Operationalizing Value-Based Business Strategy

Joshua S. Gans Glenn MacDonald

Melbourne Business School Olin School of Business
University of Melbourne Washington University in St. Louis
Michael Ryall
Melbourne Business School
University of Melbourne
26th April, 2005

Abstract
Brandenburger and Stuart (1996) identi…ed coalitional games as a means of provid-
ing precise notions of value to evaluate strategic opportunities. In this paper, we show
how coalitional game theory can be utilized to operationalize these approaches. In
particular, we demonstrate the importance of considering full competitive interactions
(rather than simply added value) when applying coalitional game theory and also how
this can be employed to provide insights into the workings of an existing economic
activity as well as to suggest ways that the activity might be altered to a …rm’s ad-
vantage. We illustrate with an application in which an innovator considers whether to
commercialize a new technology.

This paper builds on and replaces an earlier paper by MacDonald and Ryall (2003). We thank Douglas
Hirai for his careful proof reading of the original version.
1 Introduction
‘Value-based business strategy’ is a term coined in a highly in‡uential paper by Branden-
burger and Stuart (1996) who o¤er an exact de…nition of the value that can be created by
…rms transacting with suppliers and buyers. This naturally leads to the notion of ‘added
value,’a measure of a …rm’s contribution to the aggregate value produced in a given market.
As Brandenburger and Stuart (1996) argue, a …rm’s added value imposes an upper bound
on the value it can appropriate. They go on to demonstrate how the use of certain strategies
allow a …rm to create favorable asymmetries within a market to ensure its added value is
strictly positive. The notion that …rms should focus on their added value is popularized in
Brandenburger and Nalebu¤ (1996) has become a central feature of strategy and economics
as taught in business schools.1
The added value logic described above is underpinned by coalitional game theory. In par-
ticular, application of the “core”as a solution concept reinforces the importance of positive
added value as a necessary condition for positive appropriation.2 However, as MacDonald
and Ryall (2004) demonstrate, the core can also be used to characterize the maximum a
…rm might appropriate (something that is no greater than their added value but may be less
than it) as well as the least amount guaranteed by participating in the market (something
that is no less than their reservation price but may be greater than it). The latter notion is
important because, regardless of how the indeterminacy in a …rm’s range of core-consistent
appropriation levels is resolved (often by appealing to notions of “bargaining”strength; see
Brandenburger and Stuart, 2004 for a formal treatment), the minimum is assured by com-
petitive forces alone. For this reason, MacDonald and Ryall (2004) de…ne a …rm’s absolute
competitive advantage to be its minimum level of appropriation consistent with a core distri-
bution. An absolute competitive advantage is su¢ cient to guarantee a …rm’s participation
in the contemplated market (i.e., regardless of its relative bargaining skill, etc.).3
The purpose of this paper is to take the use of the core seriously as a means of opera-
1
See, for example, Saloner, Shepard and Podolny (2001); Besanko, Dranove, Shanley and Schaefer (2004);
and Gans (2005).
2
See MacDonald and Ryall (2004) for a formal proof.
3
Brandenburger and Stuart (1996) and MacDonald and Ryall (2004) both show how various strategic
options may be used to guarantee a positive competitive advantage.

1
tionalizing value-based business strategy.4 Our goal is to demonstrate how this methodology
can be used analyze the e¢ cacies of a …rm’s various strategic opportunities.5 We go beyond
qualitative speculation with respect to so-called “generic”strategies based upon incomplete
analysis of the market to quantitative assessments of speci…c strategies in complex envi-
ronments using the complete set of competitive restrictions implied by the given market.
A key theme is that, while useful, focussing entirely on added value often comes up short
when assessing strategic opportunities. In some cases, a strategy that both creates value and
increases a …rm’s added value unambiguously leads to a reduction in its potential range of
appropriation. Only a full and quantitative analysis of competitive interactions can uncover
these risks and also illustrate the drivers of these seemingly counter-intuitive outcomes.
In this respect, this paper is completely distinct from MacDonald and Ryall (2004). That
paper provides general necessary and su¢ cient conditions that determine whether the lower
bound of a …rm’s core allocation is positive or not. In contrast, the present paper deals
with the calculation of the entire set of core allocations and hence, a di¤erent set of issues.
As such, it is presented more in the style of Brandenburger and Stuart (1996) – utilizing
examples with the explicit goal of making that paper useful in practice. In this respect, it
is a synthesis of MacDonald and Ryall (2004) (the competitive part) and Brandenburger
and Stuart (2004) (the bargaining part); showing both how to apply them in practice and
4
Hereafter, we follow the convention adopted in MacDonald and Ryall (2004) of substituting the term
“competitive”for “core.”The intention is to emphasize the e¤ects of a …rm’s transaction opportunities (which
are of central importance) and to help the intuition of those who may be unfamiliar with this framework.
5
Lippman and Rumelt (2003a, 2003b) critique the traditional use of economic-based approaches in the
evaluation of strategic opportunities and suggest coalitional game theory may overcome the perceived short-
comings. Their critique is based on the observation that models of perfect competition are unable to de-
termine who might appropriate rents or quasi-rents (i.e., supra-normal pro…ts) in a market. Moreover, the
presumption of perfect competition may often be unwarranted and, hence, give rise to misleading answers
to important questions of strategic interest. While suggesting that coalitional game methods may overcome
this di¢ culty, they highlight de…ciencies in current approaches. In particular, they …nd the core’s generic
indeterminacy problematic.
Typically, however, the core places signi…cant bounds on the range of appropriation outcomes available
to an agent in a given strategic interaction. Moreover, as we demonstrate here, it does so in ways that
cannot be readily extracted from indicators of value creation and appropriation (such as the concept of
added value). We conclude that examing strategy through the lense of coalition game theory can actually
give very meaningful predictions and facilitate the evaluation of a …rm’s strategic opportunities.

