this situation, cost transparency will decrease purchase interest. In
other words, if there is nothing special in the voluntary firm disclosure of sensitive (cost)
information, and the effects of cost transparency are simply a byproduct of making costs salient,
then cost transparency should only increase purchase intentions when it positively disconfirms
consumers’ beliefs about profit – i.e., when it tells consumers that firm profit is not as high as
they had thought. By contrast, if, as we posit, there is something special about it, then cost
transparency should increase purchase interest even in the case of negative price disconfirmation.
Finally, in line with our theoretical account, we show that merely disclosing costs is not
enough to spur sales if the disclosure is forced (Study 5). Rather, a firm must proactively and
voluntarily disclose its costs, in much the same way as individuals must voluntarily self-disclose
in order to increase trust and liking (Collins and Miller 1994; Wheeless and Grotz 1977). If, on
the other hand, the effects of cost transparency are simply a byproduct of making costs salient,
then whether they are made salient by voluntary versus involuntary firm action should not
matter.
Overview of Studies
In six studies, we test the effect of cost transparency on purchase interest. We begin with
a natural field experiment conducted with an online retailer, in which cost transparency increased
sales (Study 1A). Five subsequent controlled lab experiments replicate and extend this basic
effect (Studies 1B – 5). Specifically, Study 1B replicates the field experiment, distilling the
7
critical ingredients of cost transparency that are causally responsible for boosting purchase
interest. Study 2 provides evidence of the specificity of the effect, showing that cost transparency
in particular, as opposed to other forms of transparency, such as the mere provision of
information on production processes (operational transparency), is particularly potent in boosting
purchase interest. Finally, guided by our theoretical framework, Studies 3-5 test shed light on
when and why the beneficial effect of cost transparency emerges. Specifically, Studies 3 and 4
show that the effect is mediated by perceptions of firm trustworthiness, with Study 4 showing
that this mediator explains variance above and beyond perceptions of price fairness. Study 4 also
shows that cost transparency increases purchase intent even when the costs are revealed to be
lower than, and hence the profit margins higher than, consumers’ expectations – a result not
predicted by dual entitlement. Finally, Study 5 demonstrates the critical role of the voluntary
nature of the disclosure, by showing that cost transparency boosts purchase interest only when
voluntarily instated by the firm, as opposed to involuntarily (e.g., as required by law). Attesting
to the robustness of the effect, we demonstrate it across different instantiations of cost
transparency, and across a variety of different brands and product categories.
STUDY 1A: FIELD EVIDENCE
On December 2, 2013, a privately-held online retailer launched a holiday gift shop with a
single email to its mailing list, promoting a leather wallet offered in five colors (burgundy, black,
grey, bone, and tan) and priced at $115.00. At the end of January, in an effort to boost postholiday sales, the retailer decided to add a cost transparency infographic to the online product
detail pages of each of the wallet’s five color combinations. As the wallets differed only in color,
8
the company intended to use the same infographic for every wallet in the line.
But what the company intended to do was not what actually happened. Serendipitously
(for us at least), the company inadvertently failed to introduce the infographic for two of the
wallet colors (bone and tan). Thus the infographic was implemented for only three of the five
wallet colors (burgundy, black, and grey), a mistake that was overlooked for five weeks, creating
a natural experiment enabling us to test the impact of cost transparency on sales.
Procedure
Operationalization of Cost Transparency. Along with the total cost to produce the wallet,
the infographic broke the total cost into its components, delineating the specific costs associated
with the given materials and processes required to produce and import the wallet: leather
($14.68), construction ($38.56), duties ($4.26) and transportation ($1.00). The infographic also
included additional benchmark information stating that the wallet had a 1.9x markup, compared
to the 6x markup charged by a competitor (this extraneous information is absent in the
subsequent lab replication Study 1B).
Empirical Approach. The inadvertent provision of cost information for some, but not all,
of the wallet colors served as an exogenous shock that created sets of comparable treatment (cost
transparent) and control (nontransparent) products (i.e., wallets). This treatment provides a
conservative test of cost transparency since customers browsing multiple wallet colors may have
been exposed to the infographic and (correctly) inferred that the same cost information applied to
all colors. While the benefits of the infographic likely accrued to both groups, our identification
comes from the fact that every customer who browsed wallets in the treatment group was
exposed to the infographic, while customers who browsed wallets in the control group may not
9
have been exposed.
We used a difference-in-differences approach to compare the daily sales between the
treatment and control groups before versus after the infographic was introduced. By doing so, we
isolated the effect of cost transparency on the daily count of wallets sold in each category. We
analyzed the sales performance of five color combinations over a 92-day period (N = 460)
starting with the launch of the site on December 2, 2013 and ending on March 6, 2014. The
infographic was introduced on January 28, 2014.
We estimated the following linear fixed effect specification, using a Newey West
estimator for standard errors that accounts for autocorrelation and heteroskedasticity within
colors with a small number of products (Newey and West 1987; Schaffer 2010):1
⎛ α 0 + α 1POSTt + α 2POSTt ×TREATMENTc + α 3VIEWSc ,t + α 4VIEWS 2c ,t + ⎞
COUNTc ,t = f ⎜
⎟
⎝ α 5NOVIEWSc ,t + α 6NOVIEWSc ,t−1 + α 7NOSALEc ,t−1 + β c + εc ,t
⎠
!
(1)
In the specification above, COUNTc,t represents the count of wallets sold for color c on
day t. POSTt is a dummy variable denoting observations after the introduction of the infographic.
Although the cost transparency treatment is subsumed by the color fixed effect (i.e., burgundy,
black, and gray color fixed effects perfectly identify the wallets that received the cost
transparency treatment), βc, POSTt × TREATMENTc is a dummy variable that specifically
highlights observations in the cost transparency treatment conditions after the introduction of the
infographic and is the focal independent variable of our analysis.
1
We did not use clustered standard errors since clustering requires a large number of clusters – far more than the
five product colors we have in our dataset – to approach the true variance of the error term. Using a Newey West
estimator enabled us to leverage our fairly long panel of sales data on each wallet color to generate a consistent
estimator that accounts for the autocorrelation and heteroskedasticity within each wallet.
