This article was downloaded by: [68.181.16.

134] On: 05 April 2024, At: 16:36

Publisher: Institute for Operations Research and the Management Sciences (INFORMS)
INFORMS is located in Maryland, USA

Management Science
Publication details, including instructions for authors and subscription information:
http://pubsonline.informs.org

Agency or Wholesale? The Role of Retail Pass-Through

Honggang Hu, Quan Zheng, Xiajun Amy Pan

To cite this article:

Honggang Hu, Quan Zheng, Xiajun Amy Pan (2022) Agency or Wholesale? The Role of Retail Pass-Through. Management Science
68(10):7538-7554. https://doi.org/10.1287/mnsc.2021.4262

Full terms and conditions of use: https://pubsonline.informs.org/Publications/Librarians-Portal/PubsOnLine-Terms-and-

Conditions

This article may be used only for the purposes of research, teaching, and/or private study. Commercial use
or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher
approval, unless otherwise noted. For more information, contact permissions@informs.org.

The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness
for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or
support of claims made of that product, publication, or service.

Copyright © 2022, INFORMS

Please scroll down for article—it is on subsequent pages

With 12,500 members from nearly 90 countries, INFORMS is the largest international association of operations research (O.R.)
and analytics professionals and students. INFORMS provides unique networking and learning opportunities for individual
professionals, and organizations of all types and sizes, to better understand and use O.R. and analytics tools and methods to
transform strategic visions and achieve better outcomes.
For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
 MANAGEMENT SCIENCE
Vol. 68, No. 10, October 2022, pp. 7538–7554
https://pubsonline.informs.org/journal/mnsc ISSN 0025-1909 (print), ISSN 1526-5501 (online)

Agency or Wholesale? The Role of Retail Pass-Through

Honggang Hu,a Quan Zheng,b,* Xiajun Amy Panc
a
Advanced Institute of Business, Tongji University, Shanghai 200092, China; b International Institute of Finance, School of Management,
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.

University of Science and Technology of China, Hefei, Anhui 230026, China; c Warrington College of Business, University of Florida,
Gainesville, Florida 32611
*Corresponding author
Contact: honggang.hu@warrington.ufl.edu, https://orcid.org/0000-0002-3807-844X (HH); benzheng@ustc.edu.cn,
https://orcid.org/0000-0001-5760-8920 (QZ); amy.pan@warrington.ufl.edu, https://orcid.org/0000-0002-2944-6183 (XAP)

Received: December 10, 2020
Revised: May 22, 2021; August 22, 2021
Accepted: September 2, 2021
Published Online in Articles in Advance:
March 2, 2022
https://doi.org/10.1287/mnsc.2021.4262
Copyright: © 2022 INFORMS
Abstract. With the rapid growth of e-commerce, agency selling is currently gaining popu-
larity among online retailers (e-tailers). Prior research implicitly abstracts away cross-
brand pass-through under traditional wholesale selling (i.e., how the retail price of another
brand adjusts to changes in a given brand’s wholesale price) and suggests that a shift to
agency selling benefits e-tailers but harms suppliers. As an important counterweight to this
result, we discover that an e-tailer’s choice regarding selling format is critically moderated
by retail pass-through behavior. On the one hand, agency selling can improve channel effi-
ciency compared with wholesale selling. On the other hand, the relative intensity of supplier
competition between these two selling formats is ambiguous. We show that the existing result
applies only for a nonnegative cross-brand pass-through rate (e.g., under linear demands);
otherwise (e.g., under multiplicative or exponential demands), the opposite may hold. Inter-
estingly, we find that the conflict over the preference of selling format may not arise, and that
all channel members could be better off with wholesale selling. Compared with the case of
Bertrand competition, the e-tailer is even more likely to adopt wholesale selling under Cour-
not competition. Finally, under agency selling with an endogenous commission fee, we ad-
vise caution regarding the seemingly innocuous normalization of suppliers’ marginal costs.
Surprisingly, suppliers may benefit from higher marginal costs. Overall, our findings not
only shed light on the theory of agency selling but also provide a plausible explanation for
the observation that wholesale selling continues to prevail in online markets.

History: Accepted by Eric Anderson, marketing.

Funding: Q. Zheng is partially supported by the National Natural Science Foundation of China [Grants
72122020, 72102219, 71921001, and 71832011], USTC (University of Science and Technology of Chi-
na) Research Funds of the Double First-Class Initiative (YD2040002014), the Fundamental Research
Funds for the Central Universities (WK2040000021), and the Major Program of the National Natural
Science Foundation of China [Grants 72091210/72091215]. X. A. Pan is partially supported by a
Summer Research Grant awarded by the Warrington College of Business.

Keywords: platform • selling format • retail pass-through • cost • competition mode

1. Introduction commissions (a percentage of revenue from each sale).

Online retailing has steadily risen over the last few
years. The COVID-19 pandemic boosted e-commerce
further: In 2020, consumers spent $861.12 billion on-
line with U.S. merchants, and retail e-commerce sales
worldwide amounted to $4.28 trillion, which consti-
tute remarkable increases of 44.0% and 27.6%, respec-
tively, compared with the prior year.1 Currently,
products as diverse as e-books, apps, car insurance,
tickets, and hotel reservations are sold online. In on-
line markets, there are two major selling formats: (1)
traditional wholesale selling, where e-tailers purchase
products from suppliers and then resell them to con-
sumers, and (2) agency selling, an emerging practice
in which e-tailers allow suppliers to sell products to
consumers directly on their platforms in return for
It is important to recognize that commission fees are
not fixed and are often strategically determined by
e-tailers. For example, Amazon reassesses its referral
fees annually based on revenue and sellers’ responses.2
Prior research suggests that agency selling typically
promotes lower prices, benefits e-tailers, and hurts
suppliers (Johnson 2017). In practice, however, whole-
sale selling continues to prevail across different mar-
kets. In the consumer goods market, most electronics
retailers such as Best Buy adopt wholesale selling and
operate as resellers. Although Amazon’s online market-
place continues to grow, about 50% of its 2020 sales still
comes from the wholesale channel.3 Moreover, some
e-tailers such as Zappos.com have even completely
switched from agency selling to wholesale selling. In