2
demonstrating why a complete analysis (beyond simple added value) is necessary.
The paper proceeds as follows. In Section 2, we provide the basic notation necessary
to assess strategic options. An essential feature of this material is the sharp distinction
made among (i) the physical process of value creation, formally described by what we call a
value opportunity; (ii) the competitive forces that in‡uence how the value produced in a value
opportunity is shared among the participating agents, described by a competitive distribution
of value; and (iii) any other factors that in‡uence how created value is ultimately distributed
(e.g., bargaining, negotiation, luck, psychology,...), formalized by an appropriation factor (as
in Brandenburger and Stuart, 2004).
Section 3 turns to the potential issues that arise when evaluating generic strategies (such
as product di¤erentiation and productivity improvements). In particular, we demonstrate
that productivity improvements may be strategically risky for …rms and provide an example
whereby these unambiguously reduce a …rm’s appropriation possibilities. One of the issues
stems from the fact that traditional speculation with respect to generic strategies seems
to assume that …rms begin from a position of zero appropriation. However, we see no
good reason to restrict the analysis in this way. In our approach, …rms may already be
appropriating value at the time they consider new strategic moves. We also consider how
strategic investments may change …rms’competitive positions along a vertical value chain.
Section 4 provides an extended application of how to conduct a complete analysis of
strategies using data based on real world decision problems. In particular, we examine
whether commercialization of a process innovation is worthwhile when the innovator ne-
gotiates with large but asymmetric buyers. We show how to compute the returns to this
commercialization as well as how to identify strategies that can enhance returns further.
This illustrates the power of utilizing coalitional game theory in applied settings.

2 Notation and preliminary analysis

Following MacDonald and Ryall (2004), we consider a strategic interaction among n agents,
one of which we refer to interchangeably as “the …rm”and as “agent 1”. To begin, we require
some notation to keep track of groups of agents. Our notation for a group of agents that
includes at least one, but not every, agent, is a vector g; consisting of zeros and ones, where

3
a one in the ith component of g indicates agent i belongs to g: Thus (1; 0; :::; 0) is the group
including only agent 1; (0; 1; :::; 1) is the group of all agents except agent 1; etc. Since 2n
groups can be formed from n agents, one of which is empty and another that includes all
agents, there are 2n 2 groups that a vector g might describe. The group of all agents and
the group of all agents except agent i, both of which play a special role, are denoted 1 and
1 i ; respectively.
The next element of our analysis is a formal description of the value creation possibilities
available to the agents through arms-length transactions. This is done by de…ning the
available value opportunities (hereafter, VO) to be a pair (V; v) ; where V > 0 is a scalar and
v is a non-negative vector of length 2n 2: The …rst component, V; is the aggregate value
anticipated as a result of the economic activities of the agents. The second component, the
vector v; describes the alternative value creation opportunities available to subsets of agents.
Speci…cally, to every group, g; there corresponds a component of v; v g ; specifying the value
that group could generate were it to do so independently of any additional agents.
The collection of agents and their value-producing alternatives, as summarized by (V; v),
are the key inputs to our analysis; values each agent might appropriate are its output.
De…ne a distribution of value, denoted ; to be a vector of length n whose ith component,
i; represents the value appropriated by agent i: Below, we detail how this value is to be
calculated in applications. For the moment, it can be thought of simply as whatever value
agent i ultimately appropriates.
Inner product is indicated by “ ”; for example, the number of agents in g is 1 g: A
distribution of value, ; is competitive if it meets the following two conditions:

1 = V; (1)

and for every group g;

g vg : (2)

Condition (1) is simply an adding up condition, i.e., a competitive distribution must dis-
tribute the value actually produced, V . Condition (2) requires that no group, g; be able to
improve over what it would receive, given the competitive distribution, by creating value on
its own. The logic behind this condition is that, if, instead, g < v g for some group g,
the agents in g could make themselves strictly better o¤ by ignoring the other agents, doing

4
whatever it takes to generate v g ; and then sharing this value among the group. Thus, if some
distribution involves g < v g for some group g, it is di¢ cult to see how the distribution
of value might occur. To put it the other way around, competitive distributions are those
that might occur because the alternatives available to agents do not undermine value being
distributed in the contemplated way. A distribution of value being competitive is a necessary
condition for its being a plausible candidate for the outcome of a strategic interaction.6
Note that, in this context, “competition” has a very speci…c meaning: agents have al-
ternatives and the freedom to act on them if they choose to do so. The alternatives are
competing activities in which agents might engage. Conditions (1) and (2) formalize how
competition limits the way value can be distributed.
min max
Given a VO, there are numbers and with the feature that for any appropriation
level by the …rm, 1; there is a competitive distribution of value in which the …rm receives
min max min
1 if and only if 1 : Equivalently, is the smallest value of 1 that is
max
consistent with both (1) and (2); and likewise for being the maximum. More speci…cally,

min
minn f 1 j1 = V; and for all g; g vg g (3)
2R+

and
max
maxn f 1 j1 = V; and for all g; g vg g : (4)
2R+

(3) and (4) describe min

and max
as the optimized value of a pair of linear programs.7 Thus,
min max
once data have been gathered and (V; v) speci…ed, and can be computed (e.g., by
using the Excel solver). This computation is useful since it tells us the …rm’s appropriation
min max
possibilities, i.e., at least but no more than : In addition it yields some ancillary
information that, as we show below, can be very useful.
min
Solutions to (3) and (4) are competitive distributions that result in 1 = and
max
1 = ; respectively. These distributions (there might be several) describe the ways value
min max :
can be distributed among agents such that the …rm appropriates or . Further,
6
There is, of course, a generic issue regarding the existence of competitive distributions with these prop-
erties – the ‘empty core’ problem. In all of the examples considered in this paper, the core exists. Stuart
(1997) provides a more comprehensive examination of existence issues in coalitional games of the form we
consider here.
7 min max
Note that and vary across the agents. Given our focus on …rm 1, we do not require additional
notation to keep track of these values for other agents.