10
Since cost transparency was implemented by wallet color, and there were only five
colors, we also controlled for possible confounders: a proxy for time-varying color popularity:
the number of views each wallet color’s page received; and a proxy for inventory levels: for a
given color, complete absence of both page views and sales. Controlling for these variables was
important because they could be (and in fact, were) independently correlated to both the
treatment and outcome. In the case of popularity, more popular colors (which are viewed more
frequently) could happen to have received the treatment. Popularity would also be logically
independently correlated to sales (popular items sell better). Hence the model includes variables
representing the number of views each wallet color’s page received (VIEWSc,t) as well as the
square of this number (VIEWS2c,t.), to accommodate a non-linear relationship between page
views and online sales (sales rise in page views at a diminishing rate). By controlling directly for
the daily page views of each product (which may fluctuate if, for example, some product colors
are more popular holiday gifts than others), our focal coefficients uniquely reveal the effect of
cost transparency on converting browsers to buyers, holding constant time-varying differences in
latent demand.
Similarly, we also included inventory proxies as covariates to control for the effect of
stock-outs. Stock-outs are negatively correlated with sales, since a company cannot sell items it
does not have in inventory, and stock-outs are also plausibly correlated with the treatment, since
some colors are likely to be more popular than others, and may therefore be more prone to stockouts. Although the company did not keep historical records of its inventory position over time,
two sets of variables proxy for the effects of stock-outs. The first reflects the page view
implication of stock-outs, capturing the complete lack of page views on any given day
(NOVIEWSc,t), as well as on the preceding day (NOVIEWSc,t−1). When the product was out of
11
stock on the company’s website, a “SOLD OUT” message blacked out the product on search
results pages, reducing the likelihood that a customer would view the page. The second proxy
variable reflects the sales implication of stock-outs, capturing the complete absence of sales on
the preceding day (NOSALEc,t−1).
(Insert Figure 1 about here)
Results and Discussion
Figure 1 depicts the average daily unit sales as a function of the treatment over the period
of analysis (units are withheld to protect confidential company data). Overall, sales declined over
the period, reflecting diminished demand in the post-holiday season. More importantly however,
there was an interaction between time and treatment: specifically, the post-holiday sales decline
was smaller in the treatment condition relative to the control condition. Cost transparency
therefore helped sales by serving as a buffer against the post-holiday sales decline.
(Insert Table 1 about here)
Table 1, Column (1) presents the base specification in which the daily number of units
sold per color combination is modeled as a function of the time period and treatment
classification of the product group. Once confounds are controlled for, the treatment effect
emerges. Specifically, the positive treatment effect is demonstrated in Column 2 when
controlling for color popularity (coefficient = 0.579; p < 0.10 two-tailed); in Column 3 when
further controlling for the page view implications of stock-outs (coefficient = 0.660; p < 0.05
two-tailed); and in Column 4 when further controlling for the sales implications of stock-outs
(coefficient = 0.579; p < 0.10 two-tailed).
Under the most conservative assumption of full substitution – that all of the incremental
12
customers who bought wallets in the treatment condition would have otherwise bought a wallet
in the control condition in the absence of the treatment – we calculate that the cost transparency
infographic increased daily unit sales on a per-color basis by 22.0% relative to average sales
across the period of observation.2 As a robustness test of the significance of the effect, we
perform an additional analysis, which is described below.
Robustness Check
To account for the substitution patterns among wallets in the treatment and control
categories, we create an additional set of specifications that capture the daily percentage of all
wallets sold that correspond with the treatment category, TSALES_PCTt. Importantly, we note
that the ratio of wallets sold in the treatment and control conditions was time-invariant prior to
the introduction of cost transparency.
To facilitate our analysis, we also created an aggregated set of control variables
consistent with those described above, which reflect total page views and total page views
squared in each category, TVIEWSt, TVIEWS2t, CVIEWSt, and CVIEWS2t, the percentage of
colors in each category with no visits, TNOVIEWSt and CNOVIEWSt , the one-day lagged
percentage of colors in each category with no visits, TNOVIEWSt-1 and CNOVIEWSt-1, and the
one-day lagged percentage of colors in each category with no sales, TNOSALEt-1 and
CNOSALEt-1. We estimate the following linear specification with robust standard errors:
2
Given that sales in the control and treatment conditions are not completely independent, customers may substitute
a purchase of one wallet color for another. Hence, the magnitude of the effects calculated through conventional
difference-in-difference calculations would be overstated if any substitution occurs. The conservative estimate of a
22.0% increase in sales assumes 100% substitution. If on the other hand, every incremental customer who purchased
a wallet in the treatment condition would have not purchased a wallet in the absence of the treatment (i.e., 0%
substitution), then cost transparency increased sales by 44.0%. Therefore under both conservative and liberal
assumption about substitution, the cost transparency treatment had a significant positive effect on sales.
13
⎛ γ 0 + γ 1POSTt + γ 2TVIEWSt + γ 3TVIEWS 2t + γ 4CVIEWSt + γ 5CVIEWS 2t + ⎞
⎜
⎟
TSALES_PCTt = f ⎜ γ 6TNOVIEWSt + γ 7CNOVIEWSt + γ 8TNOVIEWSt−1 + γ 9CNOVIEWSt−1 + ⎟
⎜ γ TNOSALE + γ CNOSALE + ε
⎟
⎝ 10
⎠
t−1
11
t−1
t
!
(2)
(Insert Table 2 about here)
The results of this supplemental analysis are shown in Table (2). In all columns, the
variable of interest is the coefficient on POSTt, which indicates the change in the percentage of
sales coming from colors in the treatment category after the introduction of cost transparency. In
Column (1) the base specification reveals that the percentage of wallets sold in the treatment
category, relative to the sales of all wallets, rises 12.6% following the introduction of cost
transparency (coefficient = 0.126; p<0.01 two-tailed).
Columns (2-5) reveal that this significant increase in sales percentage is robust to the
inclusion of controls for the number of times products in the treatment and control conditions
were visited during the day, the percentage of colors in each category that received no visits on
the focal or preceding day, and the percentage of colors in each category that resulted in no sales
on the previous day. The best estimate provided by the full-specified version of this model
(Column 5), is that cost transparency increased the percentage of total sales represented by the
treatment wallets by 15.7% (p<0.01 two-sided).
STUDY 1B: LAB REPLICATION OF FIELD STUDY
Study 1A, a natural experiment involving a real online retailer, provides preliminary
evidence of cost transparency’s capacity to boost sales. Cost transparency was applied to some,
but not all wallets, which provided a natural experiment opportunity to exploit. However, the
information presented for the treatment versus control wallets differed in additional ways beyond
14
our conceptual definition of cost transparency: the treatment wallets also included competitive
benchmarks. Study 1B addresses this imprecision in a lab experiment using the same interface as
Study 1A but modified to distill the critical ingredients of cost transparency which we posit to be
causally responsible for boosting purchase interest.