7538
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS
the travel industry, online travel agencies (OTAs)
like Booking Holdings (owning brands such as Book-
ing.com and Priceline.com) and Expedia Group (own-
ing brands such as Expedia.com, Hotels.com, and
Hotwire.com) still profit from the wholesale selling for-
mat. By contrast, suppliers have now started to em-
brace agency selling. For example, some suppliers
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.
who are currently under wholesale contract with Ama-
zon are interested in switching to an agency contract.4
These observations motivated us to address the fol-
lowing questions in this research. First, would a
switch from wholesale selling to agency selling neces-
sarily improve an e-tailer’s profitability? Specifically, we
challenge the classic result in the literature that e-tailers
can always earn more profits under agency selling, and
we characterize fairly general conditions under which
wholesale selling should be adopted. Second, given that
e-tailers are in the position to choose contractual
agreements in online markets (Hagiu and Wright 2015,
Abhishek et al. 2016, Jerath and Zhang 2019), we further
investigate whether suppliers would embrace e-tailers’
contract choices, that is, the possibility of channel con-
flict. We are also interested in the impact of contract
types on welfare measures. Is consumer surplus always
higher under agency selling (e.g., Hao and Fan 2014)?
Which contract form is preferred by a social planner?
To capture essential features of modern online re-
tailing, we consider a common e-tailer channel where
multiple suppliers sell substitutable products to con-
sumers through a nonexclusive e-tailer. Our analysis
reveals two fundamental driving forces behind
e-tailers’ distinct choices of selling format: (vertical)
channel efficiency versus (horizontal) relative intensi-
ty of supplier competition. As is well documented in
the literature, agency selling is typically more efficient
than wholesale selling because of the revenue-sharing
mechanism (e.g., Johnson 2017). Nevertheless, because
suppliers can only keep part of their revenue but need
to bear the entire costs, agency selling still involves
the well-known double markup effect, which is driven
by suppliers’ positive marginal costs. When the costs
increase, we surprisingly find that the e-tailer may re-
duce the commission fee to moderate double margin-
alization, making suppliers with higher marginal costs
better off. This strategic interaction between suppliers’
marginal costs and agency commission has been over-
looked in previous studies that normalize the supply
costs to zero for simplicity, such as Tian et al. (2018)
and Jerath and Zhang (2019). We show that even in a
symmetric setting, normalizing suppliers’ marginal
costs to zero is not innocuous under agency selling.
More importantly, we show that the relative intensity
of supplier competition between the two selling formats,
which is essentially absent in previous literature, is am-
biguous. Specifically, the intensity of supplier competi-
tion under wholesale selling largely depends on the
7539
cross-brand retail pass-through behavior, defined as, the
extent to which the e-tailer adjusts the retail price of anoth-
er product in response to a given product’s wholesale
price changes. In contrast, previous studies (e.g., John-
son 2017, Tian et al. 2018, Jerath and Zhang 2019) focus on
the case of zero cross-brand pass-through.5 As a result,
we question the intuition that agency selling intensifies
competition among suppliers by making them face
each other directly without any buffer. We find, rather,
that with negative cross-brand pass-through, whole-
sale selling actually induces more severe supplier com-
petition and thus could be more appealing to the
e-tailer than agency selling.
When suppliers compete on prices rather than
quantities (Bertrand rather than Cournot, as studied
in most of the literature), our analysis shows that,
from the e-tailer’s perspective, agency selling domi-
nates wholesale selling only when the cross-brand
pass-through is nonnegative owing to a more intense
supplier competition in addition to a more efficient
channel under agency selling. However, the e-tailer’s
optimal choice of selling format hinges on the relative
importance of vertical (channel efficiency) and hori-
zontal (intensity of supplier competition) effects when
the cross-brand pass-through is negative. This is dem-
onstrated by two specific cases: multiplicative and ex-
ponential demands. Departing from the established
results in the literature (Johnson 2017), the competi-
tion effect dominates the channel efficiency effect
under multiplicative demand, and wholesale selling is
always more profitable for the e-tailer. In the exponen-
tial demand case, however, the competition effect is
dominant only when the product substitutability is rela-
tively high. Moreover, although a conflict of interest re-
garding the contract form arises in most cases between
the e-tailer and suppliers, we show that they would both
prefer wholesale selling when the degree of product sub-
stitution is in an intermediate range. This alignment is
further generalized to an impossibility result that they
would never prefer agency selling simultaneously.
As the relative intensity of supplier competition
plays a crucial role in the e-tailer’s choice of selling for-
mat, we also investigate the impact of competition
mode (i.e., Bertrand versus Cournot), a well-known de-
terminant of competition intensity in oligopoly theory.
Typically, Bertrand competition is appropriate under
production to order or in the case of unlimited capacity
(e.g., the digital goods market), whereas Cournot com-
petition is applicable under production in advance or
in the case of constrained capacity (e.g., the travel in-
dustry), even if firms are price-setters (Belleflamme
and Peitz 2010, pp. 66–67). Almost all previous studies
on agency selling assume that suppliers compete on
prices, but taking competition mode into account helps
us better understand the contract adoptions in different
industries. Compared with the Bertrand case, we find
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
7540
that the e-tailer is even more likely to adopt wholesale
selling in a Cournot market, owing to the reduction of
supplier competition under agency selling.
Finally, we compare consumer surplus and social
welfare between the two selling formats. As expected,
both society and consumers benefit from the selling
format that results in a lower market price, and a
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.
clear-cut price comparison result is available. Con-
trary to conventional wisdom, we show that agency
selling does not always lead to a higher consumer sur-
plus. Furthermore, we find that the e-tailer’s contract
choice is aligned with the social planner’s so that win-
win situations occur in most cases.
The remainder of this paper is organized as follows.
In Section 2, we discuss the relevant literature and high-
light our contributions in this regard. This is followed
by a description of our model and equilibrium analysis
in Sections 3 and 4, respectively. Section 5 conducts
profitability and welfare comparisons under agency
and wholesale selling. Finally, we conclude our work in
Section 6. All the proofs are relegated to Appendix B.
2. Related Literature
Our paper draws upon three broad streams of prior re-
search: distribution channels, strategic contract choices,
and retail pass-through behavior. Here, we provide an
overview of these research streams and highlight our
contributions.
Our work is related to abundant research on distribu-
tion channels in marketing and operations management.
Early studies focus on channel structures and efficiency.
McGuire and Staelin (1983) study the effect of product sub-
stitutability on the equilibrium distribution structures in a
setting of two suppliers selling through their respective ex-
clusive retailers. Choi (1991) considers the channel struc-
ture with one common retailer and multiple competing
manufacturers and compares equilibrium outcomes under
various game settings. Choi (1996), Lee and Staelin (1997),
and Trivedi (1998) further introduce competition at both
retail and manufacturer levels of a distribution channel.
Conventional wisdom in the literature suggests that
downstream retailers can soften suppliers’ competition
owing to the well-known double marginalization effect.
To eliminate channel inefficiency, a number of contrac-
tual instruments have been proposed, such as quantity
discounts (Jeuland and Shugan 1983), two-part tariffs
(Moorthy 1987), and revenue or profit sharing (Cachon
and Lariviere 2005). However, these channel coordina-
tion mechanisms are essentially built on top of the con-
ventional wholesale contract. In contrast, we examine
the channel efficiency of the agency contract, which is
becoming ubiquitous in online markets.
E-tailers’ strategic choices between wholesale and
agency selling has gained much attention in academia
recently. Prior research has explored this question from
Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS
the perspectives of marketing activities and information
asymmetry (Jiang et al. 2011, Hagiu and Wright 2015),
cross-channel effects and retail competition (Abhishek et al.
2016), digital goods (Hao and Fan 2014, Tan et al. 2016,
Tan and Carrillo 2017, Johnson 2020), and online prod-
uct reviews (Kwark et al. 2014, 2017). By comparison,
we investigate the fundamental distinctions between
the two contract forms in a more general setting. For
example, most prior studies assume that the consumer
market is linear for analytical tractability (Abhishek
et al. 2016, Tian et al. 2018, Jerath and Zhang 2019). Al-
though linear demand is commonly assumed in eco-
nomic models, demands tend to behave in a nonlinear
manner in practice, because consumer response is more
likely based on relative rather than absolute price differ-
ences. Therefore, we consider both linear and nonlinear
demand structures, extending the existing literature by il-
luminating the critical roles played by the underlying de-
mand structure and the corresponding (cross-brand) retail
pass-through in the e-tailer’s adoption of selling format.
Our analysis captures the fundamental features of agency
selling, which is not limited to online retailing. Those in-
sights are also expected to apply to the “stores within a
store” arrangement in the offline world (Jerath and Zhang
2010), where retailers rent out their retail space to suppli-
ers and give them complete autonomy on retail decisions,
such as pricing and in-store services.
Our paper is most closely related to Johnson (2017),
which proposes a general framework to compare the
two selling formats in a setting with imperfect compe-
tition at both layers of the supply chain, and shows
that retailers are typically better off, whereas suppliers
are worse off under agency selling. As an important
complement and advance, we find that taking into ac-
count the cross-brand pass-through behavior leads to
qualitatively different results.
Retail pass-through, the response of retailers to sup-
pliers’ price-off promotions, is a key element of re-
search on trade promotions in marketing; we refer the
reader to Narasimhan (2009) and Anderson and Fox
(2019) for a comprehensive review. Tyagi (1999) is
among the first to characterize the pass-through rate of
a single-product monopoly retailer, which depends on
the concavity of the demand function. Shugan and De-
siraju (2001) and Moorthy (2005) further generalize that
result by considering multiproduct retailers and retail
competition. They find that the cross-brand retail pass-
through effect can arise, and its direction can be either
positive or negative, depending on specific assumptions
imposed on demand and supply (e.g., functional forms of
demand and vertical strategic interaction). In a common
retailer channel, Sudhir (2001) reveals that the own-brand
pass-through is between zero and one, and the cross-
brand pass-through is nonpositive if the demand is logit
and multiplicative and the retailer maximizes category
profit in a Manufacturer-Stackelberg game.
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS 7541

Empirical studies have tried to estimate the rate and
determinants of pass-through (e.g., Meza and Sudhir
2006, Ailawadi and Harlam 2009, Nijs et al. 2010,
McShane et al. 2016). Consistent with theoretical re-
sults, Besanko et al. (2005) and Dubé and Gupta (2008)
find substantial empirical evidence of nonzero cross-
brand pass-through effects. Our paper complements
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.
The three assumptions in Assumption 1, respectively,
ensure that the demand of any product is downward
sloping in its own price, the products are substitutes,
and the total demand for all products is reduced as one
raises the price of any product (i.e., the own-price effect
dominates the cross-price effect). Following Shapiro
(1995) and Weyl and Fabinger (2013), we denote

the literature by showing that the (relative) intensity of 

supplier competition induced by the agency and
wholesale formats in a common e-tailer channel hinges
on cross-brand pass-through behavior.
3. Model Setup
Aside from the channel structure, our setup precisely
follows the framework of Johnson (2017), which enables
us to conduct an “apples-to-apples” comparison. Specif-
ically, we consider n ≥ 2 competing suppliers selling
substitutable products to consumers through a domi-
nant online retailer denoted by r. Each supplier has a
unique product. This is also the simplest setting to iso-
late the effect of cross-brand pass-through.6 Throughout
the paper, we use the pronoun “he” for any supplier
and “she” for the e-tailer. For ease of exposition, we as-
sume that suppliers incur the same marginal cost
cs ≥ 0,7 and the e-tailer also bears a marginal cost of cr ≥
0 for each unit of products demanded by consumers.
We first characterize the demand structure.
3.1. Demand Structure
We consider a general demand structure: the demand
of product i, i  1, 2, : : : , n
is given by a continuous and
differentiable function qi p , where p is the vector of
market prices. We make the following common as-
sumptions about the direct demand function as in the
literature (e.g., Lee and Staelin 1997, Dong et al. 2009).8
Assumption 1. The effect of pi on demand satisfies the fol-
lowing properties:
∂qi
< 0,
∂qj
≥ 0, j ≠ i and
n ∂qj
∂pi ∂pi j1 ∂pi < 0.


 2. The
demand
Assumption  function is symmetric:
q1 p  q2 p  : : :  qn p when p1  p2  : : : pn .
j≠i∂qj =∂pi
≡−
∂qi =∂pi
as the aggregate diversion ratio between products for
ease of notation. By Assumption 1, we have  ∈ [0, 1].
Importantly,  is invariant to retail prices in a symmet-
ric equilibrium for those commonly used aggregate de-
mand functions shown in Table 1.
Throughout the paper, we use the same γ > 0 to de-
note the product substitutability parameter for specific
demand functions: γ approaches zero when products are
completely independent, and it rises as products become
more substitutable. Besides structural properties, we will
also examine those specific demands for illustration.


At a symmetric  price p, we further denote q
p ≡
qi p, p, : : : , p for any supplier i, and therefore q p 
n


 

j1 ∂qj =∂pi < 0. Let λ p ≡ − q p =q p , which is a useful
measure of
the
 sensitivity of demand. In particular, the
sign of λ p determines the
log-curvature of demand
(i.e., the curvature of logq p ), because log-concavity,

log-convexity, and log-linearity are equivalent to λ p
being (globally) negative, positive, and zero, respec-
tively. As in Johnson
(2017), we make the following as-
sumptions about λ p :




Assumption 3. Both λ p and λ p 2 − λ p have slopes
strictly less than one.


4. The elasticity of λ(p) is no greater than
Assumption
one: pλ p =λ p ≤ 1.
Assumption 3 ensures that the Nash equilibria are
well behaved. In other words, profit functions are quasi-
concave. Assumption 4 is equivalent to the elasticity of

Table 1. Commonly Used Demand Specifications in the Literature


Demand model Functional form Sign λ p Sign ∂pj =∂wi

Spence-Dixit-Vives (SDV) μ [β+(n−2)γ]pi γ j≠i pj – 0
qi  β+(n−1)γ − + ,
(β−γ)[β+(n−1)γ] (β−γ)[β+(n−1)γ]
μ > 0, β > 0, 0 < γ < β
 n
Shubik-Levitan (SL) j1 pj – 0
qi  n1 μ − βpi − βγ pi − n ,
μ > 0, β > 0, γ > 0

Multiplicative qi  μpi
−β γ + –
j≠i pj , μ > 0, β > 1, 0 < γ < β
Constant Elasticity of Substitution (CES)
 1
1
γ−1
pi
+ 0
qi  γβ 1−γβ  1−β , 0 < γ, β < 1
γ 1−γβ
n γ−1
j1 pj

