Internal Incentive Structure and the Choice of
Business Strategies
Dirk Sliwkay
Felix Hö- er
March 5, 2015
Abstract
We argue in a game theoretic model that internal incentive considerations
restrict the choice of business strategies. In complex organizations units must
receive incentives to achieve operational excellence in their …elds. However, such
high powered incentives induce units to engage in wasteful in‡uence activities to
bias strategic choices favoring their own operational goals. We show that asymmetric incentive structures where only one unit has high powered incentives are
optimal if the resulting con‡icts are severe. Asymmetric structures reduce con‡ict
and make one unit dominant in the strategy process, favoring "generic" strategies e.g., either cost or quality leadership rather than balancing both objectives.
However, operational excellence of non-dominant units is reduced. Symmetric incentive structures yield "hybrid" strategies balancing all interests but come along
with wasteful haggling.
Key Words: Business strategy, in‡uence costs, incentives, con‡ict, generic
strategies, hybrid strategies.
Department of Economics, Vogelsanger Str. 321, D-50827 Cologne, Tel.: +49 (0) 221 - 27729 206,
email: felix.hoe- [email protected].
y
Department of Business Administration, University of Cologne, D-50931 Köln, Germany, email:
[email protected].
1
1
Introduction
There is a long debate as to what extent “structure follows strategy”(Chandler (1962)),
“strategy follows structure” (the title of Hall and Saias’s (1980) paper, in the …rst
volume of the Strategic Management Journal), or rather each one follows the other as
“one foot follows the other in walking” (Mintzberg (1990)). In this article, we wish
to add an incentive perspective to this debate: We argue that the choice of a …rm’s
incentive scheme, as a core part of the organizational structure, and the …rm’s business
strategy are closely intertwined. The reason is that subunits can engage in wasteful
in‡uence activities to bias the strategy in their own favor, and the internal incentive
structures determine how attractive these in‡uence activities for di¤erent subunits will
be. Hence, decisions on strategy and structure need to be closely coordinated.
To illustrate what we have in mind, consider the following example. A car manufacturer plans a new car series. When the project is presented to the top management, the
head of marketing & sales favors high quality and a fancy design which is tailored to
the …rm’s target customer group. The head of procurement may intervene: This would
require using non-standard and expensive components, which will make the product unpro…table; the product should rather be more standard and less costly. Discussions like
these are quite common in many …rms and may naturally lead to intra-organizational
con‡ict. They can also be viewed as disputes about core parts of a business strategy. If
the top management follows the advice of the marketing & sales unit, it will go towards
a “premium strategy” (high quality-high cost); if it follows the procurement o¢ cer’s
advice, it rather opts for a cost leadership strategy. However, the top management may
also try to compromise and go for an intermediate positioning.
Why do the heads of the two subunits tend to support di¤erent strategic choices?
One important reason is that such advice is in line with their incentive structure. A
sales manager is typically rewarded if sales are large, and a procurement manager if
procurement costs are low. These rewards are meant to provide incentives to deliver
high performance in each manager’s area of responsibility. The higher powered the
incentives for high performance in the subunits are, the more di¢ cult it will be to
make the subunits compromise on what is in the interest of the overall goal. With
steep individual incentives, each unit will thus have a motivation to invest resources in
2
internal haggling and in‡uence activities to tip the top management’s strategy decision
towards their own interest.
For the strategy formation, the role of decentral units and the potential for con‡ict
is well documented. Burgelman (1991), for instance, analyzes the role of autonomous
processes in strategy-making that are initiated by decentral units and acknowledges the
danger of con‡ict between subunits.1 Eisenhardt and Zbaracki (1992) review earlier
literature on the role of politics in the strategy process and conclude that “these studies
and others indicate that politics are common in strategic choice.” (p.26) and moreover
“politics creates animosity, wastes time, disrupts information channels, and ultimately
leads to poor performance.” (p. 27). More recently, Baer, Dirks, and Nickerson (2013)
propose that “heterogeneity in objectives results in team members engaging in political
maneuverings that consume scarce resources (attention, memory, time)” (p. 205).
To gain a better understanding of how attempts to solve the organizational con‡ict
may a¤ect the strategic decisions of …rms, we apply methods from agency theory and
organizational economics. We explore these issues in a formal model of optimal incentive
structure and inter-unit con‡ict. In our model, two unit managers (e.g., a sales manager
and a procurement manager) work for a principal (i.e., the CEO or the owner of the
…rm). They undertake operational e¤orts which a¤ect their unit’s value contribution;
but the value contribution also depends on the strategy choice. Each manager can
invest personal resources in order to in‡uence the …rm’s choice of a business strategy
(for instance, "premium strategy" versus "cost leadership"), which will raise their own
unit’s value contribution at the expense of the other unit. These in‡uence e¤orts
make the units’outputs interdependent and create inter-organizational con‡ict in the
choice of a business strategy.2 We study to what extent these con‡icts can be solved
1
Burgelman and Grove (2007), for instance, describe the con‡ict between groups favoring di¤erent
microprocessor architectures at Intel stating, for example, that “Distinct CISC and RISC camps had
formed and they were competing for the best engineering talent of the company” or even “The battle
between CISC and RISC within Intel had turned into ‘civil war.’“ (pp. 972.)
2
Note that Lawrence and Lorsch (1967a) already state that “di¤ erentiated subsystems often have
quite di¤ erent interests and objectives, so that the resolution of con‡ict between them may well be
the most important function of integrative devices.” (p. 42). In their analysis of strategic renewal,
Floyd and Lane (2000) claim that a strategic role con‡ict is more likely among middle managers as
“functional expertise [..] de…nes their primary organizational roles” (p. 166).
3
by appropriately designed incentive schemes and then derive how the organizational
design a¤ects strategy.
In this context, an optimal incentive scheme has to take care of two problems: First,
the subunits must be motivated to exert high operational e¤ort. This calls for a high
weight on the subunits’own value contributions. Second, the subunits’managers should
not engage too much in wasteful internal in‡uence activities. One way to reduce internal
haggling is to choose an incentive scheme in which a unit’s manager not only bene…ts
from their own value contribution but also from the value contribution of the other unit.
One way to avoid internal haggling is to use pure pro…t sharing arrangements, in which
the managers’ bonus payments only depend on the overall pro…ts of the …rm. This
eliminates con‡icts between the subunits. However, this approach reduces individual
responsibility, adds more noise to the performance measurement, and leads to freeriding with respect to operational e¤orts. The …rm has to account for the trade-o¤
between these e¤ects.