Method
Design and Procedure. Participants (N = 322, Mage = 36.8, 51% male) completed this
experiment on Amazon’s Mechanical Turk (Mturk), which is an online subject pool, in exchange
for a small fixed payment. Participants were randomly assigned to one of two experimental
conditions: no transparency, which served as our control condition, or cost transparency. In the
control condition, participants saw a screen shot of the nontransparent wallets from Study 1A. In
the cost transparency condition, the product page also included similar cost information as the
field study: materials ($11.00), hardware ($3.60), labor ($38.56), duties ($4.21) and
transportation ($1.00). Notably, unlike the field study, the infographic did not include the sum of
the costs or competitive benchmark information and is therefore a purer representation of cost
transparency (Figure 2).
(Insert Figure 2 about here)
Dependent Measures. Participants indicated their willingness to buy the wallet by
responding to the item: “Given the opportunity, how likely would you be to purchase this
product?” (7-point response scale; 1 = Not at all likely to 7 = Very likely). This and all
subsequent experiments included attention checks (assessing whether participants correctly
identified the condition-specific information they had been shown) as well as demographics (age,
gender, education, income). The effects are substantively equivalent when these variables are
15
taken into account, and hence, we do not discuss them further. We set the desired number of
participants at the outset of each experiment and did not analyze the data until that number was
reached. No data were excluded and we report all manipulations and measures.
Results
Willingness to buy. Consistent with the field evidence, willingness to buy was greater in
the cost transparency condition relative to the control condition (Mcost = 2.69, SD = 1.81; Mcontrol=
2.26, SD = 1.72; t(322) = 2.20, p < 0.05).
In sum, Study 1B replicated Study 1A in a controlled setting in which the extraneous
information included in the field study was omitted from the treatment.3
STUDY 2: COST VERSUS OPERATIONAL TRANSPARENCY
Studies 1A and 1B suggest that cost transparency can boost sales. In Study 2, we
ascertain whether revealing costs improves purchase interest above and beyond merely showing
the operational steps to produce a good. Control participants were simply shown a t-shirt and
asked to indicate their willingness to buy it. In the operational transparency condition,
participants were informed of the steps that went into making the shirt. Finally, in the cost
transparency condition, participants were additionally informed of the cost associated with each
of the steps. For robustness, orthogonal to transparency, we also manipulated the price (holding
costs constant): the t-shirt was priced at $10, $15, or $20. The study was therefore a 3x3
between-subjects design.
3
Although the retail websites used in the primary studies reported in this paper do not rely on familiar brands, in
Appendix 3, we report the results of an additional study that provides evidence consistent with our field research
indicating that the effects of cost transparency extends to both familiar and unfamiliar brands.
16
Method
Design and Procedure. Participants (N = 272, Mage = 31.5, 62% male) completed this
experiment on Mturk in exchange for a small fixed payment. Participants first indicated their
gender. Then, they were told that they would see a simulated retail website page for a product
and would be asked to indicate their purchase interest. Next, participants were randomly
assigned to one of nine experimental conditions of the 3 (Transparency: Control, Operational,
Cost) x 3 (Price: $10, $15, $20) between-subjects design. The prices were comparable to other
similar t-shirts found online at the time of the experiment, and our displayed costs were informed
by a company that invokes cost transparency on its website (Everlane 2014).
In the control condition, participants saw a baseline interface depicting a model (same
gender as participant) wearing the t-shirt, the name of the product, price, and available colors and
sizes (Figure 3, Panel A). The experimental conditions also included an infographic entitled
“What goes into the production of our Women’s [Men’s] V?” and depicting six operational
steps: cotton, cutting, sewing, dyeing, finishing, and transport. The cost transparency condition
additionally displayed the unit cost of each of the six operational steps (i.e., $2.75, $0.35, $1.35,
$0.50, $1.25, and $0.50, respectively; Figure 3, Panels B and C).
(Insert Figure 3 about here)
Dependent Measures. Participants indicated their willingness to buy the t-shirt using the
same item as Study 1B. Participants also answered a series of secondary questions assessing their
purchase experience4 (Appendix 1).
(Insert Figure 4 about here)
4
We note that results are substantively similar after controlling for individual items in the Appendix. Covariates
seldom reduced the significance of the focal independent variables, no more than would be expected by chance.
Thus, we do not describe them further.
17
Results
We conducted a 3 (Price: $10, $15, $20) x 3 (Transparency: Control, Operational, Cost)
analysis of variance (ANOVA) on willingness to buy (Figure 4). Not surprisingly, there was a
main effect of price: willingness to buy decreased as price increased (F(2,263) = 12.29, p <
0.01). More importantly, there was a main effect of transparency (F(2,263) = 4.69, p = 0.01).
Specifically, willingness to buy was greater in the cost transparency condition relative to both the
control condition (Mcontrol = 3.31, SD = 1.87; Mcost = 4.16, SD = 1.98; t(177) = 2.95, p < 0.01) and
the operational transparency condition (Moperational = 3.57, SD = 1.80; t(180) = 2.10, p = 0.04).
There was no interaction between transparency and price (F(4,263) = 0.64, p = 0.64).
In sum, Study 2 shows that disclosing costs can improve purchase interest above and
beyond merely showing the operational steps to produce a good, and that this effect holds across
different prices (and hence, profit margins).
STUDY 3: MECHANISM
Studies 1 and 2 provide evidence that cost transparency can increase sales. Study 3
replicates and extends this result by testing the theorized process underlying the effect of cost
transparency on purchase interest. Specifically, we test whether the effect is mediated by
consumer trust toward the firm (John, Barasz, and Norton 2016; Wheeless and Grotz 1977).
Method
Design and Procedure. Participants (N = 612, Mage = 35.2, 55.1% male) completed this
experiment on Mturk in exchange for a small fixed payment. Participants were randomly
18
assigned to one of two experimental conditions varying in cost transparency: no transparency,
which served as our control condition, versus cost transparency. In the control condition,
participants were shown a graphic depicting the front and back of a chocolate bar package. We
worked with a chocolate manufacturer and retailer to develop a package for a fictitious brand
called “Cocoa Passion” with realistic cost information. In the control condition, a description of
the bar (70% cacao), flavors, ingredients, and nutrition facts were listed on the packaging. In the
cost transparency condition, on the back of the packaging, participants were also shown the
actual unit cost of each of the six cost components: $0.29 (beans), $0.03 (sugar), $1.39 (cocoa
butter), $0.17 (packaging), $0.90 (labor), and $0.11 (utilities). The total cost of these
components, $2.89, was also prominent (Figure 5).
(Insert Figure 5 about here)
Dependent Measures. We measured trust by asking participants: “How trustworthy
would you consider the firm?” (c.f. John, Barasz and Norton (2015)), using a 7-point response
scale (1 = Not at all trustworthy – 7 = Very trustworthy). On the subsequent screen, willingness
to buy was measured as in Studies 1B and 2.