Exponential qi  e
μ−βpi +γ j≠i pj
, μ > 0, β > 0, 0 < γ < β 0 –
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
7542



demand pq p =q p being decreasing in p (i.e., Marshall’s
second law of demand). Assumptions 3 and 4 hold for all
the demand functions in Table 1.
3.2. Contractual Agreement and Game Structure
The two prevalent forms of contractual agreements in
practice are scrutinized in this paper: wholesale contract
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.
versus agency contract. The interactions between the
e-tailer and suppliers under the two types of contracts
are illustrated in Figure 1.
Under a wholesale agreement, the e-tailer acts as a
traditional reseller: Suppliers first announce their
wholesale prices wi in a noncooperative manner and
then the e-tailer sets the selling price pi for each prod-
uct.9 The game is solved by backward induction as fol-
lows. Given the wholesale price vector w, the e-tailer
decides the price p̂ i (w) to maximize the total profit:

n the outside option (i.e., the participation constraint is
max πr (p; w)  (pi − wi − cr ) · qi (pi , p−i ):
p≥0
i1
Anticipating the e-tailer’s best response p̂ i (w), each
supplier maximizes his profit by setting the wholesale
price independently and simultaneously:
max πi (wi , w−i )  (wi − cs ) · qi (p̂ i (wi , w−i ), p̂ −i (wi , w−i )),
wi ≥cs
i  1, 2, : : : , n:
Under an agency agreement, suppliers are given
the rights to sell products directly to consumers as
third-party sellers through the e-tailer’s online mar-
ketplace, but they need to share some revenue with
the e-tailer. Specifically, the e-tailer first sets the terms
of trade, namely, the revenue share 1 − α collected
from each supplier as a commission fee for access to
Figure 1. Schematic Diagrams of Market Structures
(a) Wholesale Selling
Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS
her customer base, where 0 ≤ α ≤ 1 is the portion kept
by each supplier.10 Upon accepting the offer, suppliers
then set retail prices p. By backward induction, given
α, each supplier sets the price p̂ i (α) to maximize his
profit independently and simultaneously:
max πi (pi , p−i ; α)  (αpi − cs ) · qi (pi , p−i ), i  1, 2, : : : , n:
pi ≥cs =α
Anticipating the suppliers’ best response p̂ i (α), the
e-tailer sets the trade term α to maximize the total profit:

n  

max πr (α)  (1 − α)p̂ i (α) − cr · qi p̂ i (α), p̂ −i (α) :
0≤α≤1
i1
Most prior works (e.g., Hao and Fan 2014, Kwark
et al. 2017, Tian et al. 2018) assume that α is exoge-
nously given or only set to guarantee that suppliers
will accept the offer from the e-tailer without seeking
binding). The reasoning behind this assumption is that
the e-tailer’s profit is always increasing in the commis-
sion fee (i.e., 1 − α), and she would extract all the trade
surplus if α is endogenized, stemming from the nor-
malization of marginal cost cs to zero in their studies.
Our work instead allows the e-tailer to strategically de-
termine the commission fee for any nonnegative cs. In-
terestingly, we identify a pitfall of the traditional cost
normalization under agency selling, as shown in Sec-
tion 5.1.
4. Equilibrium Analysis
To derive the equilibrium results in a uniform and
compact manner, we adopt the conduct-parameter
approach, which has been widely used in recent eco-
nomics literature (e.g., Weyl and Fabinger 2013,
(b) Agency Selling
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS 7543

Johnson 2017, Gaudin 2018, Miklós-Thal and Shaffer where θs ∈ [0, 1] is the conduct parameter of the
2021). In a symmetric setting, this approach allows the supplier layer. Through simple algebra manipulation,
equilibrium outcomes to be stated in terms of proper- we obtain
ties of the aggregate demand function. dp
(1 − ) dw
Under wholesale selling, the e-tailer sets the selling θs  ∂p ∂pj
,
∂wi −  ∂wi
i
prices p to maximize her total profit by solving the fol- 


lowing first-order conditions (FOCs): where dp=dw  nj1 ∂pj =∂wi  1= 1 − λ p according to

Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.

n ∂qj Equation (1).

qi + (pj − wj − cr )  0, i  1, 2, : : : , n,
j1 ∂p i Under agency selling, suppliers compete with each
other directly for end consumers given the commis-
given the wholesale prices w offered by suppliers.
sion rate α announced by the e-tailer. By Lerner’s rule
With symmetric

wholesale price w and retail price p, (e.g., Weyl and Fabinger 2013), the symmetric retail
we have q(p)+ p − w − cr q (p)  0, namely,
price p set by suppliers satisfies
p − w − cr  λ(p): (1)


 αp − cs  α(1 − )λ p : (4)
Denoting mr p as the e-tailer’s profit margin, then we
Because ms (p)  αp − cs under agency selling, we
have mr (p) p − w − cr under wholesale selling. As in
can see θs  1 −  from Equation (3). Then in the
Johnson (2017), we can rewrite the e-tailer’s profit for
e-tailer layer, we have the following FOC:
each product as mr (p)q(p) with derivative mr (p)q(p)+
mr (p)q (p). Then the optimum p satisfies dp
   dp

 p − (1 − α) λ p + (1 − α)p − cr  0: (5)
mr p
 dα dα

  θr λ p , (2) Based on suppliers’ best response (4), we have
mr p


 c s θs λ p
where θr ∈ [0, 1] represents the conduct parameter of mr p  (1 − α)p − cr  p − cs − cr −

the e-tailer layer. Generally, the conduct parameter θ p − θs λ p
is interpreted as a measure of competition intensity. It and

2
is equal to one when the market is monopolized or dp p − θs λ p
when firms collude, and approaches zero as competi- − 
 < 0:
dα cs 1 − θs λ p
tion becomes more intense. Following Johnson (2017)
and Gaudin (2018), we assume the following: Analogously, rewriting the FOC (5) as Equation (2), we
have θr  1 as expected. For clarity of exposition, we
Assumption 5. The conduct parameter θ is invariant in drop the subscripts (“r” and “s”) hereafter and refer to
the symmetric equilibrium.11 θ only as the conduct parameter of the supplier layer.
All the specific demand functions considered in Table - The detailed expressions of the conduct parameters for
1 satisfy this assumption. specific demand functions are summarized in Table
In line with Equation (1), we have θr  1 as the A.1. The equilibrium solutions under the two selling
e-tailer is in the monopolist position. Now consider formats are summarized in the following lemma.
the supplier layer. Anticipating p̂ i (w), the FOCs for
suppliers are given by Lemma 1. In the common e-tailer channel, the equilibri-
  um solutions are characterized as follows.
∂qi ∂pi  ∂qi ∂pj (i) Under wholesale selling, the equilibrium retail price pW
qi + (wi − cs ) +  0, i  1, 2, : : : , n:
∂pi ∂wi j≠i ∂pj ∂wi and wholesale price wW satisfy
  
By symmetry, we have pW − cs − cr  λ pW + θW λ pW 1 − λ pW ,

∂pi ∂pj
 
(w − cs ) −  (1 − )λ p , j ≠ i: wW  pW − c r − λ pW ,
∂wi ∂wi
Although the suppliers choose w, a useful
 alternative dp
(1−)dw
view is that they choose p as well. Let ms p denote the where θW  ∂pi ∂pj , j ≠ i;
∂wi
−∂w
suppliers’ profit margin.
Based  on the e-tailer’s best


i
(ii) Under agency selling, the equilibrium retail price pB
response (1), we have ms p  w − cs  p − cs − cr − λ p .
and commission rate αB satisfy
Then,


each
 supplier’s profit can

also
 be
written
 as  
ms p q p , with derivative ms p q p + ms p q (p). As   (1 − θB )λ
pB  + pB
1 − λ
pB 
p − cs − cr  λ p
B B
+ cs θ λ p
B B

B
2 ,
was done before Equation (2), the equilibrium value p − θB λ pA
of p is defined to implicitly solve
cs