5
since each linear program has 2n 1 constraints, associated with each solution is a (2n 1)-
min max min
vector of “shadow prices,” denoted and ; respectively. Each component of
min
corresponds to a group (either 1 or some g); and quanti…es the e¤ect on of loosening,
by one unit, the constraint with which that group is associated. That is, if some v g is reduced
min min
by $1, or V increased by $1, the corresponding components of tell us by how much
max
falls; is interpreted similarly: The majority of shadow princes turn out to be zero. These
are associated with constraints that are not binding: varying them (within limits that are
min max
easily calculated) does not a¤ect the outcome: The information embedded in and
can serve a very useful role in the …rm’s evaluation of activities that might alter the VO. We
have more to say on this subject below.
Firm 1’s added value is de…ned as av V v1 1
(i.e., the aggregate value anticipated
with the …rm’s participation minus the amount that can be produced without it). As argued
by Brandenburger and Stuart (1996), and formally shown by MacDonald and Ryall (2004),
max
av1 > 0 is necessary for > 0. It is for this reason that Brandenburger and Stuart (1996)
and, indeed, Brandenburger and Nalebu¤ (1996) argue that the …rm should focus on strate-
min max
gies that lead to av > 0. Typically, however, < < av: This is the indeterminacy
issue discussed above. Not only does the …rm typically face a range of values other than av;
but sometimes av is not even attainable as part of a competitive distribution. Limiting the
min max
analysis to av is only useful in the very special circumstances when = = av: Thus,
a key focus on operationalizing value-based strategy must be a consideration of the full set
of interactions between agents in a competitive situation. It is only by doing a complete
min max
analysis that one can specify how, when < ; the …rm should assess the opportunity
min
to take part in some VO. The …rm knows competition guarantees it at least and no
max
more than ; but this may not be enough information for it to decide how best to proceed.
For example, suppose an outsider is considering making a tender o¤er for the …rm. How
should the …rm be valued? Alternatively, suppose the …rm considers taking some costly ac-
min max
tion that will alter the VO, possibly both lowering and increasing : How should it
evaluate this opportunity? Brandenburger and Stuart (2004) present a very useful solution.
To see how it works, begin by considering, instead, the familiar von Neumann and Mor-
genstern approach to how agents value uncertain prospects, e.g., lottery tickets. In the risk
neutral case (consistent with what we study below) the value of an uncertain prospect may

6
be represented by its expected value, where the probabilities used to compute the expecta-
tion can be interpreted as the agent’s subjective beliefs about the likelihood of the various
possible outcomes. The agent behaves as if this expected value will be the outcome of the
uncertain prospect. In a similar manner, Brandenburger and Stuart, applied to our setting,
min max
show that when facing the interval of possible appropriation, [ ; ]; the value the …rm
attaches to this may be represented by

max min
^ + (1 ) ;

where 2 [0; 1]: ^ is interpreted as the …rm’s expectation about the result of factors in
addition to competition. We refer to as the …rm’s appropriation factor (AF), and write

min max min

^ + ; (5)

in which the …rst term is the minimum appropriation guaranteed by competition and the sec-
ond is the additional value the …rm believes will appropriate as a result of extra-competitive
factors. For example, when = 1; the …rm believes it will appropriate the maximum level
max
of value consistent with competitive forces; i.e., ^ = :
While the speci…cation of (V; v) typically involves a fairly direct assessment of the un-
derlying technology, consumer demand and other market data, identi…cation of may be
more elusive. In some cases, as we demonstrate below, evaluation does not rely on knowl-
edge of (i.e., a …rm’s choice is invariant with respect to ). In others, the …rm’s current
appropriation, say ^ 0 (where the subscript indicates period zero); can be employed along
with knowledge of the current VO, (V0 ; v0 ) ; to calculate the current AF, 0 :
min
^0 0
0 max min
:
0 0

With 0 in hand, the …rm can consider whether there is any good reason to expect that the
AF for the ongoing interaction might be di¤erent.
Below, we show how our framework can be employed to uncover, explore, and evaluate
any opportunities to change the AF or VO. There are a couple of key points that are critical
to that discussion, but that are best discussed while attention is focused on (3) and (4).
First, observe that the unambiguous way for the …rm to increase ^ is to raise both
min max
and : Note, however, that the constraints in (3) and (4) are the same. Thus, the

7
 min max
constraints that prevent from being higher are the same ones that prevent from
min max
being lower. Fortunately, projects designed to simultaneously increase and are not
min
precluded; nonzero components of often correspond to alternatives that are di¤erent
max
from the those identi…ed by the nonzero components of ; in which case there may be
min max
scope for increasing both and : Still, …guring out how to alter the VO to increase
^ may involve some cleverness. Experience con…rms this, as well as the considerable value
min max
of having the nonzero components of and to light the way.
Second, the constraint corresponding to the adding up condition, (1), must always be
min max
met. Thus, if V increases, but v remains unchanged, cannot rise and cannot fall.
This suggests, correctly, that actions increasing the overall value of the economic interaction
are not necessarily those that increase ^ : More generally, actions that alter the VO or AF
min max
change V and numerous alternatives in v: The calculations revealing the way ;
and ^ are a¤ected are essential to untangling what may be a very complicated collection of
interactions.
min
MacDonald and Ryall (2004) provide necessary and su¢ cient conditions for > 0
and apply those insights in several applications. Here, our goal is not to demonstrate gen-
eral theorems about generic strategies but to consider how to operationalize their model to
evaluate potentially complex strategic opportunities using real, quantitative data.

3 Evaluating Generic Strategies

The key to operationalizing value-based business strategy is to formulate the competitive
distribution problem as a linear program based on calculations of the value created by alter-
native groups as in (3) and (4). Doing so allows the analyst to consider market environments
with many …rms and multiple vertical stages. Here we demonstrate, using a simple example,
that common qualitative prescriptions regarding the desirability of opportunities may not
hold when the full range of potential transactions are taken into account. This implies that
partial approaches to evaluating strategies –e.g., based on perfect competition or Branden-
burger and Stuart’s (1996) added value –may not only be insu¢ cient but may actually be
misleading as prescriptive decision-making tools.
Our example is as follows: we suppose there are two suppliers, S1 and S2 , two buyers, B1

8
and B2 , and two …rms, F1 and F2 , who intermediate between the buyers and suppliers. For
a coalition to create value, at least one supplier, one …rm and one buyer must be present.
The following table provides the underlying parameters that generate value. Each agent has
a capacity (in units of an homogenous good) and a (constant) value contribution (per unit
of the homogeneous good). For the suppliers and …rms, this value contribution is negative
(implying a marginal cost) and for the buyers it is positive. These underlying parameters
are used to calculate the maximum value that would be created by any given coalition but in
a real world analysis, those values could be calculated based on actual industry information
regarding the nature of production technologies and buyer demand.