Results
Willingness to buy. Cost transparency increased willingness to buy relative to the control
condition (Mcost = 4.27, SD = 2.00; Mcontrol= 3.74, SD = 2.04; t(610) = 3.26, p < 0.01).
Trust in Firm. Trust was greater in the cost transparency condition relative to the control
condition (Mcost = 5.27, SD = 1.38; Mcontrol= 4.82, SD = 1.38; t(610) = 4.02, p < 0.01).
Mediation Analysis. Cost transparency predicted both trust (β = 0.45, p < 0.01) and
willingness to buy (β = 0.53, p < 0.01). When trust and cost transparency were both included in
19
the model predicting willingness to buy, trust remained significant (β = 0.80, p < 0.01), but cost
transparency was reduced to non-significance (β = 0.17, p = 0.21) providing support for full
mediation. We used a bootstrap procedure to construct bias-corrected confidence intervals for the
indirect effect based on 5,000 resamples (Preacher and Hayes 2008). The 95% bias-corrected
confidence interval excluded zero (0.18, 0.55), suggesting a significant mediation effect.
In sum, Study 3 demonstrates the capacity for cost transparency to boost sales within yet
another product category and provides explicit evidence of the process underlying the effect.
STUDY 4: COST TRANSPARENCY AND DUAL ENTITLEMENT
Taken together, Studies 1-3 provide evidence that cost transparency increases purchase
interest, and that it does so via consumer perceptions of firm trustworthiness. In Study 4, we
show that our account of cost transparency explains variance not accounted for by the principle
of dual entitlement. First, dual entitlement predicts that the capacity for cost salience (which can
be achieved through cost transparency) to increase purchase interest is driven by perceptions of
price fairness. Thus if cost transparency’s effect on purchasing arises simply because it makes
costs salient, then it should be driven solely by price fairness. If, as we argue, there is something
special about a firm’s voluntary disclosure of such information, then additional variance should
be explained via consumer trust in the firm. Study 4 tests these possibilities by measuring both
price fairness and trust as potential mediators.
Secondly, Study 4 exploits a circumstance in which these accounts make different
predictions of the effect of cost transparency on purchase interest. Specifically, dual entitlement
predicts that when prices are higher than consumers’ expectations, revealing costs will reinforce
20
that the firm’s profit is relatively high, and hence purchase interest will decrease (Bolton,
Warlop, and Alba 2003; Kahneman, Knetsch, and Thaler 1986). If, as we predict, there is
something special about a firm’s disclosure of its costs, then cost transparency should increase
purchase interest even when the price is higher than expectations.
In Study 4, participants were first asked to estimate the price of a travel package. Next,
the actual price was revealed, and manipulated to be either higher or lower than the guessed
price. Thus, this setup enabled us to manipulate whether the price was surprisingly high or
surprisingly low for a given participant. Half of participants were subsequently randomized to
also receive the firm’s disclosure of its costs. The study was therefore a 2x2 condition betweensubjects design in which we manipulated price relative to expectations (Low versus High) and
transparency (None versus Cost).
Method
Participants (N = 419, MAge = 34.6, 55.1% male) completed this experiment on Mturk in
exchange for a small fixed payment. First, participants saw the description of a travel package
for a guided 6-night trip to Washington, D.C. consisting of: 6 nights of accommodations, select
meals, welcome and farewell receptions, and gratuities. Participants were then told to “Guess the
price of this tour,” and provided with a small text-box to input their estimate.
Manipulation of Price Relative to Expectations. After having estimated the price of the
trip, participants were told the actual price of the trip, which we manipulated to be either 25%
below the given participant’s estimated price (low) or 25% above the given participant’s
estimated price (high). For example, a participant who had estimated a price of $1000 would
subsequently be told: “The actual price of this trip is $1250” if randomized to the high price
21
condition, or “The actual price of this trip is $750” if randomized to the low price condition.
Manipulation of Cost Transparency. In the no transparency condition, participants were
shown no additional information. In the cost transparency condition, participants were told:
“The tour operator voluntarily reveals their costs for each component of the tour,
on a website that is available for prospective clients. That is, prospective clients
can see how much each component below costs the tour operator to supply.”
Subsequently, participants in the cost transparency condition were shown the cost of each
component of the trip, plus the sum of those costs (i.e., the total cost of the trip). The values were
dynamic based on the price condition, such that total costs represented 80% of the price. For
instance, in the low price condition, a participant who had estimated a price of $1000 was told
that the price was $750 and that costs totaled $600 (i.e., $150 in profits); in the high price
condition, the price in this situation would be $1,250, with the costs totaling $1000 (i.e., $250 in
profits) (Appendix 2).
Dependent Measures. On the following screen, participants were asked: “How likely
would you be to buy this tour?” (7-point response scale: 1 = Not at all likely to 7 = Very likely).
On the next two pages, we measured the two possible mediators. For trust, participants were
asked: “How trustworthy is this travel agency?” (7-point response scale: 1 = Not at all
trustworthy to 7 = Very trustworthy). For price fairness, participants were asked: “How fair is the
price of the trip?” (7-point response scale: 1 = Not at all fair to 7 = Very fair; c.f. Bolton,
Warlop, and Alba (2003)).
Results
Price Estimates. Participants estimated an average trip price of $1,537.55 (SD = 992.33).
22
The median estimate was $1,400.00. The bottom quartile of participants estimated prices below
$850.00, while the top quartile of participants estimated prices above $2,000.00.
Willingness to buy. We conducted a 2 (Price: Low versus High) x 2 (Transparency: None
versus Cost) ANOVA on willingness to buy. Not surprisingly, there was a main effect of price:
willingness to buy was higher when prices were lower than expected relative to when they were
higher than expected (F(1,415) = 27.91, p < 0.01). More importantly, there was a main effect of
cost transparency (F(1, 415) = 6.79, p = 0.01). The interaction between transparency and price
was not significant (F(1, 415) = 0.08, p = 0.79). In other words, the positive effect of cost
transparency on purchase interest held across prices (Figure 6).