 αB 
,
ms p
 pB − θB λ pB


  θs λ p , (3)
ms p where θB  1 − .
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
7544
Besides product substitutability and number of
competing varieties that are captured by , Lemma 1
demonstrates that the intensity of supplier competition
under wholesale selling (i.e., 1 − θW ) is also affected by
retail pass-through behavior. Prior works in marketing
(e.g., Besanko et al. 2005) define retail pass-through as
the rate at which changes in the wholesale price of a
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.
product are passed through by the retailer to the market
price, including both own-brand and cross-brand pass-
through. Although those commonly-used demand struc-
tures result in nonpositive cross-brand pass-through,
namely, ∂pj =∂wi ≤ 0 (see Table A.1), both empirical and
analytical studies (e.g., Besanko et al. 2005, Moorthy
2005) suggest that the cross-brand pass-through can be
positive, namely (i.e., ∂pj =∂wi > 0). Thus, for all the sce-
narios with different retail pass-through behaviors, we
theoretically characterize the structural properties and
provide managerial implications accordingly. The corre-
sponding equilibrium results for specific demands are
summarized in Tables A.4 and A.5.
Channel efficiency is well established in the litera-
ture as the first driving force behind firms’ preferences
between different selling formats (e.g., Johnson 2017).
That is, agency selling leads to a more efficient distri-
bution channel than wholesale selling, although the
well-known double marginalization problem remains.
We now compare the level of competition among sup-
pliers under wholesale and agency selling in the fol-
lowing proposition to reveal the second driving force,
which has been neglected previously. To facilitate ex-
position without affecting our qualitative results, we
hereafter normalize the e-tailer’s marginal cost to
zero, cr  0 (see Lemma 1).
Proposition 1. The relative intensity of supplier competi-
tion between wholesale and agency selling is contingent on
the sign of the e-tailer’s cross-brand pass-through (i.e.,
θW < θB if ∂pj =∂wi < 0, j ≠ i, and θW ≥ θB otherwise).
Although it is plausible that supplier competition is
abated by the moderating e-tailer under wholesale sell-
ing (Tian et al. 2018), Proposition 1 implies that the rela-
tive intensity of supplier competition between the two
contract forms hinges on the e-tailer’s cross-brand
pass-through behavior. Specifically, when the e-tailer
exercises a negative cross-brand pass-through, a reduc-
tion in the wholesale price of product i (e.g., trade pro-
motions) leads to a retail price increase for competing
products j, j ≠ i. As stated in Besanko et al. (2005), a
negative cross-brand pass-through may reflect the
e-tailer’s attempt to drive consumers from competing
brands to the “trade-promoted” brand. Intuitively, this
practice further intensifies competition among suppli-
ers under wholesale selling, because a relatively small
wholesale price disadvantage gets magnified as it is
passed through to the retail price charged to consumers.
Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS
Thus, the competition intensity in the supplier layer is
higher under wholesale selling than under agency sell-
ing (i.e., θW < θB ). Notably, this competition intensity is
the same across both selling formats (i.e., θW  θB ), if
and only if the cross-brand pass-through is zero.
Conversely, a positive cross-brand pass-through
implies that wholesale price reductions for product i
also reduce the retail prices of competing products j.
That is, the trade promotions offered by a single brand
lead to retail promotions on all the brands sold
through the e-tailer, which causes a “free-rider” prob-
lem. In practice, a substantial portion of suppliers’
marketing budgets is channeled through retailers in
the form of trade promotion spending (Anderson and
Fox 2019). Anticipating a diminished return from their
trade spending because of the e-tailer’s positive cross-
brand pass-through behavior, suppliers would be dis-
couraged from offering price-off promotions, thereby
lessening supplier competition under wholesale selling.
That is why suppliers compete more fiercely under
agency selling (i.e., θW > θB ).
5. Agency vs. Wholesale
We now proceed to examine channel members’ pref-
erences between the two selling formats. Let
λ(pB )−pB λ (pB )
KB  1−λα pB 1 + pB −θB λ pB .12
B
( ) ( )
Proposition 2. In the common e-tailer channel, the fol-
lowing hold:
(i) If the cross-brand pass-through is nonnegative (i.e.,
∂pj =∂wi ≥ 0, j ≠ i), the equilibrium retail price under agency
selling is no greater than that under wholesale selling (i.e.,
pB ≤ pW ). Consequently, a shift from wholesale selling to
agency selling raises the e-tailer’s profit (i.e., πBr ≥ πW
r ), but
lowers each supplier’s profit (i.e., πi ≤ πi );
B W
(ii) Otherwise (i.e., ∂pj =∂wi < 0), the equilibrium retail
price can be lower under wholesale selling. Specifically, pW <
pB if and only if θW < θB min{KB , 1}. As a result, wholesale
selling can be more profitable for the e-tailer, while agency
selling may benefit the suppliers.
Conceivably, when the cross-brand retail pass-through
is nonnegative (e.g., under linear or CES demand), agen-
cy selling induces a relatively more intense supplier com-
petition (see Proposition 1). Therefore, both effects of
supplier competition and channel efficiency work in the
same direction and lead to the e-tailer’s preference for
agency selling. On the contrary, suppliers still prefer to
be wholesalers, owing to the first-mover advantage and
reduced competition.
Our Proposition 2(i) resembles lemma 4 and proposi-
tion 6 in Johnson (2017), but we complement and ex-
tend the results of Johnson (2017) in two ways. First,
Johnson’s results hold only when the cross-brand pass-
through rate is nonnegative in our setting. At first
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS
glance, our common e-tailer channel seems to be a spe-
cial case of the previous analysis because the intensity
of competition represented by a conduct parameter can
vary from monopoly to perfect competition. As a matter
of fact, with symmetric-at-symmetric prices (Weyl and
Fabinger 2013), Johnson (2017) sidesteps the potential im-
pact of cross-brand pass-through by assuming θW  θB
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.
under bilateral imperfect competition.13 The common
e-tailer channel enables us to identify the pivotal role
of cross-brand pass-through, calling for caution when
applying symmetric-at-symmetric prices to the chan-
nel equilibrium analysis. Second, the sufficient condi-
tion of Propositions 3 and 6 in Johnson (2017) is
slightly stronger than ours. Assumption 4 admits
more log-convex demands, including but not limited
to the constant-elasticity demand.
A negative cross-brand pass-through implies that
suppliers compete more fiercely under wholesale sell-
ing than under agency selling (see Proposition 1). In-
terestingly, Proposition 2(ii) shows that the retail price
can be lower under wholesale selling when it brings a
sufficiently higher degree of supplier competition,
which offsets the stronger double marginalization ef-
fect compared with agency selling. Accordingly, the
e-tailer’s choice of contract form is now determined
by the relative dominance between the intensity of
supplier competition and channel efficiency. That is,
wholesale selling will be more appealing to the e-tailer
because of the intensified supplier competition, where-
as agency selling is better at mitigating the double
marginalization effect. In comparison, to avoid the
negative impact of intensified competition, suppliers
Figure 2. (Color online) Firms’ Contract Preferences Under Two Specific Demand Structures (n  2, μ  1, β  3)
Note. The (light blue) region at the bottom of (b) corresponds to the existing result in the literature (Johnson 2017).
7545
may be inclined to switch to agency selling, even at the
expense of the first-mover advantage they enjoy under
wholesale selling. Note that, in practice, this negative
cross-brand pass-through would be moderated by re-
tail competition.
To further illustrate how the horizontal and vertical
forces shape firms’ contract preferences, we now exam-
ine two specific demand models that entail negative
cross-brand pass-through: multiplicative and exponen-
tial demands (see Table A.2 for a summary).
Example 1. For the multiplicative demand, the equi-
librium retail price is always higher under agency
selling (i.e., pB > pW ) As is shown in Figure 2(a), a shift
from wholesale selling to agency selling lowers the
e-tailer’s profit (i.e., πBr < πWr ) but raises each suppli-
er’s profit (i.e., πBi > πWi ).
In contrast to the cases of linear and CES demands,
the horizontal effect of supplier competition always
dominates the vertical effect of channel efficiency in
a market with multiplicative demand.14 Similar to Prop-
osition 2, the e-tailer and suppliers also have unaligned
preferences over contracting forms, but now the e-tailer
prefers wholesale selling so that she can fully use her
pricing power to moderate the competition, whereas
suppliers prefer agency selling because the first-mover
advantage cannot offset the revenue loss caused by the
intensified competition. This finding is opposite to the
results in Johnson (2017) because that work does not ac-
count for the role of cross-brand pass-through.
Under the exponential demand structure, the hori-
zontal competition effect is not strong enough to always
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
7546 Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS

dominate the vertical efficiency effect. Indeed, whole-
sale price competition is less intense under exponen-
tial demand relative to multiplicative demand. That
is, the value of the conduct parameter θW associated
with exponential demand is larger (see Table A.1).
Therefore, the preferences of contract form vary with
the degree of product substitution as in Example 2.
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.
Proposition 3. Regardless of cross-brand retail pass-
through behavior, there is no possibility that both the e-tailer
and suppliers prefer agency selling simultaneously.
5.1. Impact of Suppliers’ Marginal Cost and
Endogenous Commission

In a common e-tailer channel with the SDV demand

Example 2. For the exponential demand, the equilibri-
um retail price is lower under wholesale selling than un-
der agency selling for highly substitutable products (i.e.,


pB < pW if γ ∈ 0, γ̂ , and pB ≥ pW otherwise). In terms of
profits, there also exist product substitutability thresh-


olds γ† and γ‡ γ‡ > γ† such that (see Figure 2(b)):
• When 0 < γ ≤ γ† , πBi < πW
i and πr ≥ πr ;
B W
† ‡
• When γ < γ < γ , πi < πi and πr < πW
B W B
r ;
• When γ ≥ γ‡ , πBi ≥ πWi and π B
r < π W
r .
Figure 2(b) provides a graphical illustration of the
impact of product substitutability γ on firms’ contract
preferences under exponential demand. Increased
substitutability further enhances the effect of supplier
competition. Example 2 states that when the degree of
product substitution is low (0 < γ ≤ γ† ), the vertical ef-
fect dominates the horizontal effect, and the e-tailer
prefers agency selling because of channel efficiency,
while suppliers profit more from leading the game
under wholesale selling. For highly substitutable
products (γ ≥ γ‡ ), the competition effect becomes
dominant. As a result, suppliers tend to switch to
agency selling to lessen competition whereas the
e-tailer benefits from the intensified supplier competi-
tion under wholesale selling. Surprisingly, when the
degree of product substitution is moderately large
(γ† < γ < γ‡ ), the preferences of the e-tailer and sup-
pliers for the selling format are aligned: they all desire
wholesale selling. In such a scenario, the intensity of
supplier competition under wholesale selling can off-
set the e-tailer’s concern about channel inefficiency,
but the effect is not strong enough for suppliers to
give up the first-mover advantage. We further gener-
alize this observation in the following proposition.
that reflects zero cross-brand pass-through, Jerath and
Zhang (2019) show that the e-tailer prefers agency
selling for less substitutable products but prefers
wholesale selling for more substitutable products.
This contrasts with our theoretical prediction that the
e-tailer would always prefer agency selling in such a
case (see Figure 3). In their setting, the suppliers’ mar-
ginal cost cs is normalized to zero at the outset for
simplicity (see Table 2). Although this normalization is
innocuous in studies on traditional wholesale selling, it
actually changes the nature of agency selling. We next
investigate the impact of this marginal cost on the
e-tailer’s contract selection and firms’ profits under agen-
cy selling.
By dividing both sides of Equation (4) by α, the sup-
pliers’ FOC under

agency selling can be written as
p − cs =α  (1 − )λ p . This is identical to the FOC of
vertically integrated suppliers with the marginal cost
cs =α. Therefore, studies that presume cs  0 would in-
evitably ignore the strategic effect of α on the equilibri-
um retail price (i.e., α is cancelled out from both
sides

of Equation (4)). When cs → 0+ , we have pB  λ pB in
equilibrium (see Lemma 1), which corresponds to the
scenario where a central decision maker set prices to

maximize the total profit ni1 pi qi (i.e., the first best so-
lution). This is why the equilibrium outcomes obtained
by setting cs → 0+ for a general equilibrium with any
cost cs do not converge to the ones under cs  0 nor-
malization. By not normalizing cs to zero, we can also
study how the commission and firms’ profits vary
with the supply cost, as Proposition 4 shows.
Figure 3. (Color online) E-tailer’s Optimal Profit Under the
SDV Demand (n  2, μ  1, β  1)

πr
Table 2. Equilibrium Results Under Agency Selling with
0.5
the SDV Demand (n  2)