We show that an organization optimally chooses between one of only two speci…c
subtypes of incentive structures. In a symmetric incentive scheme, both subunits’incentive schemes have an identical structure. For each unit, its incentive scheme puts
a larger weight on the value contribution of that particular unit. In turn, both managers’operational e¤orts are high, but both engage to some extent in wasteful in‡uence
activities by trying to in‡uence the strategy in their favor. The resulting symmetry
of in‡uence e¤orts imply that these e¤orts neutralize each other and an intermediate
business strategy is adopted. This is optimal only if internal con‡icts are not too important; in particular, if the interdependency between the subunits is weak or there is
little room for haggling. With respect to the relation between strategy and structure,
this implies that if in‡uence e¤orts are not very e¤ective, …rms will …nd it easiest to
implement (in the language of Porter (1985)) ‘hybrid’business strategies that balance
the interests of di¤erent subunits.
If, however, in‡uence activities are prominent –for instance, if the input of decentral
units in the strategy process is very important – an organization should opt for an
asymmetric solution to suppress wasteful internal haggeling. At …rst sight, one may
consider it optimal to provide symmetric pure pro…t sharing schemes in such situations.
4
This avoids haggling since managers then internalize the full externality on the other
unit. However, as laid out above, this comes at a cost: Each unit is now fully ‘liable’for
the output of the other unit as well, which increases uncertainty and invites free-riding
behavior. Our key result is that such symmetric pure pro…t sharing arrangements are
dominated by an asymmetric incentive structure: To eliminate wasteful in‡uence e¤ort,
it su¢ ces to align (by pro…t sharing) only one of the units with overall company goals.
Since this unit’s manager now has no incentives to engage in any in‡uence activity,
the other unit can implement her favored strategy without internal resistance. This
dominant unit is rewarded based only on its own value contribution. This provides
high powered incentives for its functional excellence. Hence, the total value created
and the strategy choice are both driven to a large extent by this dominant unit. Thus,
we …nd that a “generic business strategy”, i.e., a strategy focusing on the interests of a
particular unit, naturally arises in an environment without any exogenous asymmetry
–just in order to assign clear responsibilities and to avoid internal con‡icts.
Therefore, even when external market conditions and available resources and capabilities may suggest to call for a strategy that balances di¤erent strategic elements,
e.g., customer bene…ts and cost leadership, internal incentive considerations can lead
to the superiority of a strategy in which only one of the dimensions dominates. Hence,
it may be misleading to design a strategy in a two-step procedure by thinking about
the business strategy …rst before the appropriate incentive strategy. Our analysis of the
incentive perspective advises a close co-ordination of both steps.
The remainder of the paper proceeds as follows. After discussing the related literature, section three describes the model. Section four analyzes the unit manager’s
reactions to a given incentive structure. Section …ve derives the optimal incentive structure as the main result of the paper. Section six concludes and discusses implications
for management practice. All mathematical proofs, unless otherwise stated, can be
found in the appendix.
5
2
Related Literature
In line with Chandler’s “structure follows strategy”paradigm, Hall (2004) argues that
it is useful to think of a …rm’s strategy as consisting of two sub-strategies: a business
strategy de…ning how and where to compete and an organizational strategy for the
execution of this business strategy. A key component of an organizational strategy is an
incentive strategy describing the way in which performance is measured and rewarded.
In our analysis, we show that there is also an e¤ect in the opposite direction which
needs to be considered: The set of available business strategies may be restricted by
the way in which these strategies can be supported by an appropriate internal incentive
structure.
The reason for this is that any organization has to account for the problem of
"in‡uence activities" (Milgrom and Roberts (1988)) that give rise to internal con‡icts,
"policies" and "political maneuvering" in the strategy process, as stressed by Burgelman
(1991), Eisenhardt and Zbaracki (1992), Burgelman and Grove (2007), or Baer, Dirks,
and Nickerson (2013).
The idea that in‡uence costs can lead to one unit “living at the expense of another
unit”is illustrated, e.g., in Kumar (2013), who shows that units which are successful in
internal discussions improve their own results (in our model: their "value contribution")
at the expense of other units. Bidwell (2012) provides evidence that important strategic
decisions, like outsourcing of important business processes, are often better explained by
the individual agenda of unit managers than by the overall …rm pro…tability. We add to
this line of literature that the importance of wasteful con‡ict is usually not exogenously
determined but can well be the result of how the internal incentive structure is designed.
The idea that subunits have con‡icting interests that are not naturally aligned with
the overall goals has long been discussed in the literature on organizational behavior (Simon (1962), Mintzberg (1980), Siggelkow and Levinthal (2003), Rivkin and Siggelkow
(2003)). As already argued by Lawrence and Lorsch (1967a,1967b), the need for specialization generates at the same time a need for integration and for di¤erentiation, the
latter meaning, in their terminology, the segmentation of an organization in subsystems
which tend to develop their own goals and attributes, shaped by the speci…c needs of the
environment. But this comes along with a need for integration to achieve some “unity
6
of e¤ort” in order to accomplish the organization’s task. But, as Lawrence and Lorsch
already claim, there seems to be a natural tension between those objectives which “are
essentially antagonistic, and [..] one can be obtained only at the expense of the other”
(Lawrence and Lorsch (1967a), p. 47).
Instruments to align the behavior of subunits with the organizational goals are
widely discussed in the literature. For instance, organizations may set up pro…t sharing plans or grant stock options in order to align the individual units’interests with a
common goal (Wageman and Baker (1999), Kretschmer and Puranam (2008)). Alternatively, …rms may invest in creating a common culture of shared values and, hence, create
forms of “normative integration” (see, e.g., Van Maanen and Schein (1977), Ghoshal
and Nohria (1989)). If this component of collective joint alignment has a strong weight,
managers should be more inclined to compromise on the strategic decision and …nd a
common goal. We add to this view that, if incentives for joint outcomes are particularly
large, this may weaken personal responsibilities for the particular sub-objectives of the
di¤erent units.
To focus on the e¤ects of internal organization on strategy, we deliberately make
the extreme assumption of perfect symmetry ex ante. However, of course, additional
factors and asymmetries are also important. If units are not symmetric ex ante but
may exhibit di¤erent productivities in the di¤erent strategic dimensions, the degree of
productivity di¤erences will contribute to explaining whether (and if so, how) …rms
specialize into one dimension or choose a ‘hybrid’strategy (according to the resourcebased approach, Barney (1991); for a recent empirical survey, see Crook, Ketchen Jr,
Combs, and Todd (2008)). We show that “strategic purity”may arise even in an ex-ante
symmetric environment and can be driven by internal concerns alone.