Agent S1 S2 F1 F2 B1 B2
Capacity 50 50 20 20 20 20
Unit value $5 $5 $2 $2 $10 $10
Given these parameters and the fact that $10 > $5+$2, both buyers’demand will be satis…ed
and there will be excess capacity in supply. The homogeneity of the products at each stage
means that it is not relevant which buyer consumes output from which …rms/suppliers. Thus,

V = (10 2 5) min[K(S1 ) + K(S2 ); K(F1 ) + K(F2 ); K(B1 ) + K(B2 )] = 3(40) = $120:

where K(i) is the capacity of agent i. In this case, alternative coalitions have a value $120,
$60 or $0 depending upon whether both …rms and buyers are members, only one buyer or
…rm are members or there are no members from one vertical stage. Moreover, it is easy to
calculate the following:

Agent S1 S2 F1 F2 B1 B2
Added Value 0 0 60 60 60 60
min
0 0 0 0 0 0
max
0 0 60 60 60 60
The suppliers have excess capacity – both individually and jointly – and so can be easily
substituted without any loss in coalitional value. Consequently, they do not have positive
added value and cannot expect to appropriate any surplus. In unit pricing terms, the price
for inputs will be $5 per unit. In contrast, the total capacity of …rms equals total buying

9
capacity. Thus, each of those four agents has a positive added value equal to the total value
their capacity generates. This implies that to create that value any …rm and buyer pair must
be receiving at least that value or else they could deviate and form a separate vertical chain.
However, this means that the amount of surplus each receives is indeterminate; that is for
any given buyer and …rm, the sum of their distribution must be at least $60 but there is no
restriction on their individual payo¤. So, for each, their expected return ranges from $0 to
$60.8
Taking this as a base case, we now consider several alternative strategic opportunities for
F1 (our focus for this exercise). The idea is to consider strategies that are generally considered
to create, in Brandenburger and Stuart’s (1996) terminology, “favorable asymmetries.”These
are strategies that make …rms di¤erent from their competitors. Here we consider (i) F1
specializing its product to B1 and (ii) an increase in F1 ’s productivity before turning to
examine how vertical chain issues impact strategy evaluation.

3.1 Specialization

In the simple example, buyers are indi¤erent as to the …rm they purchase from. However,
one ‘generic’strategy involves a …rm improving the quality of its product in the eyes of one
or more buyers. For instance, suppose that F1 is considering an ex ante investment that
would result in its product being more valuable to B1 and only B1 ; so that B1 might value
products bought from F1 at $11 (rather than $10). Absent any other changes, this would
lead to V = $140 and the following:

Agent S1 S2 F1 F2 B1 B2
Added Value 0 0 80 60 80 60
min
0 0 0 0 0 0
max
0 0 80 60 80 60
Thus, the specialization increases total value and also F1 ’s added value. However, it does
8
In the context of competitive markets, value is apportioned through the “terms of trade”which certainly
include prices but may also be more broadly construed. Although these speci…cs are outside our formalism,
it may be instructive to note that unit pricing of the …nal good between $7 – $10 is consistent with this
range of expected returns.

10
 min
not change as both F1 and B1 must participate in creating this additional value; the
competitive constraints a¤ect the sum of their payo¤s only.
If, however, this specialization also increased B1 ’s demand for the product (say to 25
units), then it is straightforward to show that, while neither V nor F1 ’s added value changes
min
(both are $140 and $80 respectively), as the following table indicates, F1 ’s becomes $15.

Agent S1 S2 F1 F2 B1 B2
Added Value 0 0 80 60 80 45
min
0 0 15 15 0 0
max
0 0 80 60 65 45
This is primarily because the increased demand from B1 reduces B2 ’s added value (to $45).
The fact that demand exceeds the …rms’ total supply means that F2 ’s minimum becomes
$15 even though its added value and maximum remain unchanged. Thus, both …rms gain
from F1 ’s investment as it expands the market. However, the maximum appropriation for
B1 falls to $65 (below its added value).
Interestingly, this also allows us to analyze the likelihood of potential reactions from
other players. For instance, it may be imagined that B2 may encourage F2 to make a similar
specialization. The results of this are as follows:

Agent S1 S2 F1 F2 B1 B2
Added Value 0 0 80 80 80 80
min
0 0 15 15 0 0
max
0 0 80 80 65 65
This follow-on investment would raise V and also F2 ’s added value (to $160 and $80). While
max
B2 ’s added value would rise (and B1 ’s would fall to $65), this would only raise its .
So, for a given , its gain is equivalent to B1 ’s initial gain. Overall, however, specialization
bene…ts each …rm but harms each buyer. So F1 is not harmed if F2 imitates its strategy.
Of course, this raises the question as to whether buyers might be better o¤ if suppliers
engaged in specialized investments to them. Those inputs would still be converted by …rms
to buyer products but those products would have more value. So starting from our base
case, if S1 specialized its inputs for B1 expanding its demand by 5 units and increasing unit

11
value by $1, this could lead to V = $140, B1 ’s, B2 ’s and S1 ’s added values becoming $80, $45
max
and $20 respectively. It is easy to show that, while this raises S1 ’s (to $20), it does not
min max
change B1 ’s and only raises its to $65 just as in the case of a similar investment
by F1 .
The point here is that, while specialization may be just the kind of strategy that both
creates value and can enhance the added value of the parties involved, it may have asymmetric
e¤ects if this changes the balance of supply and demand in a market. Moreover, the external
e¤ects of these changes can lead to imitative reactions from others that may change the
overall balance of appropriability across the vertical chain.

3.2 Productivity increase

We turn now to consider another ‘generic’ business strategy: increasing productivity. An

increase in a …rm’s productivity generally means being able to produce more at a lower unit
cost. In our example, one interpretation of this may be a straight decrease in F1 ’s unit
costs, say to $1.8 per unit. In this case, V will increase to $124 and as can be seen from the
following table F1 ’s added value also rises, by $4 to $64 (that is, the entire level of the cost
saving).

Agent S1 S2 F1 F2 B1 B2
Added Value 0 0 64 60 60 60
min
0 0 4 0 0 0
max
0 0 64 60 60 60
In this situation, an investment to reduce unit costs is potentially pro…table for F1 . Moreover,
min max
F1 has to potential to appropriate the entire value of this increase because its and
increase by the total increase in value.
However, it is rare that such a straight unit cost reduction would be what is achieved by
a productivity increase. Starting from the basis where its unit costs were $1.8, suppose that
F1 increased the operating e¢ ciency of its business, not only reducing its unit costs further
to $1.5 but increasing production capacity to 35 units. In this case, the increase in value
is even greater to V = $138 and F1 ’s added value increases a further $14 to $78. Again,
this investment has the quality of increasing F1 ’s added value by precisely the level of the

12
increase in total surplus. From the following table, however, this does not translate into an
increase in F1 ’s pro…ts.