Agency under Agency under 0.4

cs  0 cs → 0+

Conduct parameter θB 1− 1 0.3

μ
Retail price pB μ(β−γ)
2
2β−γ
0.2
Commission fee 1 − αB 1 1
μβ μ
Demand qB 2(β+γ)
(2β−γ)(β+γ) 0.1
Suppliers’ profit πBi 0 0
μ2
E-tailer’s profit πBr 2β(β−γ)μ2
2(β+γ)
(2β−γ) (β+γ)
2
0.2 0.4 0.6 0.8 1.0
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS
Proposition 4. Under agency selling, the optimal commis-
sion rate α is nondecreasing in cs (i.e., dαB =dcs ≥ 0). More
specifically, for log-concave and log-linear demands, α is
strictly decreasing in cs, but for log-convex demands, α can
be independent of cs (e.g., multiplicative and CES de-
mands15). In terms of profits, we have the following16:
• When dαB =dcs  0, each supplier’s profit is decreasing
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.
in cs (i.e., dπBi =dcs < 0). In contrast, when dαB =dcs > 0, an
increase in cs can benefit the suppliers with relatively low
marginal costs (i.e., there exists a threshold c†s > 0 such that
dπBi =dcs > 0 if cs < c†s );
• The e-tailer’s profit is decreasing in cs (i.e.,
dπBr =dcs < 0).
Indeed, in practice (e.g., Amazon, Walmart, eBay,
and Jet.com17), a low commission fee (< 10%) is usual-
ly charged for products with high marginal cost, such
as, personal computers and cell phone devices, where-
as a relatively high commission fee (> 15%) is charged
for low-cost products, such as books, clothing, and
shoes (see Figure 4(a)). Acknowledging that e-tailers
set referral fees taking multiple factors into consider-
ation, we may partially attribute this practice to
e-tailers’ strategic decisions on commission fees based
on suppliers’ marginal costs.
Interestingly, although the impact of increasing cs
on the e-tailer’s profit is always negative, each suppli-
er’s profit may vary nonmonotonically in the marginal
cost (see Figure 4(b)). To understand the intuition be-
hind this result, consider the SDV demand: When cs is
zero, the e-tailer extracts all the channel profits. As
this cost increases from zero, the e-tailer strategically
reduces the commission fee. It is exactly this strategic
effect of endogenous commission that enables suppli-
ers to benefit from a higher cost in the low range. When
the strategic effect is absent (i.e., ∂αB =∂cs  0), an in-
crease in cs always leads to a profit loss for suppliers,
because of the higher selling prices. In summary,
Figure 4. (Color online) Impact of the Supplier’s Marginal Cost on the Commission Fee and Suppliers’ Profits
(a) Commission Fees for Major Product Categories across
Online Retailers
7547
e-tailers have an incentive to help suppliers reduce
costs. For example, Taobao.com subsidizes sellers who
fulfill orders through Cainiao (Alibaba’s smart logistics
network). Nevertheless, these “benevolent” services
on cost reduction or other process innovations may
backfire for the suppliers, because a lower marginal
cost will prompt the e-tailer to raise her commission
fee.
5.2. Role of Competition Mode
To be compatible with prior literature, the analysis to
this point has focused on the case of Bertrand compe-
tition to identify the fundamental tension between
(horizontal) degree of supplier competition and (verti-
cal) channel efficiency. Because the competition mode
is also closely related to the horizontal effect, we now
turn our attention to a Cournot market. Under whole-
sale selling, deciding the prices p or the market clear-
ing quantities q makes no difference for the e-tailer
when she maximizes the category profit. Therefore,
we only distinguish different competition modes un-
der agency selling. We use superscripts B and C to de-
note Bertrand and Cournot competition, respectively.
The corresponding equilibrium solution under Cour-
not competition has the exact same form as under Ber-
trand competition (see Lemma 1), but θC  (n−1n−1+
)−(n−2)
.
Proposition 5. Cournot competition is more likely to in-
duce the e-tailer to adopt wholesale selling, whereas agency
selling would be more appealing to suppliers.
Consistent with conventional wisdom in oligopoly
theory, it can be shown that price competition among
suppliers gives rise to a more competitive outcome
compared with quantity competition under agency
selling (i.e., θB < θC ). Thus, given θW, the incidence of
θC > θW is larger than that of θB > θW , implying that
the competition effect is more likely to dominate. This
is also reflected by the e-tailer’s contract preference for
π iB
0.020 γ = 0.2
γ = 0.5
0.015 γ = 0.8
0.010
0.005
cs
0.2 0.4 0.6 0.8 1.0
(b) Supplier’s Optimal Profit under the SDV
Demand (n = 2, µ = β = 1)
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
7548
specific demands as summarized in Table A.2. For ex-
ample, under linear demands, agency selling is no lon-
ger the dominant choice for the e-tailer, as products are
more substitutable. Higher product substitutability fur-
ther amplifies the relative strength of the competition
effect, persuading the e-tailer to shift to wholesale sell-
ing. The channel efficiency advantage of agency selling
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.
is critical to the e-tailer only when the degree of
product substitution is low. From the perspective of sup-
pliers, mild competition becomes the key for them to pre-
fer agency selling in the Cournot competition case.
Our finding of the e-tailer’s inclination toward whole-
sale selling under Cournot competition may partially
explain the prevalent wholesale selling format in online
travel agencies, in which capacity-constrained service
providers (i.e., airlines, hotel chains and car rental com-
panies) compete in a Cournot manner. Expedia Group
and Booking Holdings are prominent examples. In
2019, Expedia Group made about 67% of its revenue
through wholesale selling (Expedia 2019). For Booking
Holdings, the revenue from wholesale selling ac-
counted for 28.2% in 2019, increasing from 25% in 2018,
whereas the revenue earned from agency selling de-
creased by 3.5% over the same period (Booking 2019).
5.3. Welfare Comparison
Having compared the firms’ profitability, we now
take the perspective of a social planner to examine
consumer surplus and social welfare differences be-
tween the alternate contract forms. Consumer surplus
is an economic measure that reflects the total differ-
ence between the maximum price that a consumer is
willing to pay and the actual selling price:
 p̄


CS  n q p dp,
pe


where pe is the equilibrium retail price and q p̄  0.
Accordingly, social welfare, which is the sum of con-
sumer surplus and channel profit, is



SW  n pe − cs q pe + CS:
It is obvious that consumer surplus is decreasing in
pe, and the following lemma shows that social welfare
has the same property.
Lemma 2. Social welfare SW is decreasing in pe.
Lemma 2 implies that a lower retail price simulta-
neously achieves higher consumer surplus and higher
social welfare. Therefore, the welfare analysis is fully
determined by the comparison between equilibrium
retail prices. The following proposition characterizes
the sufficient and necessary conditions for agency sell-
ing to be socially better than wholesale selling.
Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS
Proposition 6. Agency selling socially outperforms whole-
sale selling (i.e., CSA ≥ CSW and SW A ≥ SWW ) if and only
if θW ≥ θA min{KA , 1}, A ∈ {B, C}.
Proposition 6 indicates that the competition effect
has a significant impact on welfare implications. Con-
trary to conventional wisdom, agency selling does not
always generate more consumer surplus. Consumers
can be better off under wholesale selling when it indu-
ces a sufficiently high degree of supplier competition.
Table A.3 summarizes the socially superior contract
form for the commonly used demand structures. Inter-
estingly, we can observe that the contract form pre-
ferred by the e-tailer agrees with the social planner’s
preference in most cases, leading to a win-win situation.
Notably, under the exponential demand structure,
wholesale selling can even achieve a win-win-win goal
for suppliers, the e-tailer, and the social planner when
the degree of product substitution is intermediate.
6. Conclusion
The rise of agency selling is an intriguing phenomenon
observed worldwide in online markets. It is generally
believed that agency selling outperforms wholesale
selling for e-tailers because of channel efficiency. How-
ever, this view is at odds with the anecdotal evidence
that many e-tailers are still operating as resellers. To
deepen our understanding of online retail formats, this
paper develops a parsimonious model to reexamine
the “wholesale versus agency” debate and to establish
general conditions under which it is optimal for a non-
exclusive e-tailer to adopt wholesale selling. We also
investigate how the e-tailer’s contract choice affects
other stakeholders.
Our theoretical analysis highlights the importance
of accounting for both channel efficiency and supplier
competition when assessing the relative performance
of wholesale and agency selling. In contrast to previ-
ous studies, we discover that wholesale selling may
enable the e-tailer to enhance supplier competition by
exploiting a negative cross-brand pass-through, consti-
tuting a force toward improving the e-tailer’s profit.
Although agency selling is indeed more efficient than
wholesale selling, this vertical effect can be out-
weighed by the horizontal competition effect, and the
e-tailer would be more profitable under wholesale sell-
ing. More interestingly, the channel conflict between
suppliers and the e-tailer over contract form may not
arise under wholesale selling. This trade-off is espe-
cially relevant in a Cournot market, owing to further
reduced supplier competition under agency selling.
Remarkably, for agency selling with an endogenous
commission fee, we show that cost reduction on the
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS 7549

supply side does not necessarily benefit suppliers, and
we caution that assuming away suppliers’ marginal
costs may qualitatively obscure the insights. Finally, our
analysis indicates that the competition effect also plays
an important role in welfare implications.
Complementing the existing literature, this paper
provides an additional perspective for managers on
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.
We should caution that the presence of any con-
tract form we observe in practice may be a joint ef-
fect of multiple driving forces. Previous literature has
identified several factors that could also have an impor-
tant bearing on the choice of selling format, such as
demand spillover generated by marketing activities, infor-
mation asymmetry, and the cross-channel effect. They may
even become more first-order than cross-brand pass-

how to choose the appropriate contract form, guidance

that can be broadly applied to a number of markets.
Especially with the advent of Big Data and Artificial
Intelligence, the estimation of consumer demand has
been improved dramatically, and the cost of pricing
multiple products has been reduced significantly. Ma-
jor e-tailers increasingly set retail prices automatically
and dynamically through smart pricing technologies
taking all wholesale prices as inputs, facilitating cross-
brand pass-through. Such retailers will find our results
useful in deciding whether to persist with the whole-
sale price model or switch to the agency format. By the
same token, retailers who continue to use suboptimal
pricing heuristics such as cost-plus pricing (Lee 1987,
Dholakia 2018) will see little of value here.
through. For example, as Sussman (2019) points out, the
desire for tighter control over the brands sold on the plat-
form is the primary goal for Amazon. Meanwhile, cross-
brand pass-through behavior would be moderated by
retail competition (Moorthy 2005) or informational issues
(what price is best). It would be worthwhile for future re-
search to investigate the effect of cross-brand pass-through
on contract adoptions in more complex settings.
Acknowledgments
The authors are very grateful to the department editor,
Eric Anderson, the associate editor, and two anonymous
reviewers for constructive comments that resulted in a sig-
nificantly improved manuscript.

Appendix A. Tables
Table A.1. Cross-Brand Pass-Through and Conduct Parameters for Commonly Used Demand Functions

Demand function ∂pj =∂wi θW θB θC

SDV 0 β−γ β−γ β

β+(n−2)γ β+(n−2)γ β+(n−1)γ
n n n+γ
SL 0 n+(n−1)γ n+(n−1)γ n(1+γ)
Multiplicative γ(β−1+γ)(β−(n−1)γ) (β−(n−1)γ)M β−(n−1)γ β−(n−2)γ
− β β+γ
(β−1−(n−1)γ)M β3 −βγ2 +(n−1)(1−γ)γ2 −β2 (1+(n−1)γ)
CES 0 n(1−γ) n(1−γ) n+γ−(n−β)γ
n−γ−(n−1)γβ n−γ−(n−1)γβ n(1−γβ)
γ β−(2n−1)γ β−(n−1)γ β−(n−2)γ
Exponential − β−(2n−1)γ
β−γ β β+γ



Note. M  β2 − β 1 + 2(n − 1)γ + γ n − 1 − (2n − 1)γ .