Since in our model the subunits’ managers in‡uence activities can drive strategy
choices either to extreme (like in the car manufacturer example, cost versus quality
leadership) or to intermediate outcomes, our paper is also related to the discussion on
“generic” versus “hybrid” business strategies in the tradition of Porter. Porter (1985)
advises against intermediate or “hybrid”business strategies, as …rms should avoid getting ‘stuck in the middle’: “Becoming stuck in the middle is often a manifestation of
a …rm’s unwillingness to make choices about how to compete. It tries for competitive
7
advantage through every means and achieves none, because achieving di¤erent types of
competitive advantage usually requires inconsistent actions”.3 However, there seems to
be no clear empirical result as to whether generic business strategies may indeed be
superior to hybrid strategies. While Campbell-Hunt (2000) …nd only weak evidence,
Thornhill and White (2007) detect a signi…cant positive relationship between strategic
purity and performance. However, they also show that the bene…ts of a pure relative to
a hybrid business strategy di¤er between industries. Indeed, our model does not predict
that a pure business strategy is always superior. If interdependencies and the need for
coordination between units are weak, then hybrid strategies may dominate pure strategies. A key advantage of hybrid strategies is that operational excellence is achieved in
both dimensions, while an incentive structure implementing a generic strategy leads to
lower operational performance in one dimension.
Our paper is also related to and builds on ideas that have appeared in the literature
in organizational economics analyzing in‡uence activities.4 Our modeling approach
on this idea integrates Tullock (1980)-type rent-seeking contests5 with multitasking
incentives (Holmström and Milgrom (1991)). To the best of our knowledge, our paper
is the …rst to study how incentive schemes govern intra-organizational con‡ict in a
multitasking environment and thus a¤ect the choice of business strategies.
Potential bene…ts of narrow business strategies have been analyzed in formal economic models by Rotemberg and Saloner (1994), Corts (2000), Rotemberg and Saloner
(2000), or Van den Steen (2005). Of this literature, the most closely related contribution is probably Rotemberg and Saloner (1995), who also study the connection between
in‡uence activities and business strategy in a formal model but do not study the role of
incentive schemes. In their model, a di¤erence in objectives arises because of di¤erences
in speci…c human capital; whereas in our model, di¤erent interests arise because of the
3
See Porter (1985), page 17. See also Treacy and Wiersema (1993) who distinguish operational
excellence, product leadership, and customer intimacy and claim that “companies that have taken
leadership positions [..] typically have done so by narrowing their business focus, not broadening it”.
4
Applications include Meyer, Milgrom, and Roberts (1992) on divestitures as a means to avoid
in‡uence activities in declining organizations, Schaefer (1998) on resistance to organizational change,
Scharfstein and Stein (2000) on the misallocation of capital within …rms, Inderst, Müller, and Wärneryd
(2007) on divisionalization, or Münster and Staal (2010) on the choice of multi-tiered structures.
5
See Konrad (2009), pp. 9, or Congleton, Konrad, and Hillman (2008) for recent surveys on the
literature on rent seeking contests.
8
need to provide incentives for the di¤erent operational tasks of the subunits. They argue that the choice of a generic business strategy is a commitment device to implement
a certain structure and thus avoids con‡ict. Hence, in this sense their contribution is
in the tradition of “structure follows strategy” while our study endogenizes the strategy choice and shows that the nature of the incentive structure a¤ects the choice of a
strategy.
3
A Formal Model
To analyze how internal con‡icts between subunits may a¤ect the relation between
strategy and structure, we propose a formal model. The model should capture a situation in which a …rm has to make an important decision on its strategic position, e.g.,
either to go for a low-cost, no-frills strategy or to opt for high quality and high service
levels. This may be thought of as a decision on the design of one of the …rm’s core
products but may also be a general strategy decision. Our key point is that di¤erent
units in the …rm are a¤ected by the strategic alternatives di¤erently. Since the units
and their managers will be a¤ected by the decision, they will undertake e¤orts to manipulate the strategy discussion in their favor. The owners of the …rm, who themselves
have no preference concerning the strategy decision, have to take these e¤ects into account when deciding on the incentive schemes proposed to the managers, which in turn
will in‡uence how strongly the managers want to manipulate the strategy discussion in
their individual favor. Therefore, by designing the incentive schemes of the managers,
the owners will also a¤ect the outcome of the strategy discussion.
We consider a …rm with two functional departments (units). Each unit employs
a manager i = 1; 2. For instance, manager 1 could be responsible for the sales and
manager 2 responsible for procurement activities. Both managers can exert e¤ort in
two di¤erent dimensions. First, they can work hard to increase their operational performance by choosing operational e¤ort ei 0; and second they can choose an ‘in‡uence
e¤ort’ di
0 to a¤ect the choice of the implemented business strategy. This e¤ort
includes any form of internal “politics” such as “buttering the boss” (Milgrom and
Roberts (1988)) or in line with Allen, Madison, Porter, Renwick, and Mayes (1979),
9
“intentional acts of in‡uence to enhance or protect the self-interest of individuals or
groups” (p. 77). Both kinds of e¤ort are costly to the agents and not prescribed in a
contract.6
Each manager produces an output Qi , and the principal’s pro…t is the output minus
the wages w1 and w2 : = Q1 + Q2 w1 w2 . When the units are divisions operating
in di¤erent markets, then Q1 and Q2 are simply the pro…ts of the two divisions. When
the units are di¤erent functions, one could think of Q1 as the value contribution of
the sales department and Q2 the value contribution of the production department.
Additive separability naturally arises for instance when demand (quantity) is rather
inelastic and the sales department’s e¤orts increase the price, and the e¤orts of the
production department reduce the costs. Since our interest is to investigate whether an
asymmetric strategic choice can arise in an a-priori completely symmetric environment,
we abstract from any asymmetries between the managers.
While operational e¤ort is speci…c to each kind of output, in‡uence e¤orts exhibit
externalities. The higher the e¤ort of manager 1 in this dimension is, the lower the
expected output of manager 2 will be, and vice versa. Think for instance of the sales
manager lobbying internally for the inclusion of high quality features in a new product,
implying higher cost, while the production department lobbies for using only standard
items which may limit the attractiveness to buyers and therefore lowers sales success.
These in‡uence activities a¤ect the business strategy adopted by the company. We
model this as a standard Tullock (1980) contest; i.e., from agent i’s perspective, the
resulting business strategy Di is given by7
Di =
(
di
di +dj
1
2
if di > 0 _ dj > 0;
if
di = dj = 0:
such that D1 + D2 = 1. Agent 1’s most preferred business strategy is D1 = 1, and the
most preferred decision of agent 2 is D2 = 1 (implying D1 = 0). We assume that if
one agent strictly prefers not to exert in‡uence e¤orts such that di = 0 while the other
6
In the language of microeconomic contract theory, they may be observable but unveri…able such
that they cannot directly be speci…ed in an enforcable contract.
7
See Konrad (2009) for a recent comprehensive survey on the theory of contests. Section 2.3 includes
a detailed discussion of the Tullock contest.