Agent S1 S2 F1 F2 B1 B2
Added Value 0 0 78 15 68 68
min
0 0 3 0 45 45
max
0 0 33 15 68 68
min max
Notice now that both and have decreased! While this productivity improvement
has dramatically reduced F2 ’s added value and has slightly increased those of the buyer’s,
min
both buyer’s have risen considerably. This is because having both buyers in a coalition
with F1 is more valuable than before where a single buyer and F1 could generate as much
value. This means that any given buyer must receive more value to prevent them from
forming a coalition with F1 alone. This arises due to the combined e¤ect of F1 ’s cost reduction
and its e¤ective capacity expansion.9
As discussed above, solutions to (1) and (2) have associated (2n 1)-vectors of shadow
min max min
prices, and : Each component of corresponds to a group (either 1 or some
min
g) and quanti…es the e¤ect on of loosening, by one unit, the constraint with which
max
that group is associated; and likewise for : We can get some insight into this e¤ect by
looking at the shadow prices when F1 has unit costs of $1.80 and capacity of 20. In this
min max
case, the computations leading to and yield the following shadow prices (omitting
multipliers equal to zero and for groups not involving F1 ):

min max
Group Group
S1 ; S2 ; F1 ; F2 ; B1 ; B2 1 S1 ; S2 ; F1 ; F2 ; B1 ; B2 1
S2 ; F1 ; F2 ; B1 1 S1 ; F1 ; B1 ; B2 1
S1 ,F1 ; B2 1 S2 ; F1 ; B1 ; B2 1
max
Looking at …rst, if characteristics can be adjusted so that more value is created overall,
9
In many respects, this is just another illustration of the “paradox of value” (Nelson, 1957). What is
signi…cant here (in contrast to examples provided by Lippman and Rumelt (2003a)) is that this paradox
arises for a strategy which increases the …rm’s added value by the full amount of the increase it total value.
This suggests that paradox may not be a curiousity but a signi…cant issue in many strategic environments.

13
 max
this will raise so long as F1 ’s position as a monopoly supplier to both buyers is not itself
signi…cantly improved. Note that when F1 ’s capacity is 20, this group will have to permit F1
to appropriate most value as groups with just F1 and a single buyer create the same value.
As capacity is expanded, however, this no longer the case. Hence, the value created by this
coalition rises faster than others meaning that buyers will need to appropriate more value in
any competitive distribution. Consequently, F1 ’s maximum appropriate must fall.
min min
Turning to ; however, creating more value overall tends to lower : This suggests
max min min
a potential di¢ culty in increasing without reducing . But it also explains why
falls in this case. Simply put, total value rises by an amount greater than the coalitional
values where just a single buyer is present. These coalitions tended to improve F1 ’s minimal
appropriation as they created competition for it among the buyers. Examining the shadow
prices illustrates how subtle some of the interactions can be even in a simple example. Below
we consider a more realistic application and use shadow prices to identify more innovative
strategic options.
To see all of this another way, note that, starting from our base case, had F1 simply
increased capacity (say to 25 units), this would not have increased total value nor F1 ’s added
value but would have reduced F2 ’s added value (to $45). This too would not be a strategy
max
worth pursuing although it is useful to note F1 ’s in this case would be $45; greater than
what it was for the higher productivity expansion considered above. The following …gure
min max
illustrates what happens to F1 ’s added value and its and for various the ranges of
unit cost reduction and capacity increases relative to our base case.

Figure: Change in F1 ’s Appropriation Relative to K1 = 20 Base

14
 Cost Reduction
πmax falls and
πmax and πmin rise πmin rises

πmax and πmin fall

Capacity
Expansion

This demonstrates a complete analysis of the outcomes of competition is necessary in

order to evaluate proposed strategic investments appropriately. All of these productivity im-
provements increase total value and increase the investor’s added value by the same amount.
Ordinarily, this would be the idea investment from a business strategy perspective. However,
when considering its full impact on coalitional values, we have been able to determine how
this will translate into an investment return. First, the maximum return may not be a …rm’s
added value and may even fall as that value rises. Second, the minimum return – i.e., the
amount a …rm is guaranteed from competition alone – may also fall as added value rises.
This translates to a certain fall in pro…tability regardless of that …rm’s . Thus, not only
may …rms not appropriate the full value of their investments but may diminish their ability
to appropriate value altogether. Again, this suggests the importance of a complete analysis
when evaluating business strategies –generic or otherwise.

3.3 Vertical Chain

The …nal issue to emphasize in the analysis of generic strategies is how value might ‡ow along
a vertical chain. Thus far, these issues have been suppressed by our assumption that suppliers
are not capacity constrained and do not, therefore, have an added value or consequent claim

15
on total surplus. What would a situation of constrained suppliers do to our evaluation of
generic strategies?
Let’s begin from a base case where suppliers have only 15 units of capacity each. In this
case, V = $90 and coalitional game outcomes are as follows:

Agent S1 S2 F1 F2 B1 B2
Added Value 45 45 30 30 30 30
min
15 15 0 0 0 0
max
45 45 30 30 30 30
Now consider a situation where F1 specialized its product to B1 (increase B1 ’s willingness
to pay by $1 but now for 25 units) so that V = $105.

Agent S1 S2 F1 F2 B1 B2
Added Value 45 45 45 30 45 30
min
15 15 0 0 0 0
max
45 45 45 30 45 30
Note that in contrast to our earlier unconstrained suppliers example, while F1 and B1 ’s
max min
added values and increase, F1 ’s stays at $0. This is because the increase in B1 ’s
demand does not make F1 ’s product any scarcer relative to the suppliers. Moreover, as the
suppliers were already scarce, their expected appropriation does not change. Hence, the
expected return to F1 is relatively lower.
A similar e¤ect occurs for productivity improvements. Even taking a ’good’improvement
that only reduced F1 ’s unit costs (say to $1.80), the outcome is as follows (where V = $94):

Agent S1 S2 F1 F2 B1 B2
Added Value 46 46 34 30 30 30
min
15 15 2 0 0 0
max
46 46 34 30 30 30
In this case, while total value and F1 ’s added value increase by the full amount of the cost
min
saving –i.e., $4 – for F1 does not rise by that much. Thus, part of the increased surplus
from this productivity improvement would ‡ow back up the vertical chain to the suppliers.

