European Journal of Operational Research 238 (2014) 221–232

Contents lists available at ScienceDirect

European Journal of Operational Research

journal homepage: www.elsevier.com/locate/ejor

Decision Support

Vertical integration with endogenous contract leadership: Stability

and fair profit allocation
Yuki Kumoi 1, Nobuo Matsubayashi ⇑
Department of Administration Engineering, Faculty of Science and Technology, Keio University, Hiyoshi, 3-14-1 Kohoku-ku, Yokohama 223-8522, Japan

a r t i c l e i n f o a b s t r a c t

Article history: This paper studies vertical integration in serial supply chains with a wholesale price contract. We con-
Received 29 August 2013 sider a business environment where the contracting leader may be endogenously changed before and
Accepted 12 March 2014 after forming the integration. A cooperative game is formulated to normatively analyze the stable and fair
Available online 25 March 2014
profit allocations under the grand coalition in such an environment. Our main result demonstrates that
vertical integration is stable when all members are pessimistic in the sense that they are sure that they
Keywords: will not become the contracting leader if they deviate from the grand coalition. We find that in this case,
Vertical integration
the grand coalition’s profit must be allocated more to the retailer and the members with higher costs.
Leader position
Cooperative game
Nevertheless, we also show the conditions under which the upstream manufacturer can have strong
Core allocation power as in traditional supply chains.
Economics Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction As another example, UNIQLO has established a ‘‘Specialty store

Modern supply chains in many industries—such as the well-
known examples in the automobile, electronics, and apparel indus-
tries—involve world-wide networks with many tiers, which are
filled by many companies playing various roles (Netessine, 2009).
In addition, in many cases, the power balance among the members
in a chain has drastically changed in the last two decades.
Specifically, we can increasingly find examples of retailers or other
downstream companies who have now become contract leaders in
supply chains, although formerly, manufacturers had the most
power and held contract leadership. In particular, the power of
retailers, including Wal-Mart, Tesco Plc, and UNIQLO, has been
increasing drastically. As a result, in consumer goods markets, pri-
vate labels by some major retailers have recently emerged wherein
the retailers design their labels and offer their manufacturers con-
tracts to produce them. This is in contrast with traditional national
brands where manufacturers design and produce products, and
then sell them widely through retailers. In fact, the share of private
labels has become about twenty percent in Wal-Mart, while in
Tesco Plc, which is the third biggest retailer in the world, it actually
exceeds forty percent.
⇑ Corresponding author. Tel.: +81 45 566 1812.
E-mail addresses: yuki.kumoi@gmail.com (Y. Kumoi), nobuo-m@pa2.so-net.
ne.jp (N. Matsubayashi).
1
Tel.: +81 45 563 1151; fax: +81 45 566 1617.
retailer of Private label Apparel’’ (SPA) model encompassing all
stages of a label’s business from design to production to sales,
although the manufacturing is outsourced. Further, at each stage,
UNIQLO plays a leading role, for example, by conducting stringent
quality control checks in outsourcing at the production stage. Fur-
thermore, Reebok, in the apparel industry, and Google, in the
smartphone industry, are examples of contract leaders in supply
chains who are neither retailers nor manufacturers. Reebok out-
sources its entire manufacturing to manufacturers it contracts to,
which are located in Southeast Asia and so on, and sells the product
through many retailers (van Dusen, 1998). Google, too, outsources
the manufacturing of its Nexus series of phones to leading manu-
facturers (preferring one manufacturer for each generation of
product: HTC for the first, Samsung for the next two, and LG for
the fourth and current), and sells them through cell phone carriers
or retail stores (Arrington, 2010). As seen from such examples, with
increasing tendency, many companies in mid-tier positions in sup-
ply chains are taking contract leaderships by utilizing their brand
power. Therefore, they focus on their brand development and man-
agement, rather than engaging in each stage of the chain, that is,
producing, retailing, and so on. All these examples show that re-
cently in supply chains, an entity in any position can attain con-
tract leadership and the sequential orders of contracts can vary
accordingly.
However, we should note that for any decentralized supply
chain, the problem of double marginalization exists, where each
firm sets a higher price in order to gain a higher profit margin. This