10
agent j strictly prefers to tip the strategy decision in the direction of her own most
preferred realization –which she can do by exerting an arbitrarily small e¤ort dj –then
agent j’s preferred strategy decision is implemented at zero e¤ort cost for this agent.8
Each agent’s value contribution is given by:
p
Qi = q ei + z Di + "i :
The …rst term is the contribution from the operational e¤ort, where q > 0 determines
the size of the impact of additional e¤orts on the created value in the unit. The last
term is a random productivity shock, where "i is a normally distributed i.i.d. random
variable with variance 2 . The middle term re‡ects how an agent can increase her value
contribution by exerting in‡uence activities to a¤ect the strategic decision in her favor.
The larger the in‡uence e¤ort di is, the larger Di will be. We speak of Di as the adopted
business strategy as it determines the relative importance of the value contributions of
both units. Using our above example, the endogenously chosen Di determines which
part of additional pro…ts are either created by excellence in marketing/sales or in manufacturing/procurement.9 Note that the parameter z > 0 is very important in our
model since it determines to what extent the adopted overall business strategy a¤ects
the individual value contributions. Therefore, z measures the degree of interdependence
between both units’actions. If z is large, the interdependency between both units is
large, and it becomes more attractive to invest in‡uence e¤ort to a¤ect the strategy
decision at the expense of the other unit. If z is small, interdependencies are weak, and
a unit’s value contribution is basically determined by its operational excellence.
For the managers’ utility, we assume that they have a CARA utility function,
u (wi ) = exp ( r wi ), with an Arrow-Pratt measure of constant absolute risk aversion r.10 Note that uncertainty and risk aversion induce agency costs in the model.
8
This is to avoid the situation in which agent j 0 s optimization problem has no solution. Alternatively, one could think of the e¤ort as chosen from the set: dj = f0; dj " > 0g ; where " is arbitrarily
small, such that we can neglect the cost of this e¤ort.
9
In this respect, if unit i is the marketing/sales department, Di = 0 may imply a cost leadership
and Di = 1 a quality leadership strategy.
10
The key advantage of the CARA utility function is its tractability since the agent’s decision
problem boils down to the maximization of certainty equivalents, which are here equal to the expected
wages minus a risk premium, which linearly depends on the variance of the wage (see, for instance,
11
The reason is that performance contingent wages come along with income uncertainty,
which risk averse agents dislike and, hence, the principal cannot implement high powered incentives at no costs.
The disutility of e¤ort for both the operational and the lobbying e¤orts are equal to
the e¤ort levels ei and di , respectively. Both agents’value contributions are veri…able
and the managers’compensation schemes are linear in the value contributions11
wi =
i
+
ii Qi
+
ij Qj :
Here, ai re‡ects a …xed payment and ii Qi
ij Qj is a variable payment depending
on the own value contribution (the value contribution of the other unit). The outside
option of both agents are equal to wo .
p
The …rst-best operational e¤ort is obtained by maximizing q ei ei , implying
2
eFi B = q4 : Since Di + Dj = 1 is a constant, both agents should abstain from any
in‡uence e¤orts, i.e. dFi B = 0. Another implication of Di + Dj being constant is that
the principal ex-ante has no direct preferences towards a speci…c business strategy.
Before we start with the formal analysis, it is useful to introduce some formal de…nitions or conventions of wording which we use in the further analysis of the model. In the
spirit of Porter, we speak of a „generic business strategy“ when Di is either exactly 0 or
exactly 1; such that one unit entirely dominates the strategy choice. “Hybrid business
strategies” are those in which 0 < Di < 1. Moreover, a “pure pro…t sharing” arrangement is a contract in which ii = ij such that a manager’s wage depends only on the
overall pro…ts of the …rm. In contrast, a contract provides “individualized incentives”
if ii > ij such that a manager bene…ts more from his own unit’s success than from
that of the other unit.
(Wolfstetter, 2002, p. 342) or (Milgrom and Roberts, 1992, pp. 210)). Hence, in other words, the
CARA utility implies tractable mean-variance preferences.
11
Hence, our model employs the Holmström/Milgrom type contracting framework (Holmström and
Milgrom (1987), Holmström and Milgrom (1991)).
12
4
Equilibrium E¤ort Choices
In this section, we take the incentive scheme for the managers as given. An incentive
scheme consists of a …xed payment and variable payments. Variable payments are
conditional on the success of the manager’s own unit; as well as on the total …rm pro…ts,
i.e., also on the performance of the other manager. Given his incentive scheme, each
manager chooses his operational e¤ort and his in‡uence e¤ort in order to maximize his
overall utility. We show that a unique equilibrium in the simultaneous e¤ort decisions
of both agents exists, and we characterize its properties. In particular, we show that,
depending on the structure of the incentive scheme, either both managers exert in‡uence
e¤orts or neither of them. Moreover, we show that an incentive scheme based on overall
pro…ts rather than on unit success eliminates the incentives for in‡uence activities.
Agent i maximizes her certainty equivalent given a wage scheme ii ; ij :12
max
ei ;di
i
+
ii
p
q ei + z
di
di + dj
+
ij
p
q ej + z
dj
di + dj
ei
di
1
r
2
2
2
ii
+
2
ij
:
(1)
The agents thus choose an optimal operational e¤ort
ei =
2 2
ii q
4
:
(2)
The …rst derivative of the agent’s objective function (1) with respect to the in‡uence
e¤ort di is:
dj
1;
(3)
z
ii
ij
(di + dj )2
which is strictly negative if ii < ij : In this case, the agents would not exert any
in‡uence e¤ort but choose the lowest possible e¤ort level di = 0.
For expositional purposes, it is useful to think of i = ii
ij as the “individualized
incentive” in agent i’s compensation scheme. A larger i implies that the agent’s
incentives have a stronger focus on his own value contribution. A small value of i
implies strong “cross incentives“ since a lot of weight in the agent’s payment scheme
12
Note that here we use the result that for normally distributed wages and exponential utility, the
certainty equivalent of an agent’s utility from wage w is equal to E [w] 2r V ar [w].
13
is attached to the value contribution of the other agent. A smaller i makes choosing
high lobbying e¤orts unattractive since a damage to the other agent’s output has a
larger e¤ect on the own wage. It is important to note that an incentive scheme which
has the property ii = ij (i.e., i = 0) eliminates the incentive for agent i to exert
lobbying e¤orts, since he then fully internalizes the externality caused by his in‡uence
activities. However, such an incentive scheme may not be optimal to induce appropriate
operational incentives.