16
The more pronounced is the suppliers’capacity constraint, the stronger this ’leakage’of value.
Moreover, these e¤ects only exacerbate the negative impacts of productivity improvements
that also e¤ectively increase F1 ’s capacity.
This demonstrates that vertical chain e¤ects can also be critical in evaluating business
strategies. Put simply, by considering all stages of value creation, the analyst can consider
where critical resource bottlenecks lie. This will have important implications for the expected
appropriation of value from investments by others along the vertical chain; something that
cannot be seen by focussing on value creation and added value alone.

3.4 Summary

Brandenburger and Stuart (1996) note that added value imposes an upper bound on an
agent’s ability to appropriate value. They then focus upon the evaluation of how generic
strategies might improve added value from a base case where a …rm has no added value at
all. However, in many business environments, …rms under analysis already have a positive
added value. In this case, it might have been presumed that actions that both increased
total value and increased added value would be unambiguously pro…table for the …rm.
The simple example considered here has demonstrated the di¢ culties associated with
this presumption. Put simply, in a complete analysis of appropriation in a coalitional game,
generic strategies that raise added value do not necessarily translate into higher pro…ts. First,
the bene…ts from an investment may ‡ow along the vertical chain to other agents and even
to competitors –as was the case with an improvement in buyer willingness to pay. Second,
some value creating investments that raise the investor’s added value may lead to increases
in e¤ective capacity and reduce the scarcity of an agent’s resources and capabilities. We
have demonstrated that this e¤ect can be so large as to reduce both the maximum and the
minimum an agent can expect to appropriate.
This highlights the critical importance of utilizing complete models of coalitional value
when evaluating business strategies. Even ’safe and sound’ strategic presumptions can be
undone by di¢ cult to predict swings on coalitional value.

17
4 Application: Technology Commercialization
We now employ an extended example to illustrate our methodology. Our objectives for the
example are three-fold: …rst, to illustrate some of the mechanics of the analytical procedure
detailed above; second, to show how the methodology may enable critical steps in the …rm’s
decision process to become comparatively straightforward; and …nally, to demonstrate how
some of the exploratory thinking emphasized above can be applied. This is admittedly not
a complete analysis of even this situation but it does demonstrate the main bene…ts and
practicality of employing these methods in a real world environment.10

4.1 Facts

A …rm is considering commercialization of a new technology that it has intellectual property

protection over. It is estimated that commercialization may involve investments up to $100
million. The new technology is a process innovation that will drastically reduce the produc-
tion cost for a special type of digital product (hereafter, DP) already with a considerable
market presence.
There are four large producers of DPs, and a larger number of small producers. The …rm
has already decided to transact only with the large OEMs (hereafter, simply “OEMs”). The
OEMs currently serve 85% of the market, distributed as:

OEM Share
A 40
B 25
C 15
D 5

The OEMs sell machines directly to downstream …rms who, in turn, provide services to
retail customers. The downstream market has many …rms, low concentration, and millions
of retail customers. The new technology is related to the innovator’s core business, but is
just one, albeit important, part of a complicated system making up the DP. For this reason,
10
The example is based on a strategic analysis conducted for a major technology company by one of the
authors. The details are altered to produce an example that retains realism while having only a “family
similarity” to the actual situation.

18
the innovator has already decided not to enter the DP market directly. Instead, if it elects
to commercialize, the innovator will patent the technology and license its use to one or more
of the OEMs.
The OEMs are believed to have similar cost structures for manufacturing DPs. Speci…-
cally, under the existing processes, marginal cost is estimated to be (over the relevant range
of production) a constant $150k per DP. The innovation provides a cost-e¤ective replace-
ment for the most costly component of the DP and allows simpli…cation of numerous other
components. Estimates are that the new technology will reduce marginal cost to about $50k.
Developing the technology, despite the potential $100 million up-front expenditure, has
some very appealing features: given patents and commercialization, the innovator will possess
a valuable, unique, non-imitable asset for which multiple …rms can compete. The …rm’s
researchers, based on their knowledge of the technology, professional activities, etc., are
con…dent there is no technology that might threaten uniqueness for several years, conceivably
a decade.
The DP is itself a second-generation machine that will surely replace the …rst gener-
ation analog product; the latter is relatively costly to produce, o¤ers slower and poorer
performance, and has inferior environmental characteristics. So, while each of the four OEM
producers made their DPs available relatively recently, sales of DPs are accelerating.
The retail segment of the market is projected to grow at 5% per year (i.e., new retail
locations requiring a DP). Aside from market growth, demand for DPs also comes from
replacement of the …rst generation. Retailers have shown no interest in replacing …rst gen-
eration products during their useful life. Given setup costs, even attempts to unload older
R
machines on eBay have been unsuccessful. Thus, given the 10 year useful life of this equip-
ment, downstream users are expected to replace about 10% of their existing base per year;
OEMs anticipate penetration will be below 10% in the …rst few years due to end-user aware-
ness and early adoption issues. The estimated di¤usion path is factored into the calculations
below.

4.2 Calculating group values

As noted above, if an OEM licenses the process innovation its unit costs will fall signi…cantly.
Some of this will be passed through to its customers. Let m be an OEM’s margin with the

19
older technology and M (> m) its margin if it licenses the new technology.11 In addition,
adopting the innovation will have an impact on an OEM’s market share. Suppose that i’s
current market share is si . Then it is assumed that if the combined market share of non-
adopting …rms is S, i’s post-adoption market share will be si (1 + (1 S)) where 2 [0; 1].
Here is a parameter that captures how much of the non-adoptees’market share those who
adopt will attract. Not surprisingly, this increased share will be lower, the more OEMs there
are who acquire a license.
Let us label the innovating …rm as 1. We can represent the group values, vt (g), for time
t, as follows:

X X
vt (g) = M ( si + (1 si ))Dt if 1 2 g
i2g i2g

X
vt (g) = m( si )(1 )Dt if 1 2
=g
i2g

where Dt is the total demand for DP units at time t. Based on market enquiries it was
determined that the following represented reasonable working assumptions: (i) M = $75,
(ii) m = $50, (iii) = 0:2 and (iv) Dt is determined by the following table (where 5 years
was regarding as a conservative time horizon for this analysis):12

Year Unit Demand (000s)

1 0.7
2 1.0
3 5.0
4 8.0
5 11.0
Here demand assumptions build in a 5% growth in the market plus additional growth due to
the attractiveness of the second-generation product. Clearly, these are working assumptions
11
In reality, both of these margins would themselves be determined by negotiations between OEMs and
their customers. However, a lack of information as to some of the key parameters in those relationships has
led us to consider a reduced form approached; albiet one informed by industry consultations and experience.
12
MacDonald and Ryall (2003) provide an extended discussion of these assumptions and some of the issues
in calculating group value for this application.