(Insert Figure 6 about here)
Mediation Analysis. Cost transparency predicted both trust in the firm (β = 0.34, p < 0.01)
and willingness to buy (β = 0.48, p = 0.01). We used a bootstrap procedure to construct biascorrected confidence intervals for the indirect effect based on 5,000 resamples, with transparency
as the independent variable, trust as the mediator, and purchase intention as the dependent
variable (Preacher and Hayes 2008). The 95% bias-corrected confidence interval excluded zero
(-0.41, -0.07), suggesting a significant mediation effect of trust. We then used a bootstrap
procedure to construct bias-corrected confidence intervals for the indirect effect based on 5,000
resamples, with transparency as the independent variable, price fairness as the mediator, and
purchase intention as the dependent variable (Preacher and Hayes 2008). The 95% bias-corrected
confidence interval excluded zero (-0.30, -0.02), also suggesting a significant mediation effect of
price fairness. We subsequently used a bootstrap procedure including both trust and fairness as
mediators, and find that the trust mediation path is significant even when controlling for
perceptions of price fairness; the 95% bias-corrected confidence interval again excluded zero (-
23
0.23, -0.03).
Consistent with the principle of dual entitlement, Study 4 shows that price fairness indeed
predicts purchase interest, and that it is partly responsible for the effect of cost transparency on
purchase interest. Importantly however, as we show, it does not fully explain the effect of cost
transparency on purchase interest; consistent with our theoretical account, additional variance is
explained by consumers' perceptions of firm trustworthiness. Moreover, the benefit of cost
transparency persists even when costs reinforce that profits are surprisingly high. Therefore
Study 4 provides direct evidence that cost transparency’s ability to boost sales does not lay
merely in the increased focus it draws to costs.
STUDY 5: VOLUNTARY VERSUS INVOLUNTARY TRANSPARENCY
Study 5 tests a critical implication of our theoretical account: disclosure needs to be
voluntary in nature in order to reap the benefit of cost transparency on sales. To test this
implication, in Study 5 participants were shown a simulated retail website page for each of two
different firms, the target firm and a competitor firm, each selling a comparable t-shirt.
Participants were asked from which firm they would prefer to buy the t-shirt. We assessed
participants’ propensity to choose the target firm as a function of whether and why it had
disclosed its costs, and whether and why the competing firm had disclosed its costs.
Specifically, in the involuntary versus nontransparent condition, the target firm had
disclosed its costs because regulation required it, whereas the competitor had made no such
disclosure. In the voluntary versus nontransparent condition, the target firm had voluntarily
disclosed its costs, whereas the competitor had made no such disclosure. We also included a
24
third, voluntary versus involuntary condition to test whether the benefit of voluntary disclosure
might be augmented in a context in which the competitor involuntarily disclosed its costs.
We predicted that participants would prefer to buy from a firm that does not disclose its
costs relative to one that does so involuntarily. In other words, we thought participants would
generally avoid purchasing from the target firm in the involuntary versus nontransparent
condition. Second, consistent with the preceding studies, we predicted that compared to a
nontransparent firm, participants would prefer to buy from a firm that voluntarily discloses its
costs. In other words, we thought participants would generally prefer to purchase from the target
firm in the voluntary versus nontransparent condition. Finally, we predicted that the tendency to
purchase from a firm that voluntarily discloses its costs would be augmented in a context in
which the competing firm only does so involuntarily. In other words, we thought participants
would be even more likely to choose the target firm in the voluntary versus involuntary condition
relative to the voluntary versus nontransparent condition.
Method
Design and Procedure. Participants (N = 617, Mage = 35.2, 56% male) completed this
experiment on Mturk in exchange for a small fixed payment. Participants first indicated their
gender. Then, they were shown two similar simulated retail website pages for two different
retailers, target Brand A and competitor Brand B, and told that the retailers operated in different
countries but were selling a comparable t-shirt for $15.
Between-subjects we varied whether and why Brand A had disclosed its costs, as well as
whether and why Brand B had disclosed its costs. Specifically, in the involuntary versus
nontransparent condition, Brand A’s page featured an infographic indicating that the total cost of
25
cotton, cutting, sewing, dyeing, finishing, and transport was $6.70 (Figure 7) and participants
were told that “Due to regulations in the country this brand is based in, the company is forced to
disclose its costs to customers. If regulation didn’t require it, this company would choose to NOT
disclose its costs to its customers.” By contrast, Brand B’s page contained no such information.
In the voluntary versus nontransparent condition, Brand A’s page featured voluntary
transparency, which consisted of the same infographic as when it had been voluntary, but
included a different rationale: “Due to the desire to be transparent to its customers, the company
voluntarily discloses its costs to its customers.” By contrast, Brand B’s page contained no such
information. Finally, in the voluntary versus involuntary condition, Brand A’s page featured
voluntary cost transparency, whereas Brand B’s page featured involuntary cost transparency.
(Insert Figure 7 about here)
Between-subjects we also counterbalanced the ascription of brands to target versus
competitor brand status, which produced four parallel conditions: (target) Brand B voluntary
versus (competitor) Brand A nontransparent; Brand B involuntary versus Brand A
nontransparent; Brand B voluntary versus Brand A involuntary. The effects were invariant to
counterbalancing; therefore we collapse across this independent variable in the results.
Dependent Measures. Participants indicated from which firm they would prefer to buy
the t-shirt by responding to the item: “Given the opportunity, from which firm would you
purchase this t-shirt?” (Choices: Brand A and Brand B).
(Insert Figure 8 about here)
Results. First, participants generally avoided buying from a firm that involuntarily
discloses its costs, when compared to one that does not disclose its costs. Specifically, in the
involuntary versus nontransparent condition, 35% (Χ2 = 17.89, p < 0.01 against indifference
26
point of 50%) chose the target brand (which had involuntarily disclosed its costs). Second,
participants generally preferred to buy from a firm that voluntarily discloses its costs when
compared to one that does not disclose its costs. Specifically, in the voluntary versus
nontransparent condition, 63% (Χ2 = 13.12, p < 0.01 against indifference point of 50%) chose the
target brand (which had instated voluntary cost transparency). Finally, the preference to buy
from a firm that voluntarily discloses its costs was augmented when compared to one that does
so only because regulation requires it. Specifically, in the voluntary versus involuntary condition,
82% chose the target brand. This propensity to choose the voluntarily cost transparent brand is
significantly higher than that in the voluntary versus nontransparent condition (i.e., 82% against
63%, Χ2 = 18.51, p < 0.01) (Figure 8).
GENERAL DISCUSSION
In this article, we examined the effect of cost transparency on purchase interest. We
began with a natural field experiment conducted with an online retailer, and found that cost
transparency increased sales (Study 1A). Five subsequent controlled lab experiments replicated
and extended this basic effect (Studies 1B – 5). Specifically, Study 1B replicated the field
experiment, narrowing the key ingredients of cost transparency that increase purchase interest.