http://dx.doi.org/10.1016/j.ejor.2014.03.022
0377-2217/Ó 2014 Elsevier B.V. All rights reserved.
222 Y. Kumoi, N. Matsubayashi / European Journal of Operational Research 238 (2014) 221–232
then leads to a higher retail price, lower demand, and (thus) lower
profit (Spengler, 1950, Tirol 1988, and so on). Further, it is well
known that one of the solutions to this problem is vertical integra-
tion among firms. In fact, in the real world, one well known
example of vertical integration is that of Toyota, the Japanese auto-
mobile manufacturer, that formed the so-called ‘‘keiretsu,’’ a busi-
ness group linked in a supply relationship. In contrast, the SPA
model and the development of private labels mentioned above
can be seen as examples where having a strong retailer leads to
vertical integration of each process in a supply chain (design, pro-
duction, wholesale, retail, and so on) into a centralized system. In
addition, as is well known, DELL established the ‘‘direct model’’
in the 1990s, where one leader firm coordinates all stages of the
supply chain and ensures virtual integration into a centralized
system (Magretta, 1998).
The theoretical analysis on vertical integration in the frame-
work of game theory has been well documented. In addition, there
exist many researches on coordination schemes for attaining the
centralized solution under decentralization. However, most of
these papers analyze supply chains with only two tiers of suppliers
and retailers, and thus multi-tier supply chains, including those
with a middle man have received very little attention. In contrast,
with regard to the sequential orders of contracts under pre-cooper-
ation, most papers assume that a manufacturer is the contract
leader and contracts are sequentially taken from upstream to
downstream members. Thus, there are remarkably few studies that
assume a downstream firm’s potential to play a leadership role in
multi-tier supply chains. As an exception, Majumder and Sriniva-
san (2006) consider the possibility of varying sequential orders of
contracts. However, with regard to vertical integration, they focus
on the formation process of the integration and do not pay atten-
tion to the stability of the vertical integration formed or the
rational allocation of profit among the members. However, we
should note that the feasibility of forming the integration (grand
coalition) and the stability of the formed coalition are quite differ-
ent issues. It is very important to make the distinction between
these two issues, as we can find this to be one of the main areas
of consideration in game theory, especially cooperative game the-
ory. In fact, the following recent example of the e-book market in
the US shows the importance of considering the allocation of the
centralized system’s profit to each of its members in order to sus-
tain the system. Specifically speaking, Amazon, which monopolizes
the downstream platform in the e-book market, has set the price of
its e-books at 9.99 dollars, but one of the major publishers, Mac-
millan, is not satisfied with this price. Consequently, in January
2009, in negotiations with Amazon, Macmillan introduced a
requirement that they could set their own e-book prices from
12.99 dollars to 14.99 dollars. In addition, in the revenue-sharing
contract between them, Macmillan asked Amazon to increase the
rate of revenue sharing Macmillan receives from 35% to 70%.
However, these negotiations broke down, and as a result, Amazon
stopped selling Macmillan’s products. Two days later, Amazon
finally decided to accept Macmillan’s requirements. According to
some press reports, this is because Amazon recognized that
Macmillan actually enjoys monopoly in major e-book markets,
including school textbooks and so on (Rao, 2010; The New York
Times, 2010).
We can interpret this fact as the break-down of the coalition
because careful consideration was not paid to the possibility of
an ex-post change of position in the contracting leader, although
the parties previously succeeded in forming the grand coalition.
As recently in the e-book market, the power balance between play-
ers in supply chains can drastically change in the early days of an
industry. This implies that the leader position in supply chains may
endogenously change before and after vertical integration. There-
fore, under such environments, managers might be concerned
about how they can maintain their vertical integration. Based on
this motivation, by using a cooperative game theory, we explore
stable and fair profit allocation under vertical integration in light
of an ex-post change in the leadership position. We specifically ad-
dress the following research questions.
 In serial supply chains with n tiers, provided that the leader can
change (or all members believe that the leadership can change),
after forming the grand coalition (centralized system),
– When is the grand coalition stable?
– How should the grand coalition’s profit be allocated to
each firm?
 With such circumstances as described above, under which con-
ditions can the upstream manufacturer have the most power as
in traditional supply chains?
In order to answer these questions, we use cooperative game
theory that normatively discusses fair profit allocations under the
grand coalition. We formulate our problem as a coalitional form
game, the most standard model of a cooperative game. However,
we should note that formulation is a troublesome issue. In order
to formulate a coalitional form game, we must set the value
(characteristic function) of a coalition as the maximum value it
can individually earn (Myerson, 1996, and so on). This value is
well-defined for standard cases in the absence of any externality,
where any formation of a coalition does not affect the payoffs of
players outside the coalition. Unfortunately, however, an external-
ity exists under our setting, which implies that we must assume
how the outsiders respond to the formation of a deviating coalition.
Hence, here, we apply the theory of coalition formation, noting that
there have been remarkable recent developments relating to game
theory in this area. Specifically, we assume that even if some mem-
bers deviate from the grand coalition, the remaining members con-
tinue with cooperation, such that the value of a coalition is defined
as the equilibrium profit of the two-person game between deviat-
ing members and the remaining coalition. The formulation in this
manner is one of the key points of this study and plays an important
role in the discussions on our results mentioned later.
We now present an overview of the results obtained in this pa-
per. First, we show that in terms of the existence of core allocation,
the grand coalition is stable only if all members are pessimistic in
the sense that they are sure they will not become the contract lea-
der after any of their coalitional deviations from the grand coali-
tion. This result suggests that although the whole of the supply
chain benefits from vertical integration (which effectively elimi-
nates double marginalization), integration is not stable if all or
some members expect contract leadership. We next analyze the
Shapley value and the nucleolus of our model, which indicates fair
allocations under vertical integrations. As a result, we show that
the grand coalition’s profit must be allocated more to the retailer
and the members who have higher costs. In other words, these spe-
cific members have relatively strong powers for vertical integra-
tion to be stably sustained. Therefore, this result suggests that
under situations facing a variable leadership and rapid cost
improvements due to technological advances, the retailer is likely
to be in a much more beneficial situation under vertical integra-
tion, while the upstream manufacturer is not. However, we show
that even under such conditions, the upstream manufacturer can
have a relatively strong power if it can vertically integrate with
multiple monopolistic retailers.
The rest of this paper is organized as follows. Section 2 reviews
the relevant literature in vertical integration including that focus-
ing on coordination and cooperation in supply chains, and clarifies
the position of this paper. In Section 3, we formulate our problem
 Y. Kumoi, N. Matsubayashi / European Journal of Operational Research 238 (2014) 221–232
of vertical integration and profit allocation as a cooperative game.
In Section 4, we analyze the core, the Shapley value, and the
nucleolus of our game; and examine stable and fair profit alloca-
tions under vertical integration. We extend our analysis to a supply
chain with one manufacturer and multiple monopolistic retailers
in Section 5. Finally, Section 6 concludes the paper. The proofs of
all lemmas and theorems are provided in Appendix.
2. Literature review
The literature related to this study comes from three research
streams. The first is the literature on strategic interaction between
players in a supply chain where some downstream player such as a
retailer is the contract leader. The second is that on the formation
of vertical integration in a supply chain. The last is that on stability
and fair profit allocation in a coalition formation. All these works
span the subjects of economics, marketing, and operations
management.
In the first stream, Choi (1991) considers the strategic interac-
tion between two competing manufacturers and one monopolistic
retailer by using a game theoretic approach. He compares three
cases with each other: the case where one of the manufacturers
first offers its wholesale price as the contract leader (manufacturer
Stackelberg), the reverse case where the retailer is the contract lea-
der (retailer Stackelberg), and the case, where all players simulta-
neously determine their prices (vertical Nash). In his follow-on
work—Choi (1996)—he extends the setting to the case of two man-
ufacturers and two competing retailers and analyzes the equilib-
rium under various leaders’ positions. The recent work of
Edirisinghe, Bichescu, and Shi (2011) investigates the stable struc-
ture of contracting under possible scenarios in the model of Choi
(1991). These papers analyze only two-tier supply chains and as-
sume that no player incurs costs other than wholesale prices, while
each retailer determines its margin when it is the contract leader.
In contrast, Majumder and Srinivasan (2006) consider multi-tier
serial supply chains where each member has marginally increasing
costs and investigate the impact of the leader’s position on the
profit of each member and retailing price. In their following paper,
Majumder and Srinivasan (2008) explore the extension of their
model to network supply chains. Our model is based on that of
Majumder and Srinivasan (2006). However, as mentioned later,
they do not pay much attention to the topic of vertical integration,
which is our primary concern.
Next, regarding the second stream, we first should note that
Spengler (1950) is the pioneering paper that introduces the double
marginalization problem and vertical integration under wholesale
price contracts from a microeconomics viewpoint. Since then, in
the subjects of both marketing and operations management, many
papers study this topic by considering efficient distribution struc-
tures. For example, McGuire and Staelin (1983) analyze the equi-
librium distribution structure under a situation, where two
competing manufacturers sell their products exclusively through
two retailers. By a comparative static approach, they compare
the profitability between the centralized (vertical integration)
and decentralized cases, under varying substitutability between
the two products. Moorthy (1988) discusses the possibility of ver-
tical integration by investigating a model with the endogenous
choices of such dealership structures. Corbett and Karmarker
(2001) focus on entry and exit under competition among firms at
each stage in an n-tier supply chain, and on vertical integration.
In contrast, Jeuland and Shugan (1983) is the first study on channel
coordination between a manufacturer and a retailer in the decen-
tralized situation. Coughlan and Wernerfelt (1989) also discuss a
coordination scheme in multi-tier supply chains as in our study.
The basic topics of such coordination contracts for resolving the
double marginalization problem, including two-part tariff and
223
quantity discounts, are comprehensively explained in Tirol
(1988). Nevertheless, various contract schemes considering more
realistic situations have been recently developed. Among these,
the revenue sharing contract is studied in many papers, and is clo-
sely related with our study in terms of profit allocation. Under the
revenue sharing contract, a supplier has ownership of the product
and sets a retailing price, while a retailer deducts a percentage,
which is set by the contract, from the retailing price and remits
the balance to the supplier; for such contracts, see Cachon and Lari-
viere (2004). Gerchak and Yunzeng (2004) attempt an application
of this scheme to assembly systems and compare its profitability
with wholesale pricing. Pan, Lai, Leung, and Xiao (2010) also com-
pare the profitability of revenue sharing and wholesale pricing un-