Table A.2. Firms’ Preferences of Selling Format

Bertrand Cournot

Demand function Conditions Supplier e-Tailer Supplier e-Tailer

When γ is low W A W A
SDV
When γ is intermediate W W
SL
When γ is high A W
Multiplicative A W A W
CES Indifference A W
When γ is low W A W A
Exponential When γ is intermediate W W W W
When γ is high A W A W
Note. W represents wholesale selling and A represents agency selling.
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
7550 Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS




Table A.3. Social Planner’s Preference of Selling Format and λ p + θW λ p 1 − λ p has a slope less than one by
Demand function Conditions Bertrand Cournot
Assumption 3, and we know θW ∈ [0, 1], we must have
  
SDV When γ is low A pB − cs > λ pB + θW λ pB 1 − λ pB : (B.1)
A
SL When γ is high W
Multiplicative W W On the other hand, because the optimality condition for
agency selling is
 
B
B

B 
CES Indifference W
 
B (1 − θ )λ p + p 1 − λ p
B
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.

When γ is low A A p − cs  λ p + cs θ λ p
B B B

B
2 ,
Exponential
When γ is high W W p − θB λ pB
Notes. W represents wholesale selling and A represents agency a contradiction arises if
 
selling. The threshold of γ is the exact solution of the equation  
1 − θB λ
pB  + pB
1 − λ
pB 
θW  θA min{KA , 1}, A ∈ {B, C}. λ p + cs θ λ p
B B B

B
2
p − θB λ pB
  
≤ λ pB + θW λ pB 1 − λ pB : (B.2)
Appendix B. Proofs
Proof of Lemma 1. The equilibrium results directly fol- Using (and slightly rewriting) Equation (B.1) yields
low from the arguments in the main text. w   
cs < pB − λ pB − θW λ pB 1 − λ pB :
Proof of Proposition 1. By Lemma 1, we have
∂pj Thus, replacing cs from Equation (B.2) with the above in-
∂wi
θW − θB  (1 − )(n − 1 + ) ∂p ∂p
: equality and simplifying, we can obtain
i
−  ∂wji 


∂wi



  

∂pj
pB 1 − λ pB + θB λ pB λ pB − pB θB − θW + θB λ pB
2
Then, we can see that θ < θ if and only if
W B
∂wi < 0. w





Proof of Proposition 2. We divide the proof in to the 1 − θW λ pB pB λ pB − λ pB ≤ 0:
following two cases:
∂p
(i) From Proposition 1, we have θW ≥ θB when ∂wji ≥ 0, j ≠ i. Therefore, pB ≤ pW .
Suppose, for the sake of contradiction, that pB > pW . Because
 


the wholesale optimality condition is Let π̇ p  p − cs − λ p q p . Because π̇ is quasi-concave
   in p by Assumption 3 and the corresponding optimal solu-
pW − cs  λ pW + θW λ pW 1 − λ pW ,



tion ṗ satisfies ṗ − cs  λ ṗ + λ ṗ 1 − λ ṗ , we can easily

Table A.4. Equilibrium Results with Agency Selling Under Specific Demand Functions

Demand functions Commission rate αA Retail price pA

SDV 2cs 2 (β+(n−2)γ) cs (cs (β+(n−2)γ)+μ(n−1)γ) cs (β+(n−2)γ)+αB (β−γ)μ

αB  L(Q, R), Q  − μ2 (β−γ)
,R  μ2 (β−γ)
pB  αB (2β+(n−3)γ)
2cs 2 (β+(n−1)γ) cs (cs (β+(n−1)γ)+μ(n−1)γ) c(β+(n−1)γ)+αC μβ
αC  L(Q, R), Q  − μ2 β
,R  μ2 β
pC  αC (2β+(n−1)γ)
SL 2cs β (n+(n−1)γ)
2 2
cs β(cs β(n+(n−1)γ)+μ(n−1)γ) cs β(n+(n−1)γ)+nμαB
α  L(Q, R), Q 
B
− nμ2
,R  nμ2
p 
B
αB β(n(2+γ)−γ)
2cs 2 nβ2 (1+γ) cs β(cs nβ(1+γ)+μ(n−1)γ) cs nβ(1+γ)+αC (n+γ)μ
αC  L(Q, R), Q  − ,R  pC 
(n+γ)μ2 (n+γ)μ2 αC β(γ+n(2+γ))
Multiplicative α α 
B C
1 − β−(n−1
1 cs β(β−(n−1)γ)
)γ p 
B
(β−1)(β−1−(n−1)γ)
cs (β+γ)(β−(n−1)γ)
2

p  β−1−(n−1)γ β2 −β 1+(n−2)γ +γ n−2−(n−1)γ

C
( )( ( ) ( ))
CES αB  αC  βγ cs ((1−β)γ−n(1−γβ))
pB  γ2 β 1−β−n 1−γβ
( ( ))
pC  β n−1+β
cs n
( )γ2
Exponential cs (β−(n−1)γ)
αB  L(Q, R) − 3 ,
cs 2 (β−(n−1)γ)(7cs β2 +β(27−5cs (n−1)γ)+(n−1)γ(9−2cs (n−1)γ))
Q− 27 ,
cs β(cs nβ(1+γ)+μ(n−1)γ) pB  αcsB + 1β
R
(n+γ)μ2
cs (β−(n−1)γ)
α  L(Q, R) −
C
3 ,
cs 2 (β−(n−1)γ) (β(27+7cs β)+9(n+2)γ+cs (12−5n)βγ−cs (n−1)(5+2n)γ2 )
2

Q− 27(β−(n−2)γ)
,
cs (β−(n−1)γ)(3(n−1)γ+cs (β−(n−1)γ)(2β+(n+1)γ)) β−(n−2)γ
R pC  αcCs +
3(β−(n−2)γ) (β+γ)(β−(n−1)γ)

 
3
2 3 3 
2 3
Note. L(Q, R)  − Q2 + Q2 + R3 + − Q2 − Q2 + R3
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS 7551

Table A.5. Equilibrium Results with Wholesale Selling Under Specific Demand Functions

Demand functions Wholesale price wW Retail price pW

SDV cs (β+(n−2)γ)+(β−γ)μ cs (β+(n−2)γ)+(3β+(n−4)γ)μ

2β+(n−3)γ 4β+2(n−3)γ

SL cs β(n+(n−1)γ)+nμ cs β(n+(n−1)γ)+(n(3+γ)−γ)μ
β(n(2+γ)−γ) 2β(n(2+γ)−γ)

cs (β −βγ −(n−1)(1−γ)γ −β (1+(n−1)γ))
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.

3 2
Multiplicative 2 2
1 + β−1−(1n−1)γ wW
(β−1) β−(n−1)(β−1) γ+(n−β)γ2 +(n−1)γ3
2 2

CES cs ((1−β)γ−n(1−γβ)) cs ((1−β)γ−n(1−γβ))

γ(1−β−n(1−γβ)) γ2 β(1−β−n(1−γβ))

Exponential β− 2n−1)γ
cs + β−γ ( β−(n−1 (2+cs β)(β−nγ)+cs (n−1)γ2
( )( )γ) (β−γ)(β−(n−1)γ)





get pB ≤ pW ≤ ṗ because λ p + θW λ p 1 − λ p ≤ λ p + Using Lemma 1, Condition (B.3) can be rewritten as


 


λ p 1 − λ p , which implies   λ pB − pB λ pB
  λ p +α θ λ p 1+
B B B B


 
1 − θB λ
pB  + pB
1 − λ
pB    pB − θB λ pB
cs θB λ pB q pB
B  π̇ pB ≤ π̇ pW  πW
i :
B
B 2   
p −θ λ p
> λ pB + θW λ pB 1 − λ pB :
From Lemma 1, each supplier’s equilibrium profit under Thus, by simple algebraic manipulation, we can get pW <
agency selling can be written as pB if and only if


 


cs θB λ pB q pB αB λ pB − pB λ pB
πBi 
 , θ <θ
W B

 1+
 :
pB − θB λ pB 1 − λ p B pB − θB λ pB
so πBi ≤ πW
i because αB λ(pB )−pB λ(pB )
It is easy to check that 1+ > 0. The





 1−λ (pB ) pB −θB λ(pB )
cs θB λ pB q pB λ pB − pB λ pB
π̇ p − πi 
B B

2 ≥ 0: profit comparisons are illustrated by Examples 1 and 2. w

pB − θB λ pB

In terms of the e-tailer’s profit, from Lemma 1 and As-

Proof of Example 1. Based on the equilibrium results
sumption 4, we have under multiplicative demand in Tables A.4 and A.5, we




  cs θB λ pB − pB λ pB   have pW < pB because
πBr  nλ pB q pB 1 +
B
2
≥ nλ pB q pB :




p − θB λ pB pW β − 1 β3 − βγ2 + (n − 1) γ − 1 γ2 − β2 1 + (n − 1)γ




  < 1:
pB β β β − 1 2 − (n − 1) β − 1 2 γ + n − β γ2 + (n − 1)γ3




Let π̃ p  λ p q p , and we have





 In terms of profits, we have πW i < πi because
B

π̃  p λ p q p + λ p q p






 
  λ p 1 − λ p > 0: πW β β2 −β 1+2(n−1)γ +γ n−1+γ−2nγ
q p q p i



πBi β3 −βγ2 + (n−1) γ−1 γ2 −β2 1− (n−1)γ
⎛
2
2
 2 ⎞β−1−(n−1)γ



⎜⎜⎜β β−1 β− (n−1) β−1 γ+ n−β γ + (n−1)γ ⎟⎟⎟
3
Because q p < 0 (Assumption 1), π̃ p is always decreasing

 ⎜⎜⎜
⎟
in p (i.e., π̃  p < 0). From Lemma 1, the e-tailer’s equilibri- ⎝ β−1
β3 −βγ2 + (n−1)
γ−1γ2 −β2
1− (n−1)γ ⎟⎟⎠
um profit under wholesale selling can be written as
  <1,
πWr  nλ p
W
q pW :
and πW r > πr because
B

Therefore, given pB ≤ pW , we have πBr ≥ πW because






 r πW β3 −βγ2 −β2 1+ (n−1)γ −γ2 (n−1) 1−γ
nλ pB q pB  nπ̃ pB ≥ nπ̃ pW  πW . r



r

 πBr β3 −βγ2 + (n−1) γ−1 γ2 −β2 1− (n−1)γ
∂p
(ii) When ∂wji < 0, we have θW < θB . Because λ p + ⎞β−1−(n−1)γ