When solving for the equilibrium outcome in lobbying e¤orts, we can restrict attention to the case i
0:13 Thus, for ii > ij ; note that from setting (3) equal to
p
zero and solving for di we obtain the reaction function di = z i dj dj : Since the
second derivative of (1) is negative for i > 0; there will be an interior best response if
the …rst derivative of (3) with respect to di is strictly positive at di = 0; which holds if
dj < z i . Thus, the agents’reaction functions are given by:
di (dj ) =
( p
z
i dj
dj if dj z i ;
if otherwise:
0
(4)
For positive choices of di ; the best response is increasing in z and is concave in the
in‡uence activity chosen by the other manager. We can derive the following result:
Proposition 1 (i) If i > 0 and j > 0, there is a unique equilibrium in pure strategies in which the in‡uence e¤ort choices of the agents are
di = z
2
i
(
i
+
j
2
j)
for i = 1; 2
and the strategy decision is Di = i +i j . (ii) If i ; j = 0; both will choose di =
dj = 0 and Di = 12 . (iii) If i > 0 = j ; agent i’s preferred strategy decision will be
implemented, i.e., Di = 1.
13
This is implied by the optimization of the principal. From (3); we know that ii < ij implies
di = 0: Thus, the principal could always reduce ij without changing the in‡uence e¤ort (which is still
zero), while leaving the operational e¤ort una¤ected (since this depends on ii only). However, a lower
ij makes the principal strictly better o¤ (see the discussion of Lemma 2). Therefore, in equilibrium,
contracts with ii < ij never occur, and we can restrict attention to non-negative i :
14
If both i are positive, i.e., if both agents’wages depend to a stronger extent on
their own rather than on the other unit’s signal, both agents exert in‡uence e¤orts. An
agent’s in‡uence e¤ort increases in the level of interdependency z and in the degree of his
own individualized incentive i . It decreases (increases) in j whenever i < j ( i >
j ). The resulting strategic decision is just a weighted average of the individualized
incentives.
This result points to a key danger in providing high powered operational incentives:
A large individualized incentive i (i.e., if ii is substantially larger than ij ) implies
that the unit managers invest high e¤orts in increasing their own value contribution.
On the one hand, this is of course good since this leads to higher operational e¤orts;
however, this also induces a higher incentive to engage in in‡uence activities which are
harmful from the organizational perspective. This e¤ect may re‡ect a part of Porter’s
claim that “employees, who are urged to seek every possible source of improvement,
often lack a vision of the whole and the perspective to recognize trade-o¤s.”14
If, however, the individualized incentives i in our model are zero, compensation
schemes boil down to pure pro…t sharing schemes. Agents then have no incentive to
engage in any in‡uence activities as they are fully aligned with the interests of the
organization.
Finally, if the individualized incentive for agent i is strictly positive and zero for
agent j, agent j strictly prefers dj = 0 to any positive e¤ort, using the same argument
as before. Agent i, however, strictly prefers to tip the business strategy in her own
favor due to i > 0 and can do so by exerting an in…nitesimally small level of in‡uence
e¤orts at negligible costs. Hence, a key insight of the analysis in this subsection is
that in‡uence activities can be avoided if the incentive scheme aligns only one unit’s
objectives with company pro…ts. As this unit no longer has incentives to invest in
haggling, the other unit does not need to invest signi…cant resources to in‡uence the
strategy choice in its favor.
14
See Porter (1996), p. 75.
15
5
Designing Optimal Incentives
The …rm owners want to design incentive contracts in order to induce managers to
work hard towards maximizing their unit’s value contributions. However, they must
take into account that each manager can increase its value contribution at the cost of
the other unit’s contribution by manipulating the …rm strategy in its own favor. In
this section, we derive the optimal incentive scheme, taking into account the previously
derived reactions of the managers. From the previous analysis, we know that we can
restrict attention to schemes that either induce in‡uence e¤orts by both managers or
by neither of the two. After characterizing the most pro…table incentive schemes for
each of these two cases separately, we identify the globally optimal scheme.
In particular, we show that only two speci…c types of incentive schemes can be
optimal, and the one that dominates depends on the interdependence between the two
units. When the interdependence is weak, it is optimal to design symmetric incentives
that reward performance of the individual subunits more than overall pro…ts. Under
such a scheme both managers exert in‡uence e¤orts, which results in the choice of
an intermediate business strategy. When the interdependencies are strong, however,
optimal incentives are asymmetric and lead to one unit dominating the …rm strategy
while the manager of the other unit is rewarded based only on overall pro…ts, which
implies lower powered incentives. Internal in‡uence activities are avoided and a “generic
business strategy” will be implemented which follows the preferences of the dominant
unit.
Formally, the principal determines the pro…t maximizing contract, i.e., the optimum
weights 11 ; 12 ; 22 ; and 21 , anticipating the result of the in‡uence e¤ort game. The
basic trade-o¤ faced by the principal is that she can not simultaneously provide high
powered incentives to exert productive e¤ort and prevent the agents from exerting
wasteful in‡uence e¤orts. To achieve the former, the principal would have to remunerate
the agent only according to her own value contribution; but this would lead the agent
to try to in‡ate their own output at the expense of the other agent’s output by exerting
higher in‡uence e¤orts.
Providing strong “cross incentives”, i.e., choosing low values of the i ; counteracts
this e¤ect. If the own salary positively depends on the other unit’s output, in‡uence
16
activities become less attractive. In the extreme, at which i = 0; the incentives for
wasteful internal haggling vanish entirely, which is equivalent to remunerating the agents
only with respect to the …rm gross pro…t, Q1 + Q2 : However, strong cross incentives
are also costly as they increase the agents’risk exposure as compensation is then also
a¤ected by noise in the other unit’s value contribution. Since the agents are risk averse,
this leads to higher risk premia and thus induces higher agency costs.
From Proposition 1, we know that there are either contracts in which in‡uence
activities occur in equilibrium (if i > 0 and j > 0) or contracts that avoid all
in‡uence activities (if at least one i = 0). To understand the logic of the model, it is
useful to compute optimal contracts for these two cases separately and then compare
the pro…ts under both regimes.
5.1
Contracts that do not eliminate in‡uence e¤orts
We start by investigating optimal incentive schemes in the class of contracts that do
not eliminate in‡uence activities (i.e., 1 ; 2 > 0).
The agents’ participation constraints require that their certainty equivalents (1)
exceed their reservation wages. The principal also has to take the incentive constraints
(2) into account, as well as the in‡uence e¤ort constraints as derived in Proposition
1. When we insert the binding participation constraints into the principal’s objective
function, substitute 12 = 11
1 and 21 = 22
2 ; and already use the optimal
operational e¤ort choices of the agents’from (2), the principal’s optimization problem
becomes:
11 ;
2 2
11 q
2
s
2
11
+(
11
max
22 ;
1
r
2
2
1;
q
4
+q
s
2
1)
This function is strictly concave in
ii
+
2 2
22 q
4
2
22
2 2
11 q
+z
+(
4
22
11
and
=
q 2 + 2r 2 i
:
q 2 + 4r 2
17
22 ;
2 2
22 q
4
2
2)
z
1
(
1+
2
2)
(5)
2wo
and optimality requires
(6)
Thus, we can restate the principal’s problem (5) as a problem of choosing the optimal
values of i and j . While the optimization problem is symmetric, it may well be that
the solution is asymmetric. However, the following lemma shows that we can restrict
attention to the class of symmetric contracts as long as 1 and 2 are strictly positive.