der different channel power structures. Moreover, Li, Zhu, and
Huang (2009) analyze the Nash bargaining solution between a
manufacturer and a retailer based on the equilibrium attained un-
der a revenue sharing contract between them. They suggest the
possibility of attaining the centralized solution as in vertical inte-
gration through a long-term negotiation. However, these papers
are quite different from ours in the following two aspects. One is
that they deal with only two-tier supply chains, while we consider
supply chains consisting of n tiers. Second, the other difference is
that they focus on coordination mechanisms—feasibilities of the
centralized outcome under a decentralized situation. In contrast,
provided that the centralized outcome is already attained, we ex-
plore how the centralized system profit should be allocated to each
player in order to stably maintain the outcome.
In this sense, the literature regarding the third stream—the
stability of the coalition and fair payoff allocation among n
players—is much more important for our study. These topics are
closely related to the theory of coalition formation in the area of
game theory. The literature in this area is mainly classified into
two approaches. One concerns the non-cooperative game theoretic
approach, where stable coalition structures are explored, under
which each player individually decides whether or not to join a
coalition provided that players can form a coalition and communi-
cate with each other regarding payoff divisions. As in our study, the
second approach is a cooperative game theoretic approach, where
a stable and fair payoff allocation among players is normatively
analyzed, provided that a grand coalition is formed. Regarding
the former approach, Granot and Sosic (2003) and Sosic (2006)
examine the models of horizontal alliances between retailers for
efficient inventory management and mainly focus on the feasibility
of the stable formation of a grand coalition. While Granot and Sosic
(2003) assume myopic retailers and thus analyze a Nash equilib-
rium, Sosic (2006) employs the concept of farsighted stability
presented by Chwe (1994). However, both studies assume that
cooperation among members is sustained in the final stage of the
game, and thus, their non-cooperative game models partly involve
cooperative games. Granot and Sosic (2005), too, explore the
possibility of a stable alliance among three suppliers under an
Internet-based exchange, by using Chwe’s stable concept. In
addition, Nagarajan and Sosic (2007) examine the farsighted stabil-
ity of coalition formation under a price competition (Bertrand com-
petition) between n retailers. In their follow-on study—Nagarajan
and Sosic (2009)—they study the possibility of a horizontal alliance
among suppliers in an assembly system where one assembler and
n suppliers face price competition. Regarding the latter approach,
on the other hand, Meca, Timmer, Garcia-Jurado, and Borm
(2004) analyze a cost allocation problem arising from cooperative
inventory control by using a coalitional form game formulated on
a typical model of inventory management. Slikker, Fransoo, and
Wouters (2005) discuss fair profit allocation arising from inventory
sharing among retailers based on the newsvendor model.
Although the details of such related literature can be seen in good
surveys by Nagarajan and Sosic (2008), Meca and Timmer (2008),
224 Y. Kumoi, N. Matsubayashi / European Journal of Operational Research 238 (2014) 221–232
Fiestras-Janeiro, Garcia-Jurado, Meca, and Mosquera (2011), it
seems that their main focus is on cooperative inventory controls
or price settings among retailers, or information sharing between
the supplier and the retailer. In contrast, vertical cooperation for
eliminating double marginalization due to wholesale pricing has
received very little attention. To the best of our knowledge, the
only exception is Guardiola, Meca, and Timmer (2007), who con-
sider a core allocation of the grand coalition’s profit between one
manufacturer and n monopolistic retailers. Indeed, our study ex-
plores the same situation as theirs in Section 5. However, our mod-
el is different from theirs in that they assume that only the
manufacturer incurs a unit cost and that the manufacturer always
determines the wholesale price as the leader in decentralized situ-
ations. In addition, our main analysis focuses on multi-tier supply
chains. However, as mentioned in the previous section, this in-
volves a troublesome problem that since an externality exists be-
tween coalitions, some assumption is required for the actions of
players outside the deviating coalition in order to appropriately de-
fine the value of the coalition. Indeed, this type of problem is com-
mon to some topics including coalition formation under a
substitutable goods oligopoly with a provision of public goods.
Therefore, this problem has long been studied in game theory.
Examples include Aumann (1959), Chander and Tulkens (1997),
Chander (2007), Hart and Kurz (1983), and Rajan (1989), which
are referred to in detail in Section 3. Nevertheless, we must note
that none of these papers discuss vertical cooperation at all.
Finally, we reiterate the difference between this paper and
Majumder and Srinivasan (2006), which is the most closely related
work to our study. Regarding vertical integration, they mainly fo-
cus on coordination mechanisms such as a two-part tariff and a
formation process for the grand coalition with successive coopera-
tion between neighbors. In contrast, in this study, we attempt to
investigate stable and fair profit allocation under the grand coali-
tion in the framework of a cooperative game, while considering
an ex-post change in the leadership position. In this sense, our con-
tribution to the literature would be complementary to theirs.
3. Model
In this section, we formulate our model of vertical integration as
a cooperative game. Specifically, in 3.1, we first introduce our base
model of leader–follower wholesale-price contracts before any
cooperation. Then, in 3.2, we consider coalitions between any
members and formulate the coalitional form game where the con-
tract leader is variable.
Following Majumder and Srinivasan (2006), our model is based
on a simple wholesale-price contract under an n-stage serial sup-
ply chain. The most upstream member is the manufacturer who
faces an increasing marginal cost, while the most downstream
member is the retailer who faces a price-elastic demand in addi-
tion to an increasing marginal cost. Any middle members are
assemblers or wholesalers who also face increasing marginal costs.
We suppose the sequential leader–follower contract among them.
Specifically, at each stage, the follower member chooses the quan-
tity it sells, in response to the wholesale price offered by the leader.
For example, if the retailer offers its wholesale price as the contract
leader, then his immediate upstream neighbor chooses the quan-
tity it sells. Conversely, if this upstream neighbor offers a whole-
sale price to the retailer as the leader, then the retailer chooses
the quantity it sells, in consideration of the retail price. In what fol-
lows, we formally describe these processes.
3.1. Model of pre-cooperation
We consider a serial supply chain with n members
N ¼ f1; 2; . . . ; ng. The manufacturer is indexed as member 1, the
retailer as member n, and assemblers or wholesalers as members
2; 3; . . . ; n  2; n  1. Let us assume that there is only one contract
leader in the supply chain, who is labeled as l. This leader first of-
fers the contract to its downstream neighbor l þ 1 and then mem-
ber l þ 1 offers the next contract to its neighbor l þ 2, and so on.
Finally, the retailer determines its selling quantity. In a similar
way, the leader also offers the contract to its upstream neighbor
l  1 and then member l  1 offers the contract to its neighbor
l  2, and so on. Finally, the manufacturer determines the quantity
it produces. We note that the direction of contracts from the leader
is always one way, and thus the contract sequence is uniquely
determined once the leader position is given. Given leader l, we de-
fine the wholesale price offered to member i as wli ði – lÞ. If i > l, this
price is offered by the upstream member i  1, while if i < l, it is of-
fered by the downstream member i þ 1. If member i chooses its
selling quantity qli , then the total payment is equal to wli qli , which
is paid from member i to member i  1 under i > l and from i þ 1
to i under i < l. Further, let C i ðqli Þ be the total cost of each member
i with regard to its process (producing, assembling, retailing, etc.)
and PðqÞ be the (inverse) demand function of the final product. Un-
der these structures, each member non-cooperatively attempts to
maximize its profit.
In this paper, following Majumder and Srinivasan (2006), the
cost and demand functions are specifically given as follows:
2
C i ðqli Þ ¼ ki ðqli Þ ; ki P 0; for all i 2 N;
PðqÞ ¼ 1  q:
The assumption of increasing marginal costs is relatively appro-
priate when congestion and queuing delays are expected at each
stage in a supply chain due to the limits of capacity or process
capability, as Majumder and Srinivasan (2006) explain in detail
by using real-world examples.
3.1.1. Profit maximization problem for members located downstream
of the leader
We first consider supply chains with an interior leader, that is,
l – 1; n (we deal with the case where member 1 or member n is the
leader in Section 3.1.3). First, let us consider the problem of down-
stream members. Each member other than the retailer (l < i < n)
determines its selling quantity qli and wholesale price to its imme-
diate downstream member wliþ1 in order to maximize its own prof-
it, subject to the wholesale price wli . Formally, we have
max pli ðwliþ1 ; qli ; wli Þ ¼ wliþ1 minðqli ; qliþ1 ðwliþ1 ÞÞ  wli qli  C i ðqli Þ: ð1Þ
wliþ1 ;qli
Note that qliþ1 is a function of wliþ1 , the wholesale price which i
offers to i þ 1. Taking the minimum in the first term in Eq. (1) im-
plies that there may be a difference between qli and qliþ1 , since
member i þ 1 can choose its purchasing quantity as long as it does
not exceed the quantity sold, qli . However, Lemma 1 of Majumder
and Srinivasan (2006) indeed ensures that in the equilibrium, both
quantities are always equivalent (there is no leftover product).
Hence, the profit maximization problem of member i can be re-
duced to the following problem choosing only wliþ1 : (This result
also holds for other cases as we will explain later; as such, we de-
scribe only the reduced problems, unless otherwise noted.)
maxpli ¼ wliþ1 qliþ1 ðwliþ1 Þ  wli qliþ1 ðwliþ1 Þ  C i ðqliþ1 ðwliþ1 ÞÞ:
wliþ1
At the end of the process, the retailer determines the selling
quantity qln of the final product that maximizes its own profit func-
tion, subject to the wholesale price wln :
maxpln ðqln ; wln Þ ¼ Pðqln Þqln  wln qln  C n ðqln Þ: ð2Þ
qln
 Y. Kumoi, N. Matsubayashi / European Journal of Operational Research 238 (2014) 221–232
3.1.2. Profit maximization problem for members located upstream of
the leader
Let us consider the members located upstream of the leader
other than the manufacturer, i.e., any member i such that
1 < i < l. Given the wholesale price wli , each i determines the
wholesale price for the next upstream member wli1 in order to
maximize the following profit function:
maxpli ðwli1 ; wli Þ ¼ wli qli1 ðwli1 Þ  wli1 qli1 ðwli1 Þ  C i ðqli1 ðwli1 ÞÞ:
wli1
Note also that qli1
is a function of wli1 ,
the wholesale price that i
offers to i  1. At the end of the process, the manufacturer deter-
mines the quantity ql1 that maximizes the following profit function,
subject to the wholesale price wl1 :
maxpl1 ðql1 ; wl1 Þ ¼ wl1 ql1  C 1 ðql1 Þ:
ql1
3.1.3. Profit maximization problem for the leader
We first consider the case where the leader is the manufacturer.
We note that the downstream members including the retailer have
their profit functions given by (1) and (2). Then, since there is no
member upstream, the manufacturer has to determine the whole-
sale price w12 that maximizes the following profit function:
maxp11 ðw12 Þ ¼ w12 q12 ðw12 Þ  C 1 ðq12 ðw12 ÞÞ:
w12
Note that q12 is a function of w12 . Similarly, for the case where the lea-
der is the retailer, the upstream members including the manufac-
turer have their profit functions given by (3) and (4). In addition,
since there is no member downstream, the retailer has to determine
the wholesale price wnn1 that maximizes the following profit
function:
max
n
pnn ðwnn1 Þ ¼ Pðqnn1 ðwnn1 ÞÞqnn1 ðwnn1 Þ  wnn1 qnn1 ðwnn1 Þ
wn1
 C n ðqnn1 ðwnn1 ÞÞ:
Note also that qnn1 is a function of wnn1 . In addition, we should note
that as explained above, in the equilibrium, the selling quantity of
the retailer is equal to the quantity sold by its immediate upstream
neighbor n  1, which implies that the wholesale price offered to
n  1 is the only decision variable for the retailer.
Finally, for any interior leader, the decision variables are the
two wholesale prices wll1 and wllþ1 , which are offered to its imme-
diate upstream and downstream neighbors, respectively. However,
again from the above discussion, the purchased quantity is always