 ⎛
2
2
 2
θW λ p 1 − λ p has a slope less than one, pW < pB if and ⎜⎜⎜β β−1 β− (n−1) β−1 γ+ n−β γ + (n−1)γ ⎟⎟⎟
3
⎜⎜⎜
⎟
only if ⎝ β−1
β3 −βγ2 + (n−1)
γ−1γ2 −β2
1− (n−1)γ ⎟⎟⎠
  
pB − cs > λ pB + θW λ pB 1 − λ pB : (B.3) > 1: w
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
7552 Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS


Proof of Example 2. To focus on the interesting part of Because λ p < 1, Equation (B.4) has only one solution, de-
the analysis, the range of γ is further restricted (i.e., noted as p∗ , and it is easy to check that the second-order
β




γ < 2n−1) to ensure that the profit margins are positive in condition 2q p + p − cs q p < 0 is satisfied at p∗ . There-
equilibrium. Based on the equilibrium results under expo- fore, Π is unimodal and achieves the maximum at p∗ . In
nential demand in Tables A.4 and A.5, we let other words, Π is increasing in p when p ≤ p∗ and decreas-



  ing in p otherwise. On the other hand, we can obtain that

 pW αB β 2 + cs β β − nγ + cs (n − 1)γ2
Eγ¢ B 
B 

 : pB > pW > p∗ because
p α + cs β β − γ β − (n − 1)γ    
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.

 pW − cs  λ pW + θW λ pW 1 − λ pW > λ pW ,
Because we can observe that E γ is decreasing in γ,
αB (2+cs β)
 αB (n(2+cs β)−1)

 (1 − θB )λ
pB  + pB
1 − λ
pB 



β
E(0)  αB +cs β > 1 and E 2n−1  n αB +c β < 1 when p − cs  λ p + cs θ λ p
B B B B
> λ pB :
( s )

β
 (pB − θB λ pB )2
0 < γ < 2n−1, the equation E γ  1 only has one real root


β
between 0 and 2n−1 by continuity, which is denoted as γ̂. Therefore, we have Π pB < Π pW , which implies that the in-
terests of suppliers and the e-tailer cannot be aligned under
Thus, pB < pW if γ < γ̂ and pB ≥ pW otherwise.
Similarly, in terms of each supplier’s profit, we let agency selling (i.e., πBr ≥ πW
r and πi ≥ πi ).
B W

Taken together, our results follow. w

cs (1−αB )(β−(n−1)γ) nβγ+(n−1)γ2 −β2
+
 Proof of Proposition 4. Because the equilibrium retail

 πW e αB
β β − (2n − 1)γ
β(β−γ)

Es γ ¢ iB 

 : price pA , A ∈ {B, C} satisfies

πi αB β − γ β − (n − 1)γ
 

   (1 − θA )λ
pA  + pA
1 − λ
pA 
It can be checked that Es γ is decreasing in γ, pA − cs  λ pA + cs θA λ pA
 ,


cs β 1 −1 −1  (pA − θA λ pA )2
B β
Es (0)  e ααB > 1, and Es 2n−1  0. Thus, there exists a
 (B.5)
β
unique γ‡ ∈ 0, 2n−1 such that πW ‡
i > πi if γ < γ and πi ≤
B W
by differentiating both sides of Equation (B.5) with respect
πBi otherwise. to cs, we obtain
For the e-tailer’s profit, we let
dpA
cs (1−αB )(β−(n−1)γ) nβγ+(n−1)γ2 −β2
+

 πW e αB β
αB β(β−γ) dcs
Er γ ¢ rB 
B 
:
πr (1 − α ) α + cs β β − (n − 1)γ
B
A 2



p + θA λ pA λ pA − pA 1 + λ pA


 ⎡⎢⎢⎢ A

A 
A  A

A 

 ⎤⎥

2 ⎢⎢⎢
 cs θ λ p − p λ p
A
p 1 − λ p + λ pA 2 − θA − θA λ pA ⎥⎥⎥⎥⎥

 pA − θA λ pA ⎢⎢⎢⎢1 − λ pA + ⎥⎥⎥
It can also be checked that Er γ is increasing in γ, ⎢⎢⎢
3 ⎥⎥⎥
⎥⎥⎦
⎣ p A − θA λ p A


 cs nβ−ααB ((2n−1
B n−1+c nβ)
s
cs β 1 −1 −1
e αB αB β e ) αB (2n−1) > 0:
Er (0)  (1−αB ) αB +c β < 1, and Er 2n−1  n(1−αB ) αB +c > 1.
( s ) (  s β) Then, we have
β
As a result, there exists a unique γ† ∈ 0, 2n−1 such that dαA ∂αA ∂αA dpA

†  +
r > πr
if γ
πW B
> γ† and πW r ≤ πr otherwise. Because Er γ
B
 dcs ∂cs ∂pA dcs

1 and Es γ‡  1, it can be readily verified that 1


B 2
 pA − θA λ(pA )
α β − γ‡
Er (γ‡ ) 
  > 1: 
cs (1 − θA λ (pA )) (pA )2 + θA λ(pA )(λ(pA ) − pA (1 + λ (pA )))


(1 − αB ) αB + cs β β − (2n − 1)γ‡ −  
cs θA (λ(pA ) − pA λ (pA ))(pA (1 − λ (pA )) + λ(pA )(2 − θA − θA λ (pA )))
(pA − θA λ(pA ))4 1 − λ (pA ) +
(pA − θA λ(pA ))3
Therefore, γ‡ > γ† . w ≥ 0:
A
∂pj Furthermore, dα
 0 if and only if
Proof of Proposition 3. When ≥ 0, j ≠ i, Proposition 2
∂wi  λ
pA  dcs
 A
has shown that the e-tailer prefers agency selling, whereas λ p  A
p
suppliers prefer wholesale selling (i.e., πBr ≥ πW r and 

2 

2



4
∂p p A − θA λ p A p A − θA λ p A − 4c2s θA pA − 4cs θA pA pA − λ pA pA − θA λ pA + pA − θA λ pA
πBi < πW i , i  1, : : : , n). For the case of ∂wi < 0, we have the
j
+
2
2cs θA pA
following two scenarios:
(i) if pB ≤ pW , following the proof of Proposition 2(i), we > 0:
can see that the result is identical to the one for the case of A

∂pj dcs  0 can only occur when q p is log-convex. For
That is, dα
∂wi ≥ 0; example, consider the constant-elasticity demands (i.e.,





(ii) if pB > pW , we denote Π p  πr + nπi  n p − cs q p as q p  ap−b , a > 0, b > 1). Direct computation using Equation
the channel profit (i.e., the total profits of suppliers and the (B.5) gives
e-tailer). The FOC associated with maximizing Π is as follows: cs b2 cs b−1

 pA  and αA  A
A
 :
p − cs  λ p : (B.4) (b − 1)(b − θ )
A
p −θ λ p
A b
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS 7553

Now, consider each supplier’s profit. By differentiating which has the exact same form as under Bertrand competi-

A
A (n−2) 2
πAi α θ λ p q p
A A
with respect to cs, we get tion. Because θB − θC  1 −  − n−1−
n−1+  − n−1+ < 0, θW <
dπA   dαA    dpA θB always implies θW < θC . Therefore, when the competi-
i
 θA λ pA q pA + αA θA λ pA 1 − λ pA q pA : tion effect dominates the efficiency effect in a Bertrand mar-
dcs dcs dcs
ket, it suffices the same pattern in a Cournot market, which
dαA dπA dpA makes the e-tailer more likely to adopt wholesale selling
If dcs  0, it is obvious that i
dcs < 0 because dcs > 0; other-
while the suppliers are more inclined to agency selling. w
A dπA
dcs > 0),
wise (i.e., dα dcs |cs 0
> 0, because αA |cs 0  0. Then, ac-
i
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.

Proof of Lemma 2. As shown in the proof of Proposition

cording to the continuity argument, there must exits a
3, pe > p∗ , e ∈ {W, B, C} because
c†s > 0 such that πA †
i is increasing in cs if cs < cs .    
q(pA )+pA q (pA ) pW − cs − cr  λ pW + θW λ pW 1 − λ pW > λ pW ,
For the e-tailer’s profit, because  pA −

A
A

q (pA )


∂[p q(p )]
A A
pA − cs − cr  λ pA + cs θA λ pA
λ p > 0, we have A  A
 q p + p q p < 0. By differen-

 ∂p
A 



A


tiating πr  n 1 − α p q p with respect to cs, we obtain
A A A (1 − θA )λ pA + pA 1 − λ pA

A
A 2
> λ pA , A ∈ {B,C}:
 A   A
A  dpA (p − θ λ p )
A
dπA A dα A ∂ p q p
r
 n −p q p
A
+ 1−α :
dcs dcs ∂pA dcs
Therefore, Π is decreasing in pe. Because the social welfare
dαA dpA is the sum of consumer surplus and channel profit, the re-
Because dcs ≥ 0 and dcs > 0, πA
r is always decreasing in cs sult follows. w
dπA
(i.e., r
< 0). w
dcs Proof of Proposition 6. Because both the consumer
Proof of Proposition 5. We first derive the equilibrium surplus CS and social welfare SW are decreasing in pe, we
solution under agency selling in a Cournot market. Given only need to compare the equilibrium retail price across
the symmetric α determined by the e-tailer, the suppliers different contract forms. According to Proposition 2, we
set the same q based on the following FOC: have pB ≤ pW if and only if θW ≥ θB min{KB , 1}. Because the
αP − cs ∂Pi =∂qi equilibrium solutions under agency selling in both Ber-
 λ(P) n : (B.6) trand and Cournot markets are in the exact same form
α k1 ∂Pk =∂qi
(see Equation (B.5)), we can similarly obtain that pC ≤ pW
From Equation (3), the conduct parameter of the supplier
∂P =∂q
if and only if θW ≥ θC min{KC , 1}. w
layer is θCs  n i i . To facilitate comparison with
k1 ∂Pk =∂qi
the Bertrand case, we rewrite θCs in the form of a direct Endnotes
demand function. Differentiating both sides of qi  1
See https://www.digitalcommerce360.com/article/us-ecommerce-
qi (P), i  1, 2, : : : , n with respect to qj , j ≠ i, we have sales/ and https://www.emarketer.com/chart/242908/retail-ecommerce-
sales-worldwide-2019-2024-trillions-change-of-total-retail-sales, accessed
n
∂qi ∂Pk
 0, on April 26, 2021.
k1
∂Pk ∂qj 2
See https://www.sellersnap.io/referral-fees-us/, accessed on
∂qi ∂Pi  ∂qi ∂Pk ∂Pi   ∂Pk
 April 26, 2021.
⇒ + ⇒ −  0 symmetry ,
∂Pi ∂qj k≠i ∂Pk ∂qj ∂qj n − 1 k≠i ∂qj 3
See https://www.statista.com/statistics/259782/third-party-seller-
  share-of-amazon-platform/, accessed on April 26, 2021.
∂Pi  ∂Pj  ∂Pk ∂Pi  ∂Pj ∂Pi
⇒ − +  − + (n − 2)  0, 4
See https://etailsolutions.com/amazon-1p-to-3p-why-when-and-
∂qj n − 1 ∂qj k≠i, j ∂qj ∂qj n − 1 ∂qj ∂qj
how-to-make-the-move, accessed on April 26, 2021.