Study 2 demonstrated that cost transparency is particularly potent in boosting purchase interest
compared to operational transparency alone. Studies 3 and 4 demonstrated that the positive effect
of cost transparency is mediated by perceptions of firm trustworthiness, with Study 4 showing
that this mediator explains variance above and beyond perceptions of price fairness. Study 4 also
demonstrated that cost transparency increases purchase intent even when the costs are revealed
27
to be lower than, and hence the profits higher than, consumers’ expectations. Notably, this result
is not predicted by dual entitlement. Finally, Study 5 demonstrated the critical role of the
voluntary nature of the disclosure, by showing that cost transparency boosts purchase interest
only when costs are voluntarily – as opposed to involuntarily – revealed by the firm. We
demonstrated our effect using different instantiations of cost transparency, encompassing both
total cost, and cost broken down into smaller components. We also demonstrated our effect
across a variety of different brands and product categories, including wallets, t-shirts, chocolate,
and travel packages.
Limitations
From a practical standpoint, there are important caveats a retailer would need to consider
before deciding to reveal its costs. Firms may not want to disclose their costs if their cost
structure is a competitive advantage. Moreover, a firm’s suppliers may not allow the firm to
make the costs associated with certain components public information. Thus, there could be
strategic risks or contractual barriers to disclosure.
Even if firms want and are able to disclose their costs, they might not have the knowledge
to do so. Disclosing the unit costs associated with the production of a single good could be
particularly difficult for retailers that are not vertically integrated. Although a retailer might be
transparent about certain aspects of price, such as taxes (Carter and Curry 2010), it might not be
feasible to comprehensively break out the costs of goods produced by a wide range of
manufacturers.
Open Questions and Concluding Comment
28
This research focuses on a set of contexts where unit costs can be readily calculated and
explained. For goods and services that are dependent on high fixed costs (R&D, overhead,
constant labor costs, etc.), imputing unit costs may be complicated or even confusing to
consumers. For example, imputing R&D costs on a per unit basis in a pharmaceutical context
may require many assumptions, as well as a campaign to educate customers about how the sales
of successful pharmaceuticals subsidize the production costs of less popular products as well as
the costs of early-stage trials and failures. We leave further investigation of the effect of cost
transparency in high fixed cost environments as an opportunity for future research.
In our research, all participants were American, with costs presented in US dollars. It is
possible that the evaluability of different currencies could moderate consumer choices when
faced with cost transparency (Hsee 1996; Raghubir and Srivastava 2002). Customers may also
make negative inferences if costs are presented in other currencies that indicate low wages and
poor working conditions (Paharia, Vohs, and Deshpandé 2013). Thus, future research could
explore the impact of framing cost transparency in different currencies.
Our research examines the effects of cost transparency for a single firm in a single industry.
Future investigation could explore what would happen if consumers evaluated two or more
products sold by the same firm, with varying levels of transparency. Cost transparency might
also seem less intimate in competitive environments when other firms also disclose their costs.
Thus, future work could further explore the competitive ramifications of cost transparency, to
better understand what happens in markets when multiple players reveal their costs for similar
products. There is still much to explore about the benefits and potential pitfalls of cost
transparency.
29
REFERENCES
Aron, Arthur, Edward Melinat, Elaine N. Aron, Robert D. Vallone, and Renee J. Bator (1997),
“The Experimental Generation of Interpersonal Closeness,” Personality and Social
Psychology Bulletin, 23 (4), 363–77.
Bazerman, Max H. (1985), “Norms of Distributive Justice in Interest Arbitration,” Industrial &
Labor Relations Review, 38 (4), 558–70.
Bertini, Marco and Luc Wathieu (2008), “Research Note-Attention Arousal Through Price
Partitioning,” Marketing Science, 27 (2), 236–46.
Bolton, Lisa E. and Joseph W. Alba (2006), “Price Fairness : Good and Service Differences and
the Role of Vendor Costs,” Journal of Consumer Research, 33 (2), 258–65.
———, Luk Warlop, and Joseph W. Alba (2003), “Consumer Perceptions of Price (Un)
Fairness,” Journal of Consumer Research, 29 (4), 474–91.
Buell, Ryan, Chia-jung Tsay, and Tami Kim (2014), “Creating Reciprocal Value Through
Operational Transparency.”
Buell, Ryan W. and Michael I. Norton (2011), “The Labor Illusion : How Operational
Transparency Increases Perceived Value,” Management Science, 57 (9), 1564–79.
Carter, Robert E. and David J. Curry (2010), “Transparent Pricing: Theory, Tests, and
Implications for Marketing Practice,” Journal of the Academy of Marketing Science, 38 (6),
759–74.
30
Chinander, Karen R and Maurice E Schweitzer (2003), “The Input Bias: The Misuse of Input
Information in Judgments of Outcomes,” Organizational Behavior and Human Decision
Processes, 91 (2), 243–53.
Collins, Nancy L. and Lynn C. Miller (1994), “Self-disclosure and Liking: A Meta-Analytic
Review.,” Psychological Bulletin, 116 (3), 457–75.
Everlane (2014), “Tees and Tanks,” (accessed March 4, 2014), [available at
https://www.everlane.com/collections/womens-tees].
Gershoff, Andrew D., Ran Kivetz, and Anat Keinan (2012), “Consumer Response to Versioning:
How Brands’ Production Methods Affect Perceptions of Unfairness,” Journal of Consumer
Research, 39 (2), 382–98.
Hsee, Christopher K (1996), “The Evaluability Hypothesis: An Explanation for Preference
Reversals Between Joint and Separate Evaluestions of Alternatives,” Organizational
Behavior and Human Decision Processes, 67 (3), 247–57.
John, Leslie K, Kate Barasz, and Michael I Norton (2016), “Hiding Personal Information
Reveals the Worst,” Proceedings of the National Academy of Sciences of the United States
of America, 6 (8), 1–6.
Kahneman, Daniel, Jack L. Knetsch, and Richard Thaler (1986), “Fairness as a Constraint on
Profit Ceeking: Entitlements in the Market,” The American Economic Review, 76 (4), 728–
41.
31
Kőszegi, Botond and Matthew Rabin (2015), “A Model of Reference-Dependent Preferences,”
The Quarterly Journal of Economics, 121 (4), 1133–65.
Lamming, Richard C., Nigel D. Caldwell, Deborah A. Harrison, and Wendy Phillips (2002),
“Transparency in Supply Relationships: Concept and Practice,” IEEE Engineering
Management Review, 30, 70–76.
Laurenceau, Jean-Phillipe, Lisa F. Barrett, and Paula R. Pietromonaco (1998), “Intimacy as an
Interpersonal Process: The Importance of Self-disclosure, Partner disclosure, and Perceived
Partner Responsiveness in Interpersonal Exchanges.,” Journal of Personality and Social
Psychology, 74 (5), 1238–51.