equivalent to the sold quantity in the equilibrium, which ensures
that the leader’s problem is reduced to the following one:
max pll ðwll1 ; wllþ1 Þ ¼ wllþ1 qllþ1 ðwllþ1 Þ  wll1 qll1 ðwll1 Þ  C l ðqll1 ðwll1 ÞÞ;
wll1 ;wllþ1
subject to qllþ1 ðwllþ1 Þ ¼ qll1 ðwll1 Þ:
Note that qll1 and qllþ1 are functions of wll1 and wllþ1 , respectively.
3.1.4. Game in extensive form
We can formulate the game with leader l as the following
extensive form game. We start from the decision of l in the first
stage, followed by two branches for members l  1 and l þ 1 that
are based on the wholesale prices offered by l. In each branch,
we have a sequence of stages, with each stage consisting of the
next member’s problem: one branch is for the members upstream
of the leader, which leads to the manufacturer, and the other is for
the members downstream of the leader, which leads to the retailer.
The equilibrium concept we adopt is the subgame perfect
225
equilibrium. As the number of members is finite, we can find the
equilibrium of this game by using backward induction.
3.2. Coalitional form game
We consider cooperation by any members on the basis of the
non-cooperative game formulated in the previous section. Specifi-
cally, while the supply chain system also employs a leader–fol-
lower contract framework, the members included in the coalition
determine their wholesale prices and selling quantities at each
ð3Þ
stage to maximize their total profit. Majumder and Srinivasan
(2006, Section 5.3) also discuss such a situation and prove that
cooperation between the leader and either of its neighbors is al-
ways profitable for them, and thus by repeating such a process,
the grand coalition where all members jointly form one coalition,
can beneficially be formed. Majumder and Srinivasan (2006, Sec-
ð4Þ tions 5.1 and 5.2) also show other formation processes of coalitions
that lead to the grand coalition. These results ensure the feasibility
of vertical integration among all members in the system. Thus,
acknowledging this, we here focus on analyzing the stability of
the grand coalition after realizing it and on fair profit allocation
among the members under the coalition. To do this, we formulate
this problem as a coalitional form game (characteristic function
game), which is known as the most standard model of a coopera-
tive game, and investigate the core, the Shapley value, and the
nucleolus of the game.
In order to formulate a coalitional form game, we must set the
value of coalition (characteristic function) v ðSÞ for each coalition
S  N in an appropriate way. However, we should note that in this
regard, there are two major issues. One of these is that we must as-
sume how each member in N  S responds to the formation of S,
since the profit S it can earn is clearly dependent on the decisions
of the members in N  S. The other problem is that under the situ-
ation where the whole system employs the leader–follower frame-
work, we must also assume the order of contracts between S and
each member in N  S, regardless of how the members in N  S
cooperate. Regarding the former, we should note that this type of
problem has long been discussed in microeconomics and game
theory, in particular in the area of horizontal mergers under oligop-
oly and coalition formation in public good provision, where play-
ers’ strategic interactions exhibit some externalities. Therefore,
the appropriate value of the coalition (characteristic function)
applicable to such situations has been developed. In particular, a-
and b-characteristic functions are perhaps the best known, where
coalition S assumes that the outside players (members in N  S)
will act so as to minimize the payoff of S (Aumann, 1959; Myerson,
1996, and so on). However, while such minimax-type actions
might be justified under a zero-sum situation, they are obviously
inappropriate under a non zero-sum situation where there is an
externality (Chander, 2007). (Nevertheless, a- and b-characteristic
functions can be justified in some models of a horizontal merger;
see for example, Zhao (1999) and Norde, Pham Do, & Tijs (2002).)
However, in fact, in our model, the worst scenario for S about the
action of N  S is that at each stage, each member in N  S offers
the wholesale price equivalent to the retail price, which is so high
as to exclude the demand of the final product. However, clearly
each member in N  S also does not benefit from this action. This
implies that it is very important to take into account strategic
interactions between a deviating coalition and outside members.
Noting this, other types of characteristic functions have recently
been developed where v ðSÞ is defined as the equilibrium payoff
of a game between S and every member in N  S. Among these
functions, there are two typically characteristic ones. One is known
as the c-characteristic function coalition (Chander & Tulkens,
1997; Chander, 2007; Hart & Kurz, 1983; Rajan, 1989, and so on),
where it is assumed that for the deviation of S from the grand
226 Y. Kumoi, N. Matsubayashi / European Journal of Operational Research 238 (2014) 221–232
coalition, the members in N  S break up into singletons and each
member reacts individually (non-cooperatively) to maximize its
own profit. That is, the value of a coalition can be obtained as
the payoff earned at the equilibrium of the jN  Sj þ 1-person game
between S and the members in N  S. The other function is known
as ‘‘model d,’’ which was introduced by Hart and Kurz (1983),
where it is assumed that given a deviation of S, the members in
N  S keep the coalition. Therefore, the value of a coalition is de-
fined as the payoff earned at the equilibrium of the two-person
game between S and N  S.
Of course, we must choose the most appropriate characteristic
function depending on the context. For example, when a horizontal
merger under Cournot competition or a coalition in public good
provision is modeled, the c-characteristic function is commonly
utilized. This is because the assumption in which each member
of N  S acts independently can be justified, since doing so contrib-
utes to avoiding the free-riding by S arising from such competition,
such that each member in N  S can earn a higher profit. However,
this assumption unfortunately lacks validity under our situation
for the following two reasons. First, as discussed in Majumder
and Srinivasan (2006, Section 6), under a vertical supply chain,
each member in N  S usually benefits from continuing to cooper-
ate. The second reason is that as mentioned above, our leader–fol-
lower framework requires that we assume the order of contracts
among members, regardless of how members in N  S cooperate.
However, this implies that if we consider the case, where each
member of N  S individually acts and the leader might change
depending on the coalition structure, we also have to appropriately
assume who becomes the new contract leader. This seems to be a
rather troublesome problem as the outcome v ðSÞ can be drastically
different depending on the assumptions. In addition, in order to
realize such a situation, we might need some process where start-
ing from the deviation of S, each member in N  S sequentially de-
parts from the grand coalition, and finally, they coordinate with the
new leader. However, as the number of members increases, this
process takes a long time. As mentioned in the introduction sec-
tion, this is not consistent with our motivation for decision-making
under rapid changes in the business environment.
Based on these points, we adopt the formulation of ‘‘model d’’
by Hart and Kurz (1983) (this model is also employed by some
other papers including Rajan, 1989). In fact, we assume that even
if S deviates from the grand coalition, each member of N  S keeps
the coalition, and the game is reduced to a two-person game be-
tween S and N  S. In contrast, regarding the order of the contract
between them, we consider the following two cases: Case 1, where
S (the deviating coalition) becomes the new leader, and Case 2,
where N  S (the remaining coalition) becomes the new leader.
We should note that in each case, there is no adjacency restriction
with regard to forming a coalition. That is, as in a usual coalitional
form game, any coalition is allowed to form, even among nonadja-
cent members in a supply chain. In the next subsection, we explic-
itly give the values of the coalitions for each case.
3.2.1. Value of the coalition
As mentioned in the previous subsection, we can set the value
of coalition v ðSÞ as the total payoff earned by the members in S
in the Stackelberg equilibrium of the two-person game between
S and N  S. Both coalitions determine their wholesale prices and
selling quantities in order to maximize their total profits, respec-
tively. We should note that while the leader–follower game is em-
ployed between both coalitions, all of the strategic variables with
regard to the members inside a coalition are determined simulta-
neously. Now, given any S, we suppose that S is the leader while
N  S is the follower. Then, for every i 2 S; S simultaneously
determines its wholesale prices wSiþ1 for its immediate downstream
neighbors and wSi1 for its immediate upstream neighbors
(regardless of them being insiders or outsiders) and its selling
P
quantities qSi , in order to maximize total profit i2S pSi . Following
this, N  S simultaneously determines the remaining wholesale
prices, that is, the prices set among N  S, and its selling quantities
qSj ðj 2 N  SÞ. Therefore, the game simply consists of two stages.
We should note that all members inside the coalition benefit from
choosing the identical selling quantity, since they must incur a po-
sitive cost for a leftover product. Therefore, with regard to selling
quantity, it suffices to consider the situation where coalition S
determines qSS while coalition N  S determines qSNS . Furthermore,
as in the non-cooperative situation previously discussed, we may
assume, according to Lemma 1 of Majumder and Srinivasan
(2006), that in the equilibrium, both quantities are always equiva-
lent (there is no leftover product). It should be noted that in this
game, there is a member who can be offered two wholesale prices
by its immediate downstream and upstream neighbors. Neverthe-
less, the following Lemma ensures that the equilibrium profit is al-
ways determined uniquely.
Lemma 1. In each case (Cases 1 and 2), the equilibrium profit of the
game is uniquely determined. The corresponding equilibria include the
one, where all wholesale prices are zero, except the price set between
the most downstream member inside the coalition excluding the
retailer and its immediate downstream neighbor. Formally, let S be the
leader and S0 be the coalition excluding the retailer n. Let iS0  maxi2S0 i.
Then, we have
(
iS0 þ 1 ðif S ¼ S0 Þ
wSi ¼ 0; i –
i S0 ðif S – S0 Þ:
By Lemma 1, for obtaining the value of coalition v ðSÞ, it suffices
that for each case, we set S as the leader or follower and obtain the
profit earned by S at the Stackelberg equilibrium in which all whole-
sale prices are zero, other than the price set between members iS0
and iS0 þ 1, which maximizes the total profit among the members
of S. In other words, the game is reduced to the simple two-person
Stackelberg game between iS0 and the retailer n, where they directly
determine the wholesale price between them and their quantities,
while incurring the total costs with regard to all members inside
their coalitions. More precisely, consider an example of a three-
person game with l ¼ 2, where the unique middleman is the leader
under pre-cooperation. For expositional purpose, we now consider
Case 2 and coalition S ¼ f1; 3g, implying that N  S ¼ f2g. Then,
since in Case 2, the first-mover is assumed to be N  S, we have
the Stackelberg game, where N  S first offers wholesale prices w1
and w3 to maximize pNS ¼ w3 q  w1 q  k2 q2 and then S determines
its quantity to maximize pS ¼ qð1  qÞ þ w1 q  w3 q  ðk1 þ k3 Þq2
(as mentioned above, we may assume that qSS ¼ qSNS  q holds in
the equilibrium). To find the equilibrium, according to the backward
induction process, we first should determine the optimal quantity
for S. However, clearly, the optimal q is dependent only on the differ-
ence w1  w3 , and not on each w1 and w3 , with regard to the first-
mover’s decision. This in turn implies that the profit of N  S is also
dependent only on the difference w3  w1 , and thus, it is sufficient to
set w1 ¼ 0 and optimize only w3 . This is reduced to the simple two-
person Stackelberg game between members 2 and 3, where member
2 first offers w3 , incurring cost k2 q2 , and then, retailer 3 determines
its quantity, incurring cost ðk1 þ k3 Þq2 . As a result, v ðSÞ is obtained as
pS earned in the equilibrium of this game. Lemma 1 indeed ensures
that in each case and for every coalition, the process of finding the
value of the coalition is reduced to such a simple model of a two-tier
supply chain.
Hence, considering the equilibrium profit of this game, it fol-
lows that the values of a coalition can be classified into two cases
according to whether or not the coalition includes the retailer. In
 Y. Kumoi, N. Matsubayashi / European Journal of Operational Research 238 (2014) 221–232 227