Lemma 1 There is no asymmetric interior pro…t maximum, i.e., it is never optimal
to set 1 6= 2 if 1 ; 2 > 0:
Hence, the only candidate for an interior solution is a symmetric contract in which
1 =
2 . It is therefore instructive to …rst restrict the analysis to the class of symmetric
contracts and seek the optimal contract within this class. In the next subsection, we
derive conditions under which such a contract is also globally optimal. By solving the
principal’s optimization problem for the case 1 = 2 , we obtain the following result:
Proposition 2 There is a cut-o¤ value z =
symmetric contract is
ii
=
4r 2 q 2
q 2 +4r 2
2q 2 z
and
2 (2r 2 + q 2 )
ij
such that for z
=
z, the optimal
z
;
4r 2
otherwise it is a pure pro…t sharing contract with
ii
=
ij
=
q2
q 2 + 4r
2
:
Hence, if the interdependency between both units, as measured by z; is su¢ ciently
small, the optimal symmetric contract will indeed entail some in‡uence e¤ort. Clearly,
if z is close to zero, in‡uence e¤ort considerations will not matter strongly for contract
choice, and incentives are based on the agents’ own value contribution. But as z increases, the power of operational incentives ii decreases while pro…t sharing via ij
is extended and the incentive scheme becomes more team-based. The reason is that
for higher values of z, there is a higher interdependency between the units, and the
agents have higher incentives to spend more e¤orts on wasteful in‡uence. As in‡uence
e¤ort is more attractive the higher powered the operational incentives are, higher values
of z increase the costs of implementing operational e¤orts. In turn, these operational
18
incentives are weaker in the optimal contract. Although pro…t sharing can reduce in‡uence e¤ort and therefore increases with z, this comes at a cost since agents have to
bear more uncertainty in their compensation packages. When z becomes very large,
the in‡uence e¤ort problem becomes so severe that the principal will choose a pure
pro…t sharing scheme when restricted to a symmetric contract. Such a pro…t sharing
scheme eliminates all in‡uence e¤ort activities but leads to substantially lower powered
incentives.
These considerations have important further implications. Whenever the principal
is, for some reason, restricted to use symmetric contracts, she never remunerates the
agents based solely on the individual performance, i.e., ij > 0: This happens despite the
fact that individual performance is perfectly observable and contractible. Introducing
(costly) noise into the incentive scheme by using team-based incentives is meant only
to discourage wasteful internal lobbying. This suggests an important caveat in the
design of incentive schemes for situations in which individual value contributions are
easy to measure. Even in this case, one should be reluctant to have high powered
individualized incentives. Before introducing such high powered incentive schemes, it
should be checked whether there is opportunity for wasteful internal in‡uence e¤ort
which would call for team-based elements. In a sense, this result re‡ects the argument
put forward, for instance, by Teece et al. (1997, p. 517) that “the essence of internal
organization is that it is a domain of unleveraged or low-powered incentives [..] that
rewards are determined at the group or organization level, not primarily at the individual
level, in an e¤ort to encourage team behavior, not individual behavior.”
Finally, it is interesting to note that the e¤ect of uncertainty on
= ii
ij ,
the degree to which the incentive scheme is ‘individualized’, is ambiguous and depends
on z. First note that both ii and ij are decreasing in r 2 . But for small values of
z; the marginal reduction of ij is smaller than that of ii ; and in turn
decreases
2
15
in r , i.e., the incentive scheme becomes more team based. The reason is that for
small values of z, the pro…t sharing incentives ij are rather small compared to the
operational incentives ii ; but as risk premia are convex in the slope parameters ; a
marginal reduction of ii leads to a stronger reduction in risk premia than a marginal
15
This is the case when z <
8r 2
8r 2
4 +4r
4 2
q
2 q 2 +q 4
.
19
reduction of ij . The marginal costs of implementing operational e¤orts thus increase
to a stronger extent than the marginal costs of avoiding in‡uence e¤ort. For large values
of z; the opposite is true: The marginal reduction of ij is larger than that of ii , and in
turn increases in r 2 ; i.e., the incentive scheme becomes more individualized. To see
why this is the case, note that when z is high, in‡uence e¤ort is a substantial problem.
A large emphasis is put on pro…t sharing and, in turn, additional risk premia from
these collective incentives are high. An increase in uncertainty then has a substantial
e¤ect on the marginal costs of pro…t sharing. Hence, when z is su¢ ciently high, the
marginal costs of increasing pro…t sharing incentives increase to a stronger extent than
the marginal costs of inducing operational incentives.
5.2
Contracts without in‡uence e¤ort
So far, we have focused on contracts in which both agents are treated equally. As shown,
any globally optimal contract in which there is still in‡uence e¤ort in equilibrium must
be a symmetric contract. But we have also shown that for large values of z, optimal
symmetric contracts become pure pro…t sharing schemes. However, as we already know
from Proposition 1, the costs of in‡uence e¤ort will vanish if only one of the agents
receives a pure pro…t sharing scheme (i.e. 2 = 22
21 = 0). Such an agent has
no incentive to engage in lobbying since he entirely internalizes all externalities that
in‡uence e¤ort would impose on the other unit. In turn, the other agent does not need
to spend resources on in‡uence e¤ort as he can tip the business strategy in his favor
at negligible costs. More intuitively, in any dispute about the …rm’s strategy, agent 2
would be indi¤erent about the outcome while agent 1 has a clear preference for D1 = 1,
which is then implemented.
This has an important implication:
Lemma 2 It can never be optimal to choose an incentive scheme in which
i
=
j
= 0.
The proof is straightforward: Suppose that such a contract would be optimal and
in turn 11 = 12 and 22 = 21 . Then the principal can always increase pro…ts, for
instance, by leaving 11 ; 12 ; and 22 unchanged and reducing 21 to zero. Such a change
does not a¤ect operational incentives, and there is still no in‡uence e¤ort. But agent
20
2’s compensation package entails a lower uncertainty, and in turn the risk premium that
has to be paid in order to satisfy his participation constraint will be reduced. In other
words, the drawback of a compensation structure in which both agents’incentives are
based on the …rm’s pro…ts is that it makes each agent ‘liable’ for the total pro…t of
the …rm. This is too costly since it su¢ ces to only make one agent liable for pro…ts to
eliminate in‡uence e¤ort incentives for both of them. Hence, the symmetric pure pro…t
sharing contracts, which are optimal within the class of all symmetric contracts when
2 q2
z > z = q4r
2 +4r 2 (as we have shown in Proposition 2), cannot be globally optimal.