Morales, Andrea C. (2005), “Giving Firms an ‘E’ for Effort: Consumer Responses to High‐Effort
Firms,” Journal of Consumer Research, 31 (4), 806–12.
Morwitz, Vicki G., Eric A. Greenleaf, and Eric J. Johnson (1998), “Divide to and Prosper :
Consumers’ Reactions to Partitioned Prices,” Journal of Marketing Research, 35 (4), 453–
63.
Newey, Whitney K and Kenneth D West (1987), “A Simple, Positive Semi-definite,
Heteroskedasticity and Autocorrelation Consistent Covariance Matrix,” Econometrica
(1986-1998), 55, 703.
Paharia, Neeru, Kathleen D. Vohs, and Rohit Deshpandé (2013), “Sweatshop Labor is Wrong
Unless the Shoes are Cute: Cognition can both Help and Hurt Moral Motivated Reasoning,”
Organizational Behavior and Human Decision Processes, 121 (1), 81–88.
32
Preacher, Kristopher J. and Andrew F. Hayes (2008), “Asymptotic and Resampling Strategies for
Assessing and Comparing Indirect Effects in Multiple Mediator Models,” Behavior
Research Methods, 40 (3), 879–91.
Raghubir, Priya and Joydeep Srivastava (2002), “Effect of Face Value on Product Valuation in
Foreign Currencies,” Journal of Consumer Research, 29 (3), 335–47.
Schaffer, Mark E. (2010), “xtivreg2: Stata Module to Perform Extended IV/2SLS, GMM and
AC/HAC, LIML and K-Class Regression for Panel Data Models,” Statalist.
Sedikides, Constantine, Keith Campbell, Glenn D. Reeder, and Andrew J. Elliot (1999), “The
Relationship Closeness Induction Task.,” Representative Research in Social Psychology,
23, 1–4.
Wheeless, Lawrence R. and Janis Grotz (1977), “The Measuremet of Trust and Its Relationship
to Self-Disclosure,” Human Communication Research, 3 (3), 250–57.
Zhu, Kevin (2004), “Information Transparency of Business-to-Business Electronic Markets: A
Game-Theoretic Analysis,” Management Science, 50 (5), 670–85.
33
(1)
(2)
(3)
(4)
Treatment Treatment Treatment Treatment
Sales %
Sales %
Sales %
Sales %
Post
-0.945***
(0.273)
-0.921***
(0.258)
-0.959***
(0.257)
-0.850***
(0.238)
0.523
(0.331)
0.579*
(0.330)
0.660**
(0.328)
0.582*
(0.311)
Views
0.0591*
(0.0312)
0.0488
(0.0316)
0.0451
(0.0302)
Views2
-0.0002*
(0.0001)
-0.0002
(0.0001)
-0.0002
(0.0001)
No views
-0.321**
(0.153)
-0.300**
(0.152)
Lagged no views
-0.158
(0.180)
-0.113
(0.183)
Post ⨉ Treatment
Lagged no sale
Observations
Adjusted R-squared
-0.460***
(0.142)
460
0.044
460
0.050
460
0.052
460
0.067
Table 1: Field Evidence (Study 1A). Units sold on a daily basis, by transparency condition.
Treatment variable subsumed by fixed effects estimation. Fixed effects coefficients withheld to
protect confidential company information. *p<0.10, **p<0.05, ***p<0.01, two-tailed. Robust
Newey West standard errors, accounting for autocorrelation and heteroskedasticity within colors,
shown in parentheses.
34
(1)
(2)
(3)
(4)
Treatment Treatment Treatment Treatment
Sales %
Sales %
Sales %
Sales %
Post
0.126***
(0.047)
0.117**
(0.048)
0.138***
(0.051)
0.157***
(0.052)
Treatment: Visits
0.002
(0.007)
0.004
(0.008)
0.001
(0.008)
Treatment: Visits2
-0.000
(0.000)
-0.000
(0.000)
-0.000
(0.000)
Control: Visits
-0.008**
(0.003)
-0.005
(0.004)
-0.005
(0.004)
Control: Visits2
0.000*
(0.000)
0.000
(0.000)
0.000
(0.000)
Treatment: No visits %
0.131
(0.130)
0.181
(0.131)
Control: No visits %
0.140
(0.095)
0.128
(0.093)
Treatment: Lagged no visits %
-0.070
(0.116)
-0.021
(0.116)
Control: Lagged no visits %
0.142
(0.095)
0.149
(0.096)
Treatment: Lagged no sales %
-0.169
(0.105)
Control: Lagged no sales %
-0.077
(0.068)
Constant
Observations
Adjusted R-squared
0.662***
(0.030)
0.713***
(0.058)
0.608***
(0.085)
0.681***
(0.090)
92
0.063
92
0.114
92
0.134
92
0.165
Table 2: Field evidence supplemental analysis (Study 1A). Daily percentage of wallet sales
attributable to the treatment condition (as a percentage of all wallets sold). *p<0.10, **p<0.05,
***p<0.01, two-tailed. Robust standard errors shown in parentheses.
35
1
(A)
Treatment
6-Mar
4-Mar
2-Mar
28-Feb
26-Feb
24-Feb
22-Feb
20-Feb
18-Feb
16-Feb
14-Feb
12-Feb
8-Feb
10-Feb
6-Feb
4-Feb
2-Feb
31-Jan
29-Jan
27-Jan
25-Jan
23-Jan
21-Jan
19-Jan
17-Jan
15-Jan
13-Jan
9-Jan
11-Jan
7-Jan
5-Jan
3-Jan
1-Jan
30-Dec
28-Dec
26-Dec
24-Dec
22-Dec
20-Dec
18-Dec
16-Dec
14-Dec
12-Dec
8-Dec
10-Dec
6-Dec
4-Dec
2-Dec
Average Daily Unit Sales per Color
Control
(B)
Average Unit Sales Per Color
Treatment
Pre Period
Control
Post Period
Figure 1: Field Evidence (Study 1A). Average daily unit sales by treatment versus control (Panel
A). Average daily unit sales by treatment versus control, collapsed across days (Panel B).
Dashed vertical line indicates the date the cost transparency infographic was added. Sales figures
withheld to protect confidential company information.
(A)
(B)
Figure 2: Stimuli for Study 1B. Panel A depicts the simulated wallet website for the control
condition; Panel B depicts the simulated wallet website for the cost transparency condition.
(A)
Control Condition Screen (Male)
Control Condition Screen (Female)
(B)
(C)
Figure 3: Stimuli for Study 1. Panel A depicts the simulated t-shirt website for all conditions
(screens were matched to the gender of the participant). Participants either saw no additional
infographic (control), the additional operational transparency infographic (Panel B; operational
transparency condition), or the additional cost transparency infographic (Panel C).