addition, the value of the grand coalition is clearly identical in both v A ðf2gÞ ¼ v A ðf4gÞ ¼ v A ðf1; 2gÞ ¼ v A ðf1; 4gÞ ¼ v A ðf2; 3gÞ
cases and equal to the monopoly profit in which the coalition 1
determines only the quantity of the final product, while setting ¼ v A ðf3; 4gÞ ¼ v A ðf1; 2; 3gÞ ¼ v A ðf1; 3; 4gÞ ¼ ;
32
all wholesale prices to be zero and incurring the total cost with re-
gard to all members in it. 1
Therefore, we can explicitly describe the values of the coalitions v A ðf5gÞ ¼ v A ðf1; 5gÞ ¼ v A ðf3; 5gÞ ¼ v A ðf1; 3; 5gÞ ¼ ;
36
for each case as follows. 1
v A ðf2; 4gÞ ¼ v A ðf1; 2; 4gÞ ¼ v A ðf2; 3; 4gÞ ¼ v A ðf1; 2; 3; 4gÞ ¼ ;
24
Lemma 2.
v A ðf2; 5gÞ ¼ v A ðf4; 5gÞ ¼ v A ðf1; 2; 5gÞ ¼ v A ðf1; 4; 5gÞ
1. For any partial cooperation S  N, the value of a coalition is
¼ v A ðf2; 3; 5gÞ ¼ v A ðf3; 4; 5gÞ ¼ v A ðf1; 2; 3; 5gÞ
given for each case as follows:
(a) For Case 1: 1
¼ v A ðf1; 3; 4; 5gÞ ¼ ;
8 28
>
> P 1
P   v AR ðSÞ ðif n 2 SÞ
>
>
<4 kþ
i2NS i
k þ1
j2N j 1
v A ðSÞ ¼ > ð5Þ v A ðf2; 4; 5gÞ ¼ v A ðf1; 2; 4; 5gÞ ¼ v A ðf2; 3; 4; 5gÞ ¼ v A ðNÞ ¼ :
20
>
>  1   v AN ðSÞ ðif n R SÞ:
:4 P
>
kþ
P
k þ2 We observe important characteristics, but some of these are
i2NS i j2N j
undesirable. Game GA does not generally satisfy supper-additivity
(b) For Case 2: as, for example, v A ðf2gÞ þ v A ðf3gÞ ¼ 160 9 1
> 32 ¼ v A ðf2; 3gÞ holds.
8 P P
> ki þ1 On the contrary, it is inessential since i2N v A ðfigÞ ¼
>
>
> P i2SP 2  v BR ðSÞ ðif n 2 SÞ 101 1
> 20 ¼ v A ðNÞ holds. The symmetric property can be verified
>
> 720
<4 kþ
i2S i
k þ2
j2N j
from the fact that every player with an identical cost other than
v B ðSÞ ¼ > P ð6Þ
the retailer is indifferent; for example, members 2 and 4 have
>
> k
> 
>
i2S i
2  v BN ðSÞ ðif n R SÞ: the same impact on the values of the coalition: v A ðf2; 3gÞ ¼
: 4 P ki þP kj þ1
>
i2S j2N v A ðf3; 4gÞ; v A ðf2; 5gÞ ¼ v A ðf4; 5gÞ, and so on. Since v A ðfigÞ > 0 for
all i, there is no dummy member, that is, a member who has no
2. For each case (Cases 1 and 2), the value of the grand coalition is
impact on any value of the coalition.
given as follows:
Next, we describe the values of the coalitions for game GB in
1 Case 2. These can be obtained according to the form in (6) as
v A ðNÞ ¼ v B ðNÞ ¼ P : ð7Þ
follows:
4 i2N ki þ 1