Since we have ruled out the optimality of any asymmetric contract in which both
1 and
2 > 0 in Lemma 1, the only remaining candidates are contracts in which one
agent receives a pure pro…t sharing scheme, i.e., without loss of generality 2 = 0. In
that case, the principal’s optimization problem reduces to:
max
11 ; 22 ; 12
q
s
2 2
11 q
4
+q
s
2 2
22 q
4
2 2
11 q
+z
4
Since
@
@
=
r
2 2
22 q
2
12
12
1
r
2
4
(
0 if
> 0 if
12
12
2
0
< 0;
2
11
+
2
12
+2
2
22
2wo :
(7)
it is always optimal to choose 12 = 0: As agent 2 has no incentive to exert in‡uence
e¤orts, there is no necessity for agent 1 to invest in in‡uence activities in order to get
his preferred strategy implemented, and therefore his incentive scheme can be based
solely on his own value contribution.
The …rst-order conditions for the remaining parameters require:16
@
@
@
@
q2
2
q2
=
2
=
11
22
11 q
2
r
2
22 q
2
Note that the second derivatives are always negative.
21
11
= 0;
(8)
2
This leads to the following result:
16
2
2r
2
22
= 0:
(9)
Proposition 3 The optimal contract in the class of contracts that eliminate in‡uence
e¤ort entirely is characterized by
ii
=
q2
q 2 + 2r
jj
=
=
ji
2
;
q2
q 2 + 4r
ij
2
= 0 and
(10)
(11)
such that one agent receives high powered operational incentives based solely on her
unit’s performance, and the other is remunerated according to a lower powered pro…t
sharing scheme.
Since in‡uence e¤ort never occurs under such contracts, it is obvious that the optimal wage schedule is independent of z. The scheme becomes higher powered with a
higher q: The more productive operational e¤ort is, the more the contract should be
focused on setting incentives to increase their own output. Finally, a higher relevance of
measurement uncertainty, i.e., high values of r or 2 ; reduces all incentive parameters.
Such an incentive scheme induces high powered operational incentives for one agent
whose compensation does not depend on the other unit’s value contribution. But for the
other agent who receives a pure pro…t sharing scheme, operational incentives are lower
powered (i.e., jj < ii ). Hence, the asymmetry in contracts creates an asymmetry in
the composition of the value created. Therefore, the ‘favored’unit’s strategy is chosen
and in addition this unit’s management exerts a higher operational e¤ort. In this sense,
an asymmetric incentive structure is optimal, under which an asymmetric strategy and
an internal dominance of one unit arises naturally.
5.3
Optimal contracts
It remains to be analyzed under what conditions which type of contract is superior.
We already know that for large values of z, the optimal symmetric contract is a pure
pro…t sharing scheme, which is always dominated by an asymmetric contract. Hence,
for large values of z, the asymmetric solution must be optimal for the principal.
However, if z is su¢ ciently small, then the asymmetric boundary solution will be
dominated by the interior solution, even though this comes along with some in‡uence
22
e¤ort in the induced equilibrium, as shown in the following result:
(2
p
2)2q 2 r 2
Proposition 4 There is a cut-o¤ value zb =
such that for z < zb; the
2
q +4r 2
globally optimal contract induces a symmetric incentive structure, which leads to an
intermediate strategy Di = 12 but entails internal in‡uence e¤ort. For z > zb; however,
the optimal contract induces an asymmetric incentive structure in which the strategy is
determined by one of the units (i.e., Di = 1 or Di = 0) and internal in‡uence e¤ort is
avoided.
The cuto¤ zb – and therefore the parameter range for z in which the symmetric
interior solution is optimal –is large if the e¤ectiveness of operational e¤ort q is large
and if r or 2 are rather small. The reason is that in these cases, providing relatively
strong operational incentives for both agents is attractive for the principal. This implies
that an ‘intermediate’strategy is realized, although this comes along with some internal
haggling.
Otherwise, the asymmetric incentive structure is optimal. The asymmetric solution
has the property that one agent receives a high powered incentive contract, strongly
remunerating her own unit’s value contribution. The other agent receives a wage which
is purely contingent on joint output. In turn, no in‡uence e¤ort occurs, and a ‘generic’
business strategy is chosen. Due to the symmetry of our model, the principal is indi¤erent to which strategy decision is implemented; however, she strictly prefers the
extreme outcomes to any interior position. Figure 1 shows the pro…ts as a function of
1 and
2 for a small and a large value of z; respectively (the other parameters are
r = q = = 1).
We already mentioned in Section 4.2 that in the asymmetric solution, the agent
with the high powered operational incentives exerts higher e¤orts than the colleague
working under the pro…t sharing scheme. It is easy to check that the e¤ort level chosen
by both agents under the optimal symmetric contract always lies between these two
e¤ort levels. Since the e¤orts are directly linked to the outputs Qi ; it is interesting to
see under which contract total e¤ort, and therefore total output Q1 + Q2 ; is larger:
23
Figure 1: Firm pro…ts for di¤erent levels of interdependency
2
2
r
Corollary 1 For zb z < q2q
2 +4r 2 , expected output E [Q1 + Q2 ] in the optimal symmetric contract exceeds the expected output of the optimal asymmetric contract, although
total pro…ts are higher under the asymmetric contract.
The reason is that the symmetric contract requires higher wage expenditures due to
the additional uncertainty imposed on the agents. The higher output is not su¢ cient
to compensate for these higher wage payments. If z becomes very large, the symmetric
contract will also yield lower outputs since it becomes increasingly costly to provide
operational incentives while at the same time curtailing internal haggling.
Finally, note that so far we have assumed that there is no technological reason
to prefer one type of strategy over the other. If the choices of the ei and di were
fully contractible in our model, the principal would be indi¤erent between all strategy
choices since the strategy would simply determine a redistribution of output between
the two units. But is important to note that our key result still holds (namely, that an
asymmetric solution may be optimal) even when the principal has a strict technological
preference for an intermediate strategy. To see this, consider the following very simple
24
extension. Suppose that the principal can still not determine the strategic choice, which
still is the result from the internal haggling between the units. But suppose that the
total value created increases by a constant S if and only if the strategy implemented is
equal to Di = Dj = 12 . Suppose for simplicity that the structure of both units’value
contributions remains unchanged and, hence, the incentive problem does not di¤er to
our previous set-up. It is straightforward to note that the principal would always strictly
prefer the symmetric strategy when the di and ei were contractible. But if this is not
the case, the previous result still holds and an asymmetric solution is chosen if S is not
too large:
Corollary 2 Even if the principal has a direct technological preference for an intermediate business strategy, such that the value created increases by a constant S if and only
if Di = 21 , she will implement an extreme strategy, Di = 0 or Di = 1 if S is not too
large and z is su¢ ciently large.