7
Control
Operational Transparency
6
Willingness to Buy
Cost Transparency
5
4
3
4.8
4.0
4.5
4.2
3.2
2
3.5
3.0
2.8
3.1
1
$10.00
$15.00
Price
$20.00
Figure 4: Willingness to buy was greater in the cost transparency condition relative to both the
control condition and the operational transparency condition (Study 2).
(A)
(B)
(C)
Figure 5: Chocolate package designs presented as stimuli. All participants saw the same ‘front’
packaging (A), and either control (B) or cost transparency (C) ‘back’ packaging (Study 3).
7
Willingness to Buy
6
5
4
Control
Transparent
3
4.4
3.9
2
3.4
3.0
1
Low
High
Price Relative to Expectation
Figure 6: The positive effect of cost transparency on purchase interest held across prices (Study
4).
Figure 7: Example of the simulated t-shirt website (screens were matched to the gender of the
participant) accompanied by disclosure of the total cost to produce the t-shirt ($6.55). The cost
transparency disclosure was identical regardless of whether the disclosure was framed as
voluntary or involuntary (Study 5).
100%
90%
80%
70%
60%
50%
82%
40%
63%
30%
20%
35%
10%
0%
Target:
Involuntary
Competitor:
None
Target: Voluntary
Competitor:
None
Target: Voluntary
Competitor:
Involuntary
% Choosing Target Transparent Brand
Figure 8: Participants preferred to buy from a firm that voluntarily discloses its costs relative to
one that does not do so, and relative to one that does so only involuntarily (Study 5).
Appendix 1
Additional Measures (Study 2)
1. How likely would you be to purchase from this website in general, either this item or
another?
2. What was the price of the product you saw?
3. How much do you think it cost to make this shirt?
4. How attractive do you find this product?
5. How reasonable is the price of this product?
6. My feelings toward this retailer can best be described as: (very unsatisfied - unsatisfied).
7. If it were made available to me, over the next year, my use of this retailer would be: (very
infrequent - very frequent).
8. This item is well made.
9. This item is one that would make me feel good.
10. This item is a good product for the price.
11. This item would make a good impression on other people.
12. This item is too expensive.
13. This item has consistent quality.
14. This item would not last a long time.
15. This site appears more trustworthy than other sites I’ve visited.
16. The site represents a company or organization that will deliver on promises made.
17. My overall trust in this product website is: (very low - very high).
18. My overall believability of the information on this site is: (very low - very high).
19. This site represents a company that engages in ethical business practices.
20. This site represents a company that pays its workers a fair wage.
Appendix 2
The cost breakdown (as a percent of the price displayed) was determined as follows: 6 days of
guided tours (13.6%), 6 nights of accommodations (40%), select meals (13.6%), welcome and
farewell receptions (6.4%), and gratuities (6.4%). Thus the total cost came to 80% of the
estimated price.
Low Price Cost Transparency Screen:
• 6 days of guided tours to the Smithsonian, Ford’s Theater, Senate, Mount Vernon, major
monuments, & Newseum [Cost to the tour operator: (Participant estimate) * 0.10 ]
• 6 nights of accommodations [Cost to the tour operator: (Participant estimate ) * 0.30]
• 6 breakfasts, 2 lunches and 3 dinners [Cost to the tour operator: (Participant estimate )
* 0.10]
• Welcome and farewell receptions [Cost to the tour operator: (Participant estimate) *
0.05]
• Gratuities to porters, waiters, guides and drivers for all group activities [Cost to the tour
operator: (Participant estimate ) * 0.05]
The total cost of the trip to the tour operator is [(Participant estimate ) * 0.60] dollars (and as
noted on the previous screen, the travel company is selling the trip for a price of [Participant
estimate) * 0.75] dollars).
High Price Cost Transparency Screen:
• 6 days of guided tours to the Smithsonian, Ford’s Theater, Senate, Mount Vernon, major
monuments, & Newseum [Cost to the tour operator: (Participant estimate) * 0.17.]
• 6 nights of accommodations [Cost to the tour operator: (Participant estimate) * 0.50]
• 6 breakfasts, 2 lunches and 3 dinners [Cost to the tour operator: (Participant estimate)
* 0.17]
• Welcome and farewell receptions [Cost to the tour operator: (Participant estimate) *
0.08]
• Gratuities to porters, waiters, guides and drivers for all group activities [Cost to the tour
operator: (Participant estimate) * 0.08]
The total cost of the trip to the tour operator is [(Participant estimate ) * 1.00] dollars (and as
noted on the previous screen, the travel company is selling the trip for a price of [Participant
estimate) * 1.25] dollars).
Appendix 3
We test whether the effect of cost transparency holds for both familiar and unfamiliar
brands. Participants were simultaneously shown two t-shirts sold by two different brands for the
same price – J. Crew (an established firm) and Wuxtry (an imagined brand devised for the
study). Between-subjects, we varied transparency, and either J. Crew or Wuxtry disclosed the
(identical) total cost to produce the shirt. Subsequently, participants were asked from which
brand they would purchase the t-shirt.
Method
Design and Procedure. Participants (N = 419, Mage = 30.5, 55% male) completed this
study on Mturk in exchange for a small fixed payment. Participants first indicated their gender.
Then, they were shown two simulated retail website pages for two different retailers, J. Crew and
Wuxtry. Between-subjects we varied whether J. Crew had disclosed its costs, as well as whether
and why competitor Wuxtry had disclosed its costs. Specifically, in the J. Crew transparency
versus Wuxtry nontransparent condition, J. Crew’s page featured an infographic indicating that
the total cost of cotton, cutting, sewing, dyeing, finishing, and transport was $6.70. By contrast,
Wuxtry’s page contained no such information. In the Wuxtry voluntary versus J. Crew
nontransparent condition, Wuxtry’s page featured voluntary transparency, while J. Crew’s page
contained no such information. Between-subjects we also counterbalanced the ascription of
brands to website formatting, which produced four parallel conditions. The effects were invariant
to counterbalancing; therefore we collapse across this independent variable in the results.
Dependent Measures. Participants indicated from which firm they would prefer to buy
the t-shirt by responding to the item: “Given the opportunity, from which firm would you
purchase this t-shirt?” (Choices: Wuxtry and J. Crew).
Results. When J. Crew was the transparent brand (and Wuxtry was not transparent), 71%
of participants chose to buy from J. Crew. However, when Wuxtry was the transparent brand
(and J. Crew was not transparent), only 45% chose to buy from J. Crew, a significantly lower
percentage (71% against 45%, Χ2 = 26.87, p < 0.01).