With these values of the coalitions, we define the coalitional

v B ðf1gÞ ¼ v B ðf3gÞ ¼ v B ðf1; 3gÞ ¼ 0;
form game corresponding to each case (Cases 1 and 2) as
GA ¼ ðN; v A Þ; GB ¼ ðN; v B Þ, respectively. As can be easily seen from
the forms in (5)–(7), the values of the coalitions depend only on
v B ðf2gÞ ¼ v B ðf4gÞ ¼ v B ðf1; 2gÞ ¼ v B ðf1; 4gÞ ¼ v B ðf2; 3gÞ
whether or not the coalition includes the retailer, and on the total 1
¼ v B ðf3; 4gÞ ¼ v B ðf1; 2; 3gÞ ¼ v B ðf1; 3; 4gÞ ¼ ;
costs of the members in the coalition. In other words, they do not 98
depend on the specific position of the leader under pre-integration.
1
Unfortunately, we should note that none of the games—GA and v B ðf5gÞ ¼ v B ðf1; 5gÞ ¼ v B ðf3; 5gÞ ¼ v B ðf1;3; 5gÞ ¼ ; v B ðf2; 4gÞ
GB —generally satisfy supper-additivity. On the contrary, only game 144
P 1
GB is necessarily essential (v B ðNÞ P i2N v B ðfigÞ holds), which en- ¼ v B ðf1; 2; 4gÞ ¼ v B ðf2; 3; 4gÞ ¼ v B ðf1; 2;3;4gÞ ¼ ;
sures that the imputation set is non-empty. However, we note that 81
this does not immediately imply that the formation of the grand
coalition is not possible, since our values of the coalitions are de- v B ðf2;5gÞ ¼ v B ðf4; 5gÞ ¼ v B ðf1; 2; 5gÞ ¼ v B ðf1; 4;5gÞ ¼ v B ðf2; 3;5gÞ
fined as ex-post values and thus can be obtained after deviating 3
¼ v B ðf3; 4;5gÞ ¼ v B ðf1; 2;3; 5gÞ ¼ v B ðf1;3; 4; 5gÞ ¼ ;
from the grand coalition. (As mentioned above, the possibility of 256
forming the grand coalition is discussed in Majumder & Srinivasan
(2006, Section 5).) Furthermore, the values have a symmetric prop- 1 1
erty in that all members other than the retailer n are indifferent v B ðf2; 4; 5gÞ ¼ v B ðf1; 2; 4; 5gÞ ¼ v B ðf2; 3; 4; 5gÞ ¼ ; v B ðNÞ ¼ :
80 20
regardless of their positions, as long as they have identical costs.
These properties characterize our games and play an important Although game GB is also not necessarily super-additive, it is
role in a later analysis. essential. Unlike Case 1, the values of individual coalitions for
In what follows, we illustrate our values of the coalitions with a members 1 and 3 who have no cost are zero (v B ðf1gÞ ¼
simple example. v B ðf3gÞ ¼ 0). Nevertheless they are not dummy members, as
v B ðf1; 2; 4; 5gÞ ¼ v B ðf2; 3; 4; 5gÞ ¼ 801 < 201 ¼ v B ðNÞ.
Example 3.1. Consider a five-stage supply chain. Let n ¼ 5 and
k1 ¼ k3 ¼ k5 ¼ 0; k2 ¼ k4 ¼ 2. We consider Case 1. According to the
forms in (5) and (7) in Lemma 2, we can obtain the values of the 4. Stability of vertical integration and fair profit allocation
coalitions. For example, v ðf1gÞ is obtained by calculating v AN ðSÞ,
the second equation in (5), since member 5 R S (i.e., the retailer is In this section, we analyze the core, Shapley value, and nucleo-
not included in the coalition). In a similar manner, we obtain the lus of our games formulated in the previous section in order to
values for all possible coalitions as follows: investigate fair profit allocation under the grand coalition. First,
we introduce the definitions of these terms; however, for details,
1 please see a standard textbook on cooperative game theory, such
v A ðf1gÞ ¼ v A ðf3gÞ ¼ v A ðf1; 3gÞ ¼ ;
40 as Myerson (1996).
228 Y. Kumoi, N. Matsubayashi / European Journal of Operational Research 238 (2014) 221–232
First, the core of a game is the set of all payoff vectors
x ¼ ðx1 ; . . . ; xn Þ satisfying
X X
xi ¼ v ðNÞ; and xi P v ðSÞ for any S  N:
i2N i2S
In words, the core is the set of payoff allocations of v ðNÞ that
satisfies coalitional rationality, where for every coalition S, the total
payoff among members in S is no less than the value of coalition
v ðSÞ. That is, if every member gains a core allocation, they have
no incentive to deviate from the grand coalition through any coa-
lition, which ensures the stability of the grand coalition. However,
note that the core is a set that can be empty, while it can also in-
clude numerous allocations. Therefore, it is useful to consider some
rational allocation rule that can always determine the payoff allo-
cation uniquely rather than considering a set of solutions. Among
such rules, consideration of the Shapley value and/or the nucleolus
is the best-known way.
Specifically, the definition of Shapley value is given as follows.
For the coalitional form game ðN; v Þ, the Shapley value of player i
is the payoff allocation for i given by the following formula:
X ðs  1Þ!ðn  sÞ!
/i ðv Þ ¼ ðv ðSÞ  v ðS  figÞÞ;
S;i2S # N
n!
where s is the number of members in S. In addition, we define
v ð/Þ ¼ 0 for the empty set.
As seen in Eq. (8), the Shapley value can be interpreted as the
average of marginal contributions to the formation of the grand
coalition. The payoff is allocated to each member in proportion
to the degree of contribution to the formation of the grand coali-
tion. More specifically, given a particular path to the grand coali-
tion, that is, an order of adding players, the marginal contribution
of each player i is defined as the marginal worth of i for joining
the coalition. Then, the Shapley value can be obtained by taking
the average of marginal contributions over all possible n! paths
to forming the grand coalition. Therefore, by calculating the Shap-
ley value, we can find a fair allocation among players under the
grand coalition, based on the evaluation of the power of each
player for forming the grand coalition.
We should note that the Shapley value need not satisfy the indi-
vidual and coalitional rationalities even if the core is non-empty,
which implies that it does not become a stable allocation if it is
not in the core. In contrast, the nucleolus is well-known as a stable
allocation, since it is ensured to necessarily be in the core if the
game has a core.
Specifically, the definition of the nucleolus is given as follows.
For any imputation x and any coalition S  N in the coalitional
form game ðN; v Þ, we first define the excess eðS; xÞ as
P to v B ðfigÞ. Therefore, we get as follows.
eðS; xÞ ¼ v ðSÞ  i2S xi . Then, given an imputation x, we arrange
jNj
all the 2  2 possible excesses in the decreasing order. Let bk ðxÞ
be the k-th highest excess under x. Then, let Að1Þ be the set of all
imputations that minimize b1 ; formally, Að1Þ ¼ argminx b1 ðxÞ. If
Að1Þ consists of a single imputation m , then m is the nucleolus.
Otherwise, we inductively find the set
argminx2Aðk1Þ bk ðxÞ; k ¼ 2; . . . ; 2jNj  1. Clearly, by this process, we
can find the nucleolus m ðv Þ, the unique imputation of the game that
minimizes the excess under the lexicographic order.
In Section 4.1, we analyze the core, and in the subsequent sub-
sections, we focus on the Shapley value and the nucleolus for the
game with non-empty core.
4.1. Analysis of core allocation
We immediately obtain the following clear results about the
existence of the core.
Theorem 4.1. The core of game GA is always empty. In contrast, the
core of game GB is always non-empty.
This theorem shows that in game GB , where we suppose pessi-
mistic members who expect that they cannot become the leader
after any deviation from the grand coalition, the core allocation al-
ways exists. In contrast, however, we also have that in game GA ,
where we suppose optimistic members who expect that they can
become the leader after any deviation, the core is always empty.
More precisely, as explained in Section 3, game GA is inessential
and thus does not have even an imputation set. As shown in
Majumder and Srinivasan (2006), in game GA , a deviating coalition
as the first mover can always obtain a higher profit even if it is a
unilateral deviation. Therefore, each member has an incentive for
the deviation, which is so high that the sum of profit allocations
that can prevent the unilateral deviation exceeds the grand coali-
tion’s profit. In contrast, in game GB , the most any member can ob-
tain from any deviation is a gain of the second mover. Hence,
members have a lower incentive for a deviation and thus the
sum of profit allocations that can prevent the unilateral deviation
does not exceed the grand coalition’s profit. Furthermore, under
our scenario where all non-deviating members keep their coalition,
there is almost no difference between the deviation by a single
ð8Þ
member and that by any group in the sense that the resulting prof-
its are equal as the equilibrium profit of the two-person Stackel-
berg game, although there is some difference between their
costs. This implies that the merit of any coalitional deviation van-
ishes, which indicates that any profit allocation satisfying individ-
P
ual rationality is a core allocation (v B ðSÞ 6 i2S v B ðfigÞ always
holds for any S  N).
4.2. Analysis of the Shapley value and nucleolus
As shown in the previous subsection, only game GB has a core
allocation, which can be stable. Therefore, as is rational for profit
allocation under the grand coalition, we analyze the Shapley value
and the nucleolus of game GB . Let /i ðv B Þ and mi ðv B Þ be the Shapley
values of member i and the allocation of member i in the nucleolus
of game GB , respectively.
Theorem 4.2.
1. With regard to the Shapley value of GB , we have the following.
(a) Let a and b be any two members other than retailer ða; b – nÞ.
If ka > kb , then /a ðv B Þ > /b ðv B Þ.
(b) If kn P ka for all a – n, then /n ðv B Þ > /a ðv B Þ.
2. The allocation of member i in the nucleolus of GB is specifically
P
v B ðNÞ i2N v B ðfigÞ
given by mi ðv B Þ ¼ v B ðfigÞ þ , which is proportional
n
(a) Let a and b be any two members other than the retailer
ða; b – nÞ. If ka > kb , then ma ðv B Þ > mb ðv B Þ.
(b) If kn P ka for all a – n, then mn ðv B Þ > ma ðv B Þ.
Indeed, the analyses of both the Shapley value and the nucleo-
AðkÞ ¼
lus derive the same conclusion, although the logic underlying each
result is slightly different. That is, part (a) of both Theorems 4.1-(1)
and 4.2-(2) means that the Shapley value as well as the profit in the
nucleolus are higher for a member who has a higher cost. This re-
sult might seem to be un-intuitive, since a more inefficient mem-
ber in terms of the cost structure is treated well. However, in
fact, any coalition benefits more from adding a member with a
higher cost, since the effect of solving the double marginalization
is relatively high. This is because a coalition including such a mem-
ber can gain a higher margin, which in turn indicates that the coa-
lition without him/her suffers from higher double marginalization.
Under our model of a wholesale price contract, the merit of vertical
 Y. Kumoi, N. Matsubayashi / European Journal of Operational Research 238 (2014) 221–232
integration is to entirely eliminate double marginalization. There-
fore, it is acceptable that a member who contributes more to this
elimination is well treated, resulting in a higher Shapley value. Fur-
ther, since the value of the coalition itself is relatively high when it
includes a member with a higher cost because of gaining a higher
margin, the profit should be allocated more to such a member for
the grand coalition to be stable, which leads to a higher allocation
in the nucleolus. Next, part (b) also implies that the retailer has a
strong influence on achieving stable vertical integration. This is be-
cause under the situation where the game is reduced to one be-
tween two coalitions including and excluding the retailer, the
retailer always can gain some margin regardless of the order of
decision, since it necessarily can set the monopoly price for the fi-
nal product. Therefore, the retailer also always obtains a higher
marginal contribution, as well as contributes to a higher value of
the coalition.
In Table 1, we present numerical results for several cases of
n ¼ 5 (including the one in Example 3.1) and n ¼ 7. It can be easily
verified that all the results are consistent with 4.2. We note that for
the case of n ¼ 5, the Shapley value is always a core allocation,
while for the case of n ¼ 7, it fails in some cases. As seen in Exam-
ple 3.1, there is no dummy member in game GB , since every mem-
ber necessarily gains a positive marginal contribution when it
eventually joins the coalition (v B ðNÞ  v B ðN  figÞ > 0). This im-
plies that in the Shapley value, the grand coalition’s profit must
be allocated to every member to some extent, even to a member
whose value of the individual coalition is zero. Therefore, if there
are enough such members (i.e., members who have no cost, other
than the retailer), the total of such ‘‘excess’’ can be sufficient to pre-
vent the Shapley value from belonging to the core.
5. Extension to cases of multiple retailers
In the analysis of the Shapley value and the nucleolus in the
previous section, we reach a conclusion that regardless of the lead-
ership position, the members who exert a stronger power for ver-
tical integration are the members incurring higher costs and the
retailer. Hence, this implies that although the retailer always has
a stronger influence, the power of the upstream manufacturer be-
comes relatively weaker, unless it is the fixed leader or it has a
higher cost. However, both these conditions seem to be unrealistic.
The rapid changes in the business environment may result in the
contract leader being changed, as explained in the introduction
section. In contrast, it is usual that technological progress can re-
solve some inefficiency, resulting in a decrease in cost. Therefore,
we have one simple question: given such circumstances, under
which conditions can the upstream manufacturer be in a beneficial
position in a grand coalition? Finding the answer to this question
might help the decision-making process of manufacturers who suf-
fer from power shifts to retailers. Based on this motivation, we ex-
Table 1
Numerical results for game GB with n ¼ 5 and n ¼ 7.
ðk1 ; k2 ; k3 ; k4 ; k5 ; k6 ; k7 Þ ð/1 ; /2 ; /3 ; /4 ; /5 ; /6 ; /7 Þ
(0, 0 ,0 ,0,1) (0.019, 0.019, 0.019, 0.019, 0.05)
(0, 0, 1, 1, 1) (0.009, 0.009, 0.014, 0.014, 0.016)
(0, 1, 1, 1, 1) (0.008, 0.010, 0.010, 0.010, 0.012)
(1, 1, 1, 1, 1) (0.009, 0.009, 0.009, 0.009, 0.009 , 0.009)
(0, 1, 10, 100, 0) (0.003, 0.003, 0.004, 0.008, 0.003)  101
(0, 2, 0, 2, 0) (0.008, 0.012, 0.008, 0.012, 0.010)
(0, 0, 0, 0, 0, 0, 1) (0.013, 0.013, 0.013, 0.013, 0.013, 0.013, 0.045)
(0, 0, 0, 0, 1, 1, 1) (0.007, 0.007, 0.007, 0.007, 0.011, 0.011, 0.014)
(0, 1, 1, 1, 1, 1, 1) (0.004, 0.005, 0.005, 0.005, 0.005, 0.005, 0.006)
(1, 1, 1, 1, 1, 1, 1) (0.004, 0.004, 0.004, 0.004, 0.004, 0.004, 0.005)
229
tend our model of a serial chain to a simple case where there are
multiple monopolistic retailers, while there is only one manufac-
turer. For expositional simplicity, we do not consider middlemen,
which implies that the supply chain consists of two tiers. In addi-
tion to Guardiola et al. (2007) introduced in Section 2, Tyagi (1999)
employs a similar setting to ours. However, he also assumes that
the manufacturer is always the Stackelberg leader. In addition, he
considers a Cournot competition among retailers in a homoge-
neous good market. With this structure, we focus on the analysis
of the Shapley value and the nucleolus, which are of most interest
to us.
5.1. Model
We consider a two-tier supply chain with one manufacturer
(indexed as 0) and n retailers (indexed as 1; . . . ; n). Let
N þ ¼ f0; 1; . . . ; ng be the set of members. We suppose that retailers
are regional monopolies or fully differentiated with regard to their
tastes/brands with each other. Let l be the contract leader in the
supply chain. If manufacturer 0 is the leader, then this manufac-
turer simultaneously offers the wholesale prices to n retailers,
and then simultaneously determines their quantities of final prod-
ucts. In contrast, if any retailer i is the leader, then this retailer of-
fers the wholesale price to manufacturer 0 at first, and then the
manufacturer determines its selling quantity to retailer i, as well
as offering the wholesale prices to other n  1 retailers. Finally,
all retailers simultaneously determine their quantities of final
products. We denote the wholesale price offered to member i as
wli (i – l). As in the previous discussions, the quantity sold by the
manufacturer to retailer i is equal to that from retailer i to its con-
sumers, which is denoted by qli . Furthermore, the cost function of
each member is given as follows. For manufacturer 0, the cost func-
tion is
X
n
2
C 0 ðql1 ; . . . ; qln Þ ¼ k0i ðqli Þ ; k0i P 0;
i¼1
and that for retailer i is
2
C i ðqli Þ ¼ ki ðqli Þ ; ki P 0:
That is, the cost of the manufacturer is additive over retailers,
implying that there is neither a scale economy nor a diseconomy.
In fact, this assumption would be appropriately realistic for cases
of sales to regional monopoly retailers under the existence of trans-
portation and/or holding costs. Finally, the inverse demand function
of the final products is identical for all retailers and is given as
follows:
Pi ðqli Þ ¼ 1  qli :
Under these settings, each member determines its wholesale price
or quantity to maximize its own profit.
ðm1 ; m2 ; m3 ; m4 ; m5 ; m6 ; m7 Þ / in the core?
(0.019, 0.019, 0.019, 0.019, 0.05) Yes
(0.006, 0.006, 0.016, 0.016, 0.020) Yes
(0.004, 0.011, 0.011, 0.011, 0.014) Yes
(0.008, 0.008, 0.008, 0.008, 0.010) Yes
(0.003, 0.003, 0.005, 0.009, 0.003)  101 Yes
(0.005, 0.015, 0.005, 0.015, 0.011) Yes
(0.013, 0.013, 0.013, 0.013, 0.013, 0.013, 0.045) Yes
(0.004, 0.004, 0.004, 0.004, 0.014, 0.014, 0.018) No
(0.001, 0.005, 0.005, 0.005, 0.005, 0.005, 0.008) No
(0.004, 0.004, 0.004, 0.004, 0.004, 0.004, 0.006) Yes
230 Y. Kumoi, N. Matsubayashi / European Journal of Operational Research 238 (2014) 221–232

5.1.1. Case where the manufacturer is the leader Lemma 3.

We consider the retailers’ problems when the manufacturer is
the leader ðl ¼ 0Þ. Given the wholesale price w0i offered to each re- 1. For any partial cooperation S  N þ , letting s be the number of
tailer i; i determines its quantity of final product q0i to maximize the retailers in S, the value of the coalition is given as follows:
following profit function: ( ns s
4ð3kþ2Þ
þ 4ð2kþ1Þ ; ðif 0 2 SÞ;
max p0i ðq0i ; w0i Þ ¼ Pi ðq0i Þq0i  w0i q0i  C i ðq0i Þ: uA ðSÞ ¼ s
q0i 4ð3kþ1Þ
; ðif 0 R SÞ;