∂Pj n − 1 ∂Pi 5
We recently interviewed a Chief Operating Officer (COO) of a
⇒  2−n+ :
∂qj  ∂qj supplier of smart instant language translator devices selling
through e-tailers such as JD.com. He stated that e-tailers do adjust
Thus, θCs  ∂Pi =∂qi∂Pi =∂qi
+(n−1)∂Pj =∂qi 
n−1−(n−2)
n−1+ . Now, we consider the selling prices of his products accordingly when altering the pri-
the e-tailer layer. According to the suppliers’ best re- ces of those rival products as a result of rivals’ wholesale price
sponse (B.6), α  P−θcCsλ(P), which implies that optimizing α changes. This confirms the existence of e-tailers’ cross-brand pass-
s
is equivalent to optimizing q for the e-tailer. Then, the through behavior in practice.
6
e-tailer’s profit can be written as Allowing for imperfect competition at both layers is certainly im-
   portant, but it also sacrifices generality of results for tractability
cs
πr  n 1 − P − cr q: (e.g., restriction on retail pass-through), which is our focus.
P − θCs λ(P) 7
This marginal cost can be interpreted as the production costs for
physical goods, royalties for digital content (e.g., e-book, music,
With the FOC: dπr
dq  0 and Equation (2), we can easily de-
apps), and service fees in the travel industry (e.g., hotel reservation,
rive that θCr  1 and the equilibrium retail price pC satisfies air ticket, car rental).
⎡ 


⎤⎥
8
The corresponding assumptions for the inverse demand function
 ⎢⎢⎢⎢⎢ 1 − θCs λ pC + pC 1 − λ pC ⎥⎥⎥⎥⎥

Pi q can be similarly stated.
C ⎢ ⎢
⎢ ⎥⎥⎥,
p − cs − cr  λ p + cs θs λ p ⎢⎢
C C C

2 ⎥⎥⎥
⎢⎢ ⎥⎥⎦
9
This Supplier-Stackelberg move order not only captures the insti-
⎢⎣ pC − θC λ pC s
tutional reality but also enables us to directly contrast with the
 Hu, Zheng, and Pan: Agency or Wholesale? The Role of Retail Pass-Through
7554
extant literature. For completeness, we also investigate the alterna-
tive case where the e-tailer acts as a price leader, and the analysis is
available on request.
10
Consistent with common practice, we implicitly assume that the
e-tailer applies the same commission fee to all suppliers in a specific
category (i.e., discrimination is prohibited). Technically, this as-
sumption also does not change the symmetric equilibrium solution
in our common e-tailer setting.
Downloaded from informs.org by [68.181.16.134] on 05 April 2024, at 16:36 . For personal use only, all rights reserved.
11
For robustness, we also examine the Multinomial Logit (MNL)
model leading to variable conduct parameters, and it turns out that
our main insights still qualitatively hold.
12
Under Cournot competition, we define KC similarly.
13
See section 2.2 in Johnson (2017, p. 1166–1167) for the detailed
discussion about this assumption.
14
We have KB  1 for the multiplicative demand.
B
dcs  0 for all constant-elasticity de-
15
More generally, we have dα


mands in the form of q p  ap−b , a > 0, b > 1.
16
The results continue to hold when suppliers engage in a Cournot
competition; see the proof in Appendix B.
17
See https://www.zentail.com/blog/marketplace-commission-rates-
comparison-jet-com-walmart-amazon-ebay, accessed on April 26, 2021.
References
Abhishek V, Jerath K, Zhang ZJ (2016) Agency selling or reselling?
channel structures in electronic retailing. Management Sci. 62(8):
2259–2280.
Ailawadi KL, Harlam BA (2009) Findings-retailer promotion pass-
through: A measure, its magnitude, and its determinants.
Marketing Sci. 28(4):782–791.
Anderson ET, Fox EJ (2019) How price promotions work: A review
of practice and theory. Handbook of the Economics of Marketing,
vol. 1 (Elsevier, New York), 497–552.
Belleflamme P, Peitz M (2010) Industrial Organization: Markets and
Strategies (Cambridge University Press, Cambridge, UK).
Besanko D, Dubé J-P, Gupta S (2005) Own-brand and cross-brand
retail pass-through. Marketing Sci. 24(1):123–137.
Booking (2019) Booking Holdings’ 2019 annual report. Accessed
October 21, 2020, https://ir.bookingholdings.com/static-files/
92c3d5b6-8f42-4686-afc1-f6bd61b94e06.
Cachon GP, Lariviere MA (2005) Supply chain coordination with
revenue-sharing contracts: Strengths and limitations. Manage-
ment Sci. 51(1):30–44.
Choi SC (1991) Price competition in a channel structure with a com-
mon retailer. Marketing Sci. 10(4):271–296.
Choi SC (1996) Price competition in a duopoly common retailer
channel. J. Retailing 72(2):117–134.
Dholakia UM (2018) When cost-plus pricing is a good idea. Harvard
Bus. Rev. Digital articles. https://hbr.org/2018/07/when-cost-
plus-pricing-is-a-good-idea.
Dong L, Narasimhan C, Zhu K (2009) Product line pricing in a sup-
ply chain. Management Sci. 55(10):1704–1717.
Dubé J-P, Gupta S (2008) Cross-brand pass-through in supermarket
pricing. Marketing Sci. 27(3):324–333.
Expedia (2019) Expedia group’s 2019 annual report. Accessed
October 21, 2020, https://s26.q4cdn.com/987552466/files/doc_
financials/2019/ar/EXPEDIA-GROUP,-INC.-FINAL-2019-ANN
UAL-REPORT.pdf.
Gaudin G (2018) Vertical bargaining and retail competition: What
drives countervailing power? Econom. J. (London) 128(614):
2380–2413.
Hagiu A, Wright J (2015) Marketplace or reseller? Management Sci.
61(1):184–203.
Hao L, Fan M (2014) An analysis of pricing models in the electronic
book market. Management Inform. Systems Quart. 38(4):1017–1032.
Management Science, 2022, vol. 68, no. 10, pp. 7538–7554, © 2022 INFORMS
Jerath K, Zhang ZJ (2010) Store within a store. J. Marketing Res.
47(4):748–763.
Jerath K, Zhang ZJ (2019) Platform retailing: from offline “stores with-
in a store” to online “marketplaces.” Ingene CA, Brown JR, Dant
RP, eds. Handbook of Research on Distribution Channels (Ed-
ward Elgar Publishing, Cheltenham, UK), 52–74.
Jeuland AP, Shugan SM (1983) Managing channel profits. Marketing
Sci. 2(3):239–272.
Jiang B, Jerath K, Srinivasan K (2011) Firm strategies in the
“mid tail” of platform-based retailing. Marketing Sci. 30(5):
757–775.
Johnson JP (2017) The agency model and mfn clauses. Rev. Econom.
Stud. 84(3):1151–1185.
Johnson JP (2020) The agency and wholesale models in electronic
content markets. Internat. J. Industry Organ. 69:102581.
Kwark Y, Chen J, Raghunathan S (2014) Online product reviews:
Implications for retailers and competing manufacturers. Inform.
Systems Res. 25(1):93–110.
Kwark Y, Chen J, Raghunathan S (2017) Platform or wholesale? A stra-
tegic tool for online retailers to benefit from third-party informa-
tion. Management Inform. Systems Quart. 41(3):763–785.
Lee JY (1987) Developing a pricing system for a small business.
Management Accounting 68(9):50–58.
Lee E, Staelin R (1997) Vertical strategic interaction: Implications for
channel pricing strategy. Marketing Sci. 16(3):185–207.
McGuire TW, Staelin R (1983) An industry equilibrium analysis of
downstream vertical integration. Marketing Sci. 2(2):161–191.
McShane BB, Chen C, Anderson ET, Simester DI (2016) Decision
stages and asymmetries in regular retail price pass-through.
Marketing Sci. 35(4):619–639.
Meza S, Sudhir K (2006) Pass-through timing. Quantitative Marketing
Econom. 4(4):351–382.
Miklós-Thal J, Shaffer G (2021) Input price discrimination by resale
market. RAND J. Econom. 52(4):727–757.
Moorthy KS (1987) Managing channel profits: Comment. Marketing
Sci. 6(4):375–379.
Moorthy S (2005) A general theory of pass-through in channels
with category management and retail competition. Marketing
Sci. 24(1):110–122.
Narasimhan C (2009) Trade promotions. Rao VR, ed. Handbook of
Pricing Research in Marketing (Edward Elgar Publishing, Chel-
tenham, UK), 283–301.
Nijs V, Misra K, Anderson ET, Hansen K, Krishnamurthi L (2010)
Channel pass-through of trade promotions. Marketing Sci. 29(2):
250–267.
Shapiro C (1995) Mergers with differentiated products. Antitrust. 10:23.
Shugan SM, Desiraju R (2001) Retail product-line pricing strategy
when costs and products change. J. Retailing 77(1):17–38.
Sudhir K (2001) Structural analysis of manufacturer pricing in the
presence of a strategic retailer. Marketing Sci. 20(3):244–264.
Sussman S (2019) How Amazon controls its marketplace. Accessed
October 7, 2020, https://prospect.org/economy/amazon-controls-
marketplace/.
Tan Y, Carrillo JE (2017) Strategic analysis of the agency model for
digital goods. Production Oper. Management 26(4):724–741.
Tan Y, Carrillo JE, Cheng HK (2016) The agency model for digital
goods. Decision Sci. 47(4):628–660.
Tian L, Vakharia AJ, Tan Y, Xu Y (2018) Marketplace, reseller, or
hybrid: strategic analysis of an emerging e-commerce model.
Production Oper. Management 27(8):1595–1610.
Trivedi M (1998) Distribution channels: An extension of exclusive
retailership. Management Sci. 44(7):896–909.
Tyagi RK (1999) A characterization of retailer response to manufac-
turer trade deals. J. Marketing Res. 36(4):510–516.
Weyl EG, Fabinger M (2013) Pass-through as an economic tool: Prin-
ciples of incidence under imperfect competition. J. Political
Econom. 121(3):528–583.