20
and a complete analysis would explore the robustness of any conclusions to changes in those
assumptions.
The new technology implies signi…cant changes in the DP market. In order to de-
cide whether commercialization is advisable, assumptions must be made about the post-
commercialization environment. We began with assumptions that were relatively conserv-
ative/pessimistic. If commercialization appears attractive under these assumptions, then a
positive decision is straightforward; otherwise, a more careful look at the speci…c assumptions
is called for.

4.3 Analysis

Given the facts and assumptions, we turn to answering the question of immediate interest
to the innovator.

4.3.1 Is commercialization worth $100 million?

The …rst question in setting up the analysis is which agents should be explicitly included
in the analysis? As we mention above, the retail market is large, both in terms of the
number of consumers and the number of retail …rms providing services. Modelling these
agents explicitly is impractical and, according to the innovator, unnecessary. Therefore, we
begin the competitive analysis with only the innovator and the four OEMs; the others, e.g.,
retailers, customers, etc., are incorporated indirectly.
How do we calculate the VO? In line with the discussion above, we assume that the
innovation will ultimately be used e¢ ciently. That is, if the innovator is imagined to restrict
e¢ cient access to the technology, there is an o¤er the excluded OEMs could make for the
technology that would be acceptable to the innovator and the other OEM(s). In this example,
the most e¢ cient use of the technology involves all four OEMs using it. The annual value
that the innovator and OEMs can share is therefore

1
Vt = ($350 $50) (:88) Dt :
4

21
Allowing for replacement of old machines and the 5% growth in demand, we have ($ millions):

Year Vt
1 46:2
2 66:0
3 330:0
4 528:0
5 726:0
min max
Given these data, solving the linear programming problems yields the and
values in the table following. Assuming, for example, an discount rate of 12%, the present
discounted value of the innovator’s assured appropriation is $244.6 million.

Year
($ millions) 1 2 3 4 5 PV (12%)
min
t 10:5 15:0 75:0 120:0 165:0 244:6
max
t 16:5 23:5 117:5 188:0 258:5 383:2
Thus, even under what is argued to be a very conservative of assumptions, OEM competition
to use the new technology should allow the innovator to appropriate about $144 million be-
yond the up-front cost of commercialization. The OEMs’alternatives to using the innovation
bound the innovator’s appropriation at $383.2 million.
Note that, in this case, assessment of the innovator’s AF is unnecessary to answer whether
commercialization should proceed. The forces of competition alone are su¢ cient to guaran-
tee a positive NPV. Essentially, this arises because OEMs without a license not only lose the
potential cost reduction but market share as well. As MacDonald and Ryall (2004) demon-
min
strate, in the absence of this type of externality, = 0 and the AF would be critical in
determining whether commercialization should proceed. Indeed, in the next section, which
contemplates an actual mechanism by which to implement the expected appropriation, the
AF does play a role.

4.3.2 What are the license fees, by OEM?

In applications such as this, it is not enough simply to identify the anticipated appropriation
stream; management must …nd a means by which these appropriation levels will be achieved.

22
Given the widely divergent market positions of the four OEMs, a volume-based licensing
scheme seems a good candidate. The innovator plans to license to all four OEMs. Anticipated
unit sales of DPs by the four OEMs are as summarized below.

DP Unit Sales (000s)

Year OEM A OEM B OEM C OEM D All Other Total
1 0:29 0:18 0:11 0:04 0:08 0:7
2 0:41 0:26 0:16 0:05 0:12 1:0
3 2:07 1:29 0:78 0:26 0:60 5:0
4 3:31 2:07 1:24 0:41 0:96 8:0
5 4:56 2:85 1:71 0:57 1:32 11:0

Recall that the OEMs expect to appropriate $75k per machine on sales to retailers. Begin-
ning with year 1, the minimum level of innovator appropriation consistent with competition
is $10.5m. In this year, the four OEMs are expected to sell 616 DPs, which comprise an
88% share of the market (i.e., of the 700 units projected to be sold in year 1). To appropri-
ate $10.5m through volume-based licensing fees, the innovator must charge $10.5m/616, or
$17.0k per unit.
At the high end, a $26.7k per unit fee delivers a total of $16.5k, the competitive maximum,
to the innovator in the …rst year. While the innovator has no experience negotiating deals in
this speci…c market, it is an established …rm with substantial bargaining experience in similar
transactions. Based on this history, as well as an analysis of the speci…c OEM business units,
the innovator believes it can negotiate a licensing fee between $24.8k/unit and $26.7k/unit.
With a licensing fee of $24.8k/unit, the innovator’s appropriation stream is as presented in
the following table.

Year
($ millions) 1 2 3 4 5 PV (12%)
^t 15:3 21:8 109:0 174:4 239:8 355:5
All of these values, as should be expected, fall within the innovator’s competitive ap-
propriation range. What is not yet clear is whether a $24.8k/unit fee results in annual
distributions of value (i.e., among all the agents) that satisfy competitive requirements (1)
and (2). If not, the innovator will need to fashion some other licensing arrangement in order

23
to be consistent with competition. The annual distributions implied by a licensing fee of
$24.8k/unit are presented in the following table. It is easy to show that these satisfy all the
constraints implied by (1) and (2).

Distributions of Value by Year ($ millions)

Year Firm OEM A OEM B OEM C OEM D Total
1 15:3 14:5 9:1 5:5 1:8 46:2
2 21:8 20:8 13:0 7:8 2:6 66:0
3 109:1 103:9 65:0 39:0 13:0 330:0
4 174:6 166:3 103:9 62:4 20:8 528:0
5 240:1 228:7 142:9 85:8 28:6 726:0

Therefore, the proposed licensing fee of $24.8k/unit is consistent with competition and results
in a net present value for the project of ($000,000): 355:5 100:0 = 255:5:
The estimation of actual value streams (as shown in the preceding two tables) is based
upon management’s speci…c judgment regarding its ability to negotiate licensing fees with
these four OEMs. Of course, the …gures in the preceding table do imply a AF. This is
calculated as in equation (??). For example, in year 1,

15:3 10:5
1 = = 0:8:
16:5 10:5

As it turns out, the innovator’s implied VF is constant over the …ve year period.