Therefore, manufacturer 0 as the leader offers its wholesale prices 8 ðnsÞk s

< þ 4ð2kþ1Þ ; ðif 0 2 SÞ;
w0i to retailer i, in order to maximize the following profit function: 4ð3kþ1Þ2
uB ðSÞ ¼
: sðkþ1Þ
; ðif 0 R SÞ:
X
n
4ð3kþ2Þ2
0 0 0
max p 0 ðw1 ; . . . ; wn Þ ¼ w0i q0i ðw0i Þ  C 0 ðq01 ðw01 Þ; . . . ; q0n ðw0n ÞÞ:
w01 ;...;w0n 2. For games HA and HB , the value of the grand coalition is
i¼1

Note that each q0i is a function of w0i , the wholesale price offered by n
uA ðNþ Þ ¼ uB ðNþ Þ ¼ :
0 to i. 4ð2k þ 1Þ

5.1.2. Case where the retailer is the leader 5.3. Analysis of the Shapley value and the nucleolus
Let us consider the case where any retailer i is the leader
ðl ¼ i – 0Þ. First, we consider the problem of retailer j other than In this subsection, we focus on the difference between the fair
the leader ðj – 0; iÞ. Given that this retailer is offered wholesale profit allocations to the manufacturer and to the retailers. To do
price wij , it determines its quantity of final product qij to maximize this, we should clarify the existence of the core in games HA and
the following profit function: HB . As can be easily inferred from Theorem 4.1, the core exists only
in game HB , which is immediately shown in the following theorem.
max pij ðqij ; wij Þ ¼ Pj ðqij Þqij  wij qij  C j ðqij Þ:
qij
Theorem 5.1. The core of game HA is always empty. In contrast, the
Next, we consider the problem of the manufacturer. Given that core of game HB is always non-empty.
manufacturer 0 is offered wholesale price wi0 , it determines the We next investigate the Shapley value and the nucleolus for
quantity qii sold to retailer i as well as wholesale prices wij offered game HB , which are denoted by /i ðuB Þ and mi ðuB Þ, respectively.
to retailers j other than the leader, to maximize the following profit The following theorem mentions how the profit allocations to
function: the manufacturer and retailers under these values change with
X the number of retailers.
max pi0 ðwij–0;i ; qii ; wi0 Þ ¼ wi0 qii þ wij qij ðwij Þ  C 0 ðqii ; qij–i;0 ðwij ÞÞ:
wij–0;i ;qii
j–i;0
Theorem 5.2.
qij
Note that is a function of wij ,
the wholesale price offered by 0 to j.
Finally, let us consider retailer i’s problem as the leader. Retailer 1. In game HB , the Shapley value is identical for all retailers, and is
i determines the wholesale price wi0 to maximize the following denoted by /R ðuB Þ. Further, the profit allocations to all retailers
profit function: in the nucleolus are identical, and are denoted by mR ðuB Þ.
max pii ðwi0 Þ ¼ Pi ðqii ðwi0 ÞÞqii ðwi0 Þ  wi0 qii ðwi0 Þ  C i ðqii ðwi0 ÞÞ: 2. In game HB , the Shapley value of the manufacturer increases in the
wi0 number of retailers, while that of retailers is constant regardless of
the number of retailers. Formally,
Note that qii is a function of wi0 .
@/0 ðuB Þ @/R ðuB Þ
> 0; ¼ 0:
5.2. Value of the coalition @n @n
Further, in the nucleolus of game HB , the profit allocation of the man-
Based on the model formulated above, we define the coalition ufacturer increases in the number of retailers, while that of retailers is
form game in a manner similar to that in the case of a single retai- constant regardless of the number of retailers. Formally,
ler. For simplicity, here, we focus on only the case of cost symme-
try: k0i ¼ ki ¼ k; for all i ¼ 1; . . . ; n. As in the previous section, we @ m0 ðuB Þ @ mR ðuB Þ
> 0; ¼ 0:
assume that after any coalition S deviates from the grand coalition, @n @n
the remaining members continue to form coalition N þ  S, so that 3. There exists threshold level n ^ B regarding the number of retailers
the value of the coalition is defined as the equilibrium profit of the such that n P n ^ B () /0 ðuB Þ P /R ðuB Þ holds. Further, there exists
Stackelberg game between S and N þ  S. We suppose that as in the threshold level n  B regarding the number of retailers such that
case of a single retailer, all members in a coalition can determine nPn  B () m0 ðuB Þ P mR ðuB Þ holds. Specifically, n ^ B , and n
 B are
their wholesale prices and quantities simultaneously. Further, given as follows:
regarding the order of the contract, we formulate two games:  
4 3 2
one where S is the leader and the other, where S is the follower. 81k þ 168k þ 126k þ 41k þ 5 5
^B ¼
n 1 < ^
n B 6 ;
The resulting coalition form games are denoted by HA ¼ ðN þ ; uA Þ 4 3 2
84k þ 156k þ 108k þ 31k þ 3 3
and HB ¼ ðN þ ; uB Þ, respectively. Then, as shown in Lemma 1, it 4 3 2  
63k þ 135k þ 106k þ 37k þ 5 5
can be easily verified that each Stackelberg game underlying HA B ¼
n 4 3 2
1<n B 6 :
99k þ 189k þ 128k þ 35k þ 3 3
and HB can be reduced to the game where in the equilibrium,
wholesale prices are determined to be positive only between the The first part of this theorem is straightforward because the de-
manufacturer and each retailer in the coalition excluding the man- mand and cost structures are symmetric across all retailers. The
ufacturer. Therefore, the values of the coalition can be classified second part of the theorem shows that both the Shapley value
into two cases according to whether or not the coalition includes and the profit allocation in the nucleolus of the manufacturer
the manufacturer. Specifically, for any S ðS  NÞ, each value of monotonically increase in the number of retailers, while those
the coalition is given as follows: for retailers stay constant. The reason why these allocations to
 Y. Kumoi, N. Matsubayashi / European Journal of Operational Research 238 (2014) 221–232
retailers stay constant is as follows: independent demands and
costs imply the simple ‘‘star’’ network, where the manufacturer
has bilateral relations with each retailer. Therefore, even if some
retailers form a coalition, any synergy effect does not occur, which
implies that the marginal contribution of each retailer is constant
in the number of retailers included in the coalition. On the other
hand, even if a retailer joins to a coalition including the manufac-
turer, each retailer can contribute only to eliminating double mar-
ginalization with regard to the bilateral link between this retailer
and the manufacturer. Therefore, the marginal contribution of a re-
tailer is also independent of the timing of joining to a coalition. In
contrast, the marginal contribution of the manufacturer is increas-
ing in the number of retailers included in the coalition, since the
manufacturer can at once resolve double marginalization with re-
gard to multiple links with those retailers. Hence, the Shapley va-
lue, which is the average marginal contribution, is increasing in n
only for the manufacturer. On the other hand, under the network
consisting of the collection of bilateral links, the nucleolus is
clearly obtained as the sum of profit allocations in the nucleoli over
those links. Specifically, if we consider the case of single retailer
(n ¼ 1), then it is obvious that the profit allocations in the nucleo-
lus for the manufacturer and the retailer are given by
uB ðf0gÞ þ uB ðf1;0gðuB ðf0gÞþu B ðf1gÞÞ
and uB ðf1gÞ þ uB ðf1;0gðuB ðf0gÞþu B ðf1gÞÞ
2 2
respectively. Therefore, in the nucleolus in the whole system, the
profit allocation for each retailer is the same as this value, while
that for the manufacturer is n times as the allocation in the case
of single link.
Therefore, it follows that the power of the manufacturer relative
to the retailers is increasing in the number of retailers. In fact, the
third part of the theorem calculates the number of retailers such
that the relationship between the fair profit allocations to the man-
ufacturer and the retailers are reversed. As proved in the previous
section, in a serial supply chain with symmetric costs, these alloca-
tions to the retailers are always higher than those to the manufac-
turer. Indeed, this is consistent with the result of n
However, Theorem 5.2 also shows that the threshold level is less
than 2, which implies that the Shapley value and the profit alloca-
tion in the nucleolus of the retailers become higher than those of
the manufacturer only under a serial supply chain. In addition, it
can be easily verified that n ^ B and n
 B monotonically decrease in
the cost parameter k. This implies that as the cost is increasing,
the minimum number of retailers such that the fair allocation to
the manufacturer exceeds those of the retailers is decreasing. As
observed in the previous section, the effect of vertical integration
on the elimination of double marginalization is significant under
higher costs. This result implies that under the existence of multi-
ple retailers, the manufacturer has a higher contribution to this
elimination, relative to the retailers.
6. Conclusion
In this paper, we discuss vertical integration in light of the con-
tract leadership position in supply chains by a normative approach.
Although there are many works on vertical integration in two-tier
supply chains, vertical integration in multi-tier supply chains has
received little attention. In addition, we can find many real-world
examples of a variety of leadership positions and rapid changes
therein. Motivated by such examples, we investigate stable and fair
profit allocations among members under vertical integration in a
serial supply chain by utilizing cooperative game theory. From
our model analysis, we obtain the following.
Regarding the stability of vertical integration, there exist core
allocations in the case where all members are pessimistic in the
sense that they are certain that they will not become the contract
leader after any coalitional deviation from the grand coalition. In
231
contrast, in the case where they are optimistic in the sense that
they are sure they will become the leader, the core is always
empty, which implies that integration is unstable. These results
suggest that vertical integration is likely to break up in situations
where all or some members decide to take contract leadership.
We also obtain some insights about fair allocations under verti-
cal integration using the analyses of the Shapley value and the
nucleolus, which are fortunately the same under the two different
allocation rules. That is, when all members expect that the leader-
ship position may change after integration, the retailer and mem-
bers who have higher costs have relatively strong power, and as
such, more of the grand coalition’s profit should be allocated to
them. This result suggests that under situations facing a variable
leadership and rapid cost improvements due to technological
advances, the retailer is likely to be in a much more beneficial
position under vertical integration; however, for the upstream
manufacturer, this is not necessarily so. In this regard, our other
analysis shows that even under such conditions, the upstream
manufacturer can have relatively strong power if it can sell its
product to multiple monopolistic retailers. In other words, forming
a coalition with multiple retailers is one of the keys for the manu-
facturer to have a stronger influence on realizing stable vertical
, integration.
These results would give us some new managerial insights
regarding strategic vertical integrations in real-world supply
chains. For instance, regarding the example of the supply chain
for providing e-books introduced in Section 1, which motivates
our study, we might be able to explain why Amazon.com and Mac-
millan failed to agree on continuing their cooperation, resulting in
the coalition breaking up. That is, it seems that both companies
might have believed to have stronger power as the leader under
vertical integration. Our result suggests that with such optimistic
players, vertical integration cannot be stably maintained. However,
our result also shows that integration among pessimistic players
^B ; n
 B > 1. can be stable. In this sense, it might be reasonable that Amazon fi-
nally softened its attitude by taking into account the counterpart’s
increasing power in the market, which led to a compromise.
Although most executives of companies in supply chains well
understand the efficiency of vertical integration, they might have
concerns regarding profit allocation among members under verti-
cal integration. From our investigation, we now have an implica-
tion for this: when it is expected that the contract leader in a
supply chain can change, it is important for each member not to
determine to take the contract leadership. In addition, regarding
profit allocation, we clarify who should be allocated more profit.
Finally, we should note that some of our results largely depend
on the assumptions employed in our study. For example, while our
model assumed increasing marginal costs, we might have another
cost structure that significantly alters the results. As Majumder and
Srinivasan (2006) pointed out, our basic Stackelberg framework
does not make sense under a serial chain with constant or dimin-
ishing cost structures, without any additional setup. However, in
multiple retailer cases employed in Section 5, it might be meaning-
ful to consider some economies of scale arising from cooperation
among retailers, as Guardiola et al. (2007) discussed. In contrast,
in this study, we assumed a serial monopolistic supply chain where
there is one member in each tier. However, it would be interesting
to consider the case of a serial chain where there is a competitive
tier with horizontal competition among several members and the
case of horizontal competition among several serial chains, as ex-
plored in some papers (Tyagi, 1999; Corbett & Karmarker, 2001;
Carr & Karmarker, 2005; Majumder & Srinivasan, 2008, and so
on). In addition, more importantly, we should note that for formu-
lating our problem as an n-person coalitional form game, we em-
ployed some restrictive assumptions on how to define the values
of the coalitions. As precisely explained in Section 3, at least in
232 Y. Kumoi, N. Matsubayashi / European Journal of Operational Research 238 (2014) 221–232
the manner employed in the related literature, we are sure that
these assumptions can be justified to some extent. However, it is
not clear whether or not even a slight change in the assumptions
might significantly alter some of our results. For example, our
study discussed two extreme cases regarding how the remaining
members react to a deviation by some members from the grand
coalition. Clearly, though rather complicated, other intermediate
coalition structures can be considered. In this regard, however, it
should be noted that how to justify the values of the coalitions is
a harder task beyond the problem of vertical integration. This is left
as an open problem in game theory as a topic of coalition forma-
tion in the presence of an externality, including its applications
to horizontal mergers, provision of public goods, and so on. Never-
theless, we are sure that our study could provide the first frame-
work of model analysis in the area of vertical integration, which
requires further investigation in the future.
Acknowledgments
We thank the Editor, Lorenzo Peccati, and two anonymous ref-
erees for helpful comments. The second author is supported by the
Grants-in-Aid for Scientific Research (C) 24510201 of the Ministry
of Education, Culture, Sports, Science and Technology of Japan.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.ejor.2014.03.022.
References
Arrington, J. M. (2010). Google Nexus One: The TechCrunch Review, 2010. <http://
techcrunch.com/2010/01/05/google-nexus-one-the-techcrunch-review/>.
Aumann, R. (1959). Acceptable points in general cooperative n-person games. In L.
Tucker (Ed.), Contributions to the theory of games IV. Annals of mathematics
studies (Vol. 40). Princeton: Princeton University Press.
Cachon, G. P., & Lariviere, M. A. (2004). Supply chain coordination with revenue
sharing contracts. Management Science, 51, 30–44.
Carr, S. M., & Karmarker, U. S. (2005). Competition in multiechelon assembly supply
chains. Management Science, 51, 45–59.
Chander, P. (2007). The gamma-core and coalition formation. International Journal of
Game Theory, 35, 539–556.
Chander, P., & Tulkens, H. (1997). The core of an economy with multilateral
environmental externalities. International Journal of Game Theory, 26, 379–401.
Choi, S. C. (1991). Price competition in a channel structure with a common retailer.
Marketing Science, 271–296.
Choi, S. C. (1996). Price competition in a duopoly common retailer channel. Journal
of Retailing, 72, 117–134.
Chwe, M. S.-Y. (1994). Farsighted coalitional stability. Journal of Economic Theory, 63,
299–325.
Corbett, C. J., & Karmarker, U. S. (2001). Competition and structure in serial supply
chains with deterministic demand. Management Science, 47, 966–978.
Coughlan, A. T., & Wernerfelt, B. (1989). On credible delegation by oligopolies: A
discussion of distribution channel management. Management Science, 35,
226–239.
Edirisinghe, N. C. P., Bichescu, B., & Shi, W. (2011). Equilibrium analysis of supply
chain structures under power imbalance. European Journal of Operational
Research, 214, 568–578.
Fiestras-Janeiro, M. G., Garcia-Jurado, I., Meca, A., & Mosquera, M. A. (2011).
Cooperative game theory and inventory management. European Journal of
Operational Research, 210, 459–466.
Gerchak, Y., & Yunzeng, W. (2004). Revenue-sharing vs. wholesale-price contracts
in assembly systems with random demand. Production and Operations
Management, 25–33.
Granot, D., & Sosic, G. (2003). A three stage model for a decentralized distribution
system of retailers. Operations Research, 51, 771–784.
Granot, D., & Sosic, G. (2005). Formation of alliances in internet-based supply
exchanges. Management Science, 51, 92–105.
Guardiola, L. A., Meca, A., & Timmer, J. (2007). Cooperation and profit allocation in
distribution chains. Decision Support Systems, 44, 17–27.
Hart, S., & Kurz, M. (1983). Endogenous formation of coalitions. Econometrica, 51,
1047–1064.
Jeuland, A. P., & Shugan, S. M. (1983). Managing channel profits. Marketing Science, 2,
239–272.
Li, S., Zhu, Z., & Huang, L. (2009). Supply chain coordination and decision making
under consignment contract with revenue sharing. International Journal of
Production Economics, 88–99.
Magretta, J. (1998). The power of virtual integration: An interview with Dell
Computer’s Michael Dell. Harvard Business Review, 76.
Majumder, P., & Srinivasan, A. (2006). Leader location, cooperation, and
coordination in serial supply chains. Production and Operations Management,
15, 22–39.
Majumder, P., & Srinivasan, A. (2008). Leadership and competition in network
supply chains. Management Science, 54, 1189–1204.
McGuire, T. W., & Staelin, R. (1983). An industry equilibrium analysis of
downstream vertical integration. Marketing Science, 2, 161–190.
Meca, A., & Timmer, J. (2008). Supply chain collaboration. In V. Kordic (Ed.), Supply
chains: Theory and application. Vienna, Austria: I-Tech Education and Publishing.
Meca, A., Timmer, J., Garcia-Jurado, I., & Borm, P. E. M. (2004). Inventory games.
European Journal of Operational Research, 156, 127–139.
Moorthy, K. (1988). Strategic decentralization in channels. Marketing Science, 7–4,
335–355.
Myerson, R. B. (1996). Game Theory: Analysis of Conflict. Cambridge, Massachusetts:
Harvard University Press.
Nagarajan, M., & Sosic, G. (2007). Stable farsighted coalition in competitive markets.
Management Science, 53, 29–45.
Nagarajan, M., & Sosic, G. (2008). Game-theoretic analysis of cooperation among
supply chain agents: Review and extensions. European Journal of Operational
Research, 719–745.
Nagarajan, M., & Sosic, G. (2009). Coalition stability in assembly models. Operations
Research, 57, 131–145.
Netessine, S. (2009). Supply webs: managing, organizing, and capitalizing on global
networks of suppliers. In P. Kleindorfer & Y. J. Wind (Eds.), The network
challenge: strategy, profit, and risk in an interlinked world. Pearson Prentice Hall.
Norde, H., Pham Do, K. H., & Tijs, S. (2002). Oligopoly games with and without
transferable technologies. Mathematical Social Sciences, 43, 187–207.
Pan, K., Lai, K. K., Leung, S. C. H., & Xiao, Di. (2010). Revenue-sharing versus
wholesale price mechanisms under different channel power structures.
European Journal of Operational Research, 203, 532–538.
Publisher Wins Fight With Amazon Over E-Books. The New York Times, January
31, 2010. <http://www.nytimes.com/2010/02/01/technology/companies/
01amazonweb.html>.
Rajan, R. (1989). Endogenous coalition formation in cooperative oligopolies.
International Economic Review, 30, 863–876.
Rao, L. (2010). Amazon Caves To Macmillan’s eBook Pricing Demands. The
TechCrunch Review, 2010. <http://techcrunch.com/2010/01/31/amazon-caves-
to-macmillans-ebook-pricing-demands/>.
Slikker, M., Fransoo, J., & Wouters, M. (2005). Cooperation between multiple news-
vendors with transshipments. European Journal of Operational Research, 167,
370–380.
Sosic, G. (2006). Transhipment of inventories among retailers: Myopic vs. farsighted
stability. Management Science, 52, 1493–1508.
Spengler, J. (1950). Vertical integration and anti-trust policy. Journal Political
Economy, 58, 347–352.
Tirol, J. (1988). The theory of industrial organization. Cambridge, Massachusetts: MIT
Press.
Tyagi, R. K. (1999). On the effects of downstream entry. Management Science, 45,
59–63.
van Dusen, S. (1998). The manufacturing practices of the footwear industry: Nike vs.
the competition. <http://www.unc.edu/andrewsr/ints092/vandu.html>.
Zhao, J. (1999). A necessary and sufficient condition for the convexity in oligopoly
games. Mathematical Social Sciences, 37, 189–204.