4.3.3 Experimentation

The preceding analysis indicates that commercialization should proceed. One question that
can be asked is whether anything can be done to enhance the project; i.e., are there initiatives
available to the innovator that will improve its competitive position and, thereby, increase its
appropriation stream? The following tables summarize the shadow prices, by year, for the
innovator’s minimum and maximum appropriations levels, respectively (groups with shadow

24
prices of zero in every year are omitted).

Shadow Prices on Minima

Group Year

Firm A B C D 1 2 3 4 5

1 1 1 1 1 1:0 0:5 2:0 2:0 0:5

1 1 1 1 0 0:5 0 1:0 1:0 0
1 1 1 0 1 0 0 0 1:0 0
1 1 1 0 0 0 0:5 0 0 0
1 1 0 1 0 0 0:5 0 0 0
1 1 0 0 1 0:5 0 1:0 0 0:5
1 0 1 1 1 0 0 1:0 0 0
1 0 1 0 0 0:5 0:5 0 0 0:5
1 0 0 1 1 0:5 0 0 1:0 0
1 0 0 1 0 0 0 0 0 0:5

Shadow Prices on Maxima

Group Year

I A B C D 1 2 3 4 5

1 1 1 1 1 1:0 1:0 1:0 1:0 1:0

0 1 1 1 0 0 0 1:0 0 1:0
0 1 1 0 1 0 0 0 1:0 0
0 1 0 0 0 1:0 1:0 0 0 0
0 0 1 0 1 1:0 0 0 0 0
0 0 1 0 0 0 1:0 0 0 0
0 0 0 1 0 1:0 1:0 0 1:0 0
0 0 0 0 1 0 1:0 1:0 0 1:0
The general patterns in the shadow prices are consistent with the theory: the innovator’s
minimum is increased by increasing the tension between the available value and the value
producible by groups including the innovator; the innovator’s maximum is increased by
decreasing the tension between the available value and the value producible by groups that
do not include the innovator. Thus, any initiative that tends to increase the value of groups

25
including the innovator while decreasing the value of those without it should, provided the
overall e¤ect on the available value is not too great one way or the other, increase both the
maxima and minima.
Given these observations and the nature of the new technology, management considers a
marketing campaign aimed at the OEMs’retail customers. The idea is to establish a brand
identity for the new technology in a fashion similar to the marketing programs by Intel and
Nutrasweet (i.e., market the quality of a key product component or, in this case, a key
technology).
After some analysis, management concludes that a sustained, properly funded marketing
campaign –$60 million in present value outlays –will have the e¤ect of increasing the share
attracted by OEMs using the technology from those not using it from 20% to 40%. Factoring
this change into the calculations for the innovator’s min, max and anticipated appropriation
levels yields

Year
($ millions) 1 2 3 4 5 PV (12%)
min
t 21:0 30:0 150:0 240:0 330:0 489:2
max
t 21:0 30:0 150:0 240:0 330:0 489:2
^t 21:0 30:0 150:0 240:0 330:0 489:2
There are two striking features of this result: 1) the marketing campaign succeeds in
substantial increases in both the minimum and maximum appropriation levels consistent
with competition; and, 2) under this plan, bargaining ( ) ceases to play a role. Previously,
the innovator indicated a fairly high degree of con…dence in its ability to appropriate on the
high end of its competitive range, with its estimated present value of $355.5 much closer to the
maximum of $383.2 then the minimum of $244.6. Of course, con…dence notwithstanding,
the actual future outcome of price negotiations with the OEMs is uncertain. Under this
marketing plan, the uncertainty surrounding the innovator’s true AF is essentially eliminated,
a potentially signi…cant side bene…t of implementing the marketing campaign. The di¤erence
in anticipated present value between the original commercialization plan and the one with
enhanced marketing is 489:2 355:5 = 133:7. Given the estimated investment of $60m
required to achieve this level of appropriation, the marketing enhancement appears to be a

26
very promising course of action.

5 Conclusion
Coalitional games have proved to be a useful framework within which general principles gov-
min max
erning value appropriation can be developed, e.g., the factors determining and :
The framework has also shown power as a tool for assessing speci…c assertions about value
appropriation, e.g., whether competition assures appropriation for the owner of an unique,
valuable and inimitable resource. The goal of this paper was to show how the same frame-
work can be employed as a decision making tool. We discussed how the ingredients the theory
identi…es as important can be quanti…ed, described the relevant calculations, and then illus-
trated with an application to a commercialization of technology decision. Throughout, we
emphasized both the iterative process required to ensure the framework is being employed
as e¤ectively as possible, and the way the framework and its calculations focus the decision
maker’s attention on opportunities for improvement.

27
References
[1] Besanko, D., D. Dranove, M. Shanley and S. Schaefer (2004), Economics of Strategy,
4th Ed., Wiley: New York.

[2] Brandenburger, A. and B. Nalebu¤ (1996), Coopetition, Harper Collins: New York.

[3] Brandenburger, A. and H. Stuart (1996), “Value-based Business Strategy,” Journal of

Economics & Management Strategy 5: 5-24.

[4] Brandenburger, A. and H. Stuart (2004), “Biform Games,”mimeo., New York Univer-
sity.

[5] Gans, J.S. (2005), Core Economics for Managers, Thomson: Melbourne, forthcoming.

[6] Lippman, S.A. and R.P. Rumelt (2003a), “The Payments Perspective: Micro-
Foundations of Resource Analysis,”Strategic Management Journal, 24: 903-27.

[7] Lippman, S.A. and R.P. Rumelt (2003b), “A Bargaining Perspective on Resource Ad-
vantage,”Strategic Management Journal, 24: 1069-86.

[8] MacDonald, G., and M. D. Ryall (2003), “New Methods for the Identi…cation and
Evaluation of Strategic Initiatives,”Working Paper, Washington University in St. Louis.

[9] MacDonald, G., and M. D. Ryall (2004), “How do value creation and competition
determine whether a …rm appropriates value?”Management Science, 50 (10): 1319-33.

[10] Nelson, R.R. (1957), “Increased Rents from Increased Costs: A Paradox of Value The-
ory,”Journal of Political Economy, 65: 387-93.

[11] Saloner, G., A. Shepard and J. Podolny (2001), Strategic Management, Wiley: New
York.

[12] Stuart, H. (1997), “The Supplier-Firm-Buyer Game and Its M-sided Generalization,”
Mathematical Social Sciences, 34: 21-7.

28