The reason is very simple: If the interdependency and thus the room for haggling z is
very large, the sum of both agents’value contributions is maximized by an asymmetric
solution as it avoids any in‡uence e¤ort. The larger z is, the larger the loss from
in‡uence e¤ort under a symmetric contract will be. Hence, even a strict technological
bene…t of having a symmetric choice will be outweighed by in‡uence e¤ort costs if z is
very large.
6
Discussion
We have analyzed how tensions within an organization between high powered operational incentives on the one hand and the avoidance of con‡ict on the other can a¤ect
the organization’s strategic positioning. Asymmetric incentive structures that make
certain internal functions or units dominant within the organization, and thereby determine its strategic focus, will avoid wasteful internal haggling while still providing
high powered incentives for the operational performance of the dominant unit. Such
asymmetries can be optimal for the organization even in an ex-ante perfectly symmetric
environment, i.e., if all units are equally important to the organization’s overall goal.
25
The underlying reason is that symmetric incentive structures with high powered incentives based on each unit’s performance will provide incentives to increase not only each
unit’s operational excellence but also incentives for internal haggling since each unit
tries to a¤ect the organizational strategic choices to boost its own value contribution at
the expense of other units. If interdependencies between the units are strong, this latter
e¤ect may dominate and render such symmetric and high powered incentive contracts
suboptimal. Alternatively, a symmetric incentive scheme could conceivably base remuneration only on overall company performance; but then incentive setting becomes less
e¤ective since units are not measured solely on the basis of their own value contribution, which weakens the link between their own operational e¤ort and the signals used
to evaluate performance. In turn, it reduces individual responsibility for operational
excellence. Therefore, we show that it is never optimal to rely on pure pro…t sharing
arrangements for both units.
Hence, when interdependencies are strong and thus there is room for internal haggling, introducing an incentive ‘bias’in favor of one unit can be bene…cial. This unit
should receive high powered incentives to pursue its own objectives and achieve ‘operational excellence’in its own domain. The other unit, however, receives lower powered
and overall pro…t based incentives such that it has no interest in diverting resources
to a¤ect the choice of the business strategy in its own favor. Our result thus may
give a novel interpretation to an empirical result by Floyd and Wooldridge (1997), who
…nd in a survey study that a higher variance in middle managers’ upward in‡uence
on business strategy is positively related to …rm performance. This is in line with our
theoretical predictions: Low variance, i.e., symmetric upward in‡uence, may create too
much intra-organizational haggling and therefore reduces performance.
Of course our model has limitations. For instance, it only captures internal organizational considerations, whereas market interactions and available resources are not
taken into account. However, it highlights that even in the absence of market-based
reasons and of speci…c capabilities favoring either side of the trade-o¤, the choice of
an incentive structure favoring a generic strategy may be preferable since it clari…es
responsibilities and in turn leads to higher performance when interdependencies are
strong.
26
The model also makes speci…c assumptions on the technology. For instance, we
assumed that the value contributions of the subunits are additive. In reality, the technology depends on the speci…c industry and type of the …rm. For instance, there may
well be industries in which there is a complementarity between the activities of the
di¤erent subunits such that, for instance, e¤orts in production become more valuable
with higher sales performance, and vice versa. Alternatively, both may also be substitutes when, for instance, an excellently manufactured product “sells itself” such that
high manufacturing performance can make up for lower sales e¤orts. While the basic
trade-o¤s should still occur under these alternative assumptions, the exact mechanisms
of course will change. One conjecture is that complementarities may reduce the opportunity for internal con‡ict since one unit needs the other in order to create value. On
the other hand, substitutability may lead to more con‡ict as the subunits can then become more direct competitors for scarce resources. In turn, generic business strategies
that avoid internal haggling may be more prominent when the unit’s contributions are
substitutes rather than complements. Hence, while the basic trade-o¤s between internal con‡ict and asymmetric operational incentives should exist across di¤erent …rms
and industries, of course not all …rms will end up with the same structure due to both
di¤erences in market structures and in technologies.
Although derived from a highly stylized formal model, our analysis has important
implications for management practice. First, it stresses that organizational choices and
strategic positions are interlinked in a two-way fashion: The set of available business
strategies may be restricted by the way in which these strategies can be supported
by appropriate internal incentive strategies. Even when external market conditions,
as well as available resources and capabilities, may call for a strategy that balances
di¤erent strategic elements, e.g., customer bene…ts and low costs, internal incentive
considerations can lead to superiority of a strategy in which one dimension dominates.
Hence, it may be misleading to design a strategy in a two-step procedure by thinking
about the business strategy …rst before the appropriate incentive strategy. Rather,
these steps should be closely intertwined.
Second, and closely related, we …nd that asymmetric strategies may be optimal
even in quite symmetric environments, which, in a way, is also a novel justi…cation for
27
Porter’s advice to focus on “generic business strategies”. However, not surprisingly,
in the perfectly symmetric environment of the model, symmetric strategies (“hybrid
strategies” in Porter’s terminology) are nevertheless often superior, namely, if interdependencies between di¤erent units are su¢ ciently small. A key point to take-away here
is that if symmetric high powered incentive contracts are installed, the top management
must take particular care about the problems of internal haggling and wasting resources
on internal lobbying.
Third, on a rather general level, our analysis suggests that it may often be optimal
to treat the equal unequally. Even if di¤erent functions or units are equally important
for the organization’s success, this need not to imply that incentive schemes should be
identical if the potential for internal haggling is large. For instance, it may be optimal
to set high powered incentives for the sales department with a bonus conditional on
sales success while remunerating the procurement department with a more balanced
and lower powered incentive contract, even if both departments are equally important
for the …rm.
7
Appendix
Proof of Proposition 1: (i) A direct implication of the reaction functions (4) is that
in any equilibrium of the design choice game, it must hold that
di
=
dj
i
;
j
and therefore
di
dj
=
p
z
1
dj
ip
) di = z
j
2
i
j
i
28
i
1=
+
;
j
(12)
implying
2
i
z(
Di =
2
i
z(
i+
j
i+
j)
2
j
j)
2
i
2
2
j
+ z(
=
i
i+
j)
2
i
2
j
+
j
i
i
=
2
j
i
+
:
j
Part (ii) directly follows from (3). Part (iii) directly follows since, if di = 0 (which holds
if i = 0); the other agent can tip the design in the direction of her desired form at
in…nitesimally small cost.
Proof of Lemma 1: The partial derivative of the principal’s expected pro…t function
( 1 ; 2 ; 11 ; 22 ) w.r.t. 1 is equal to
@
@
=
@
(
@
(
@
1;
(
1+
1;
2;
1
Let
( 1 ; 2 ) = max 1 ;
can insert (6) and obtain:
@
2
2
z
2)
(
2
=
z
1
1;
2)
=
z
2
(
@
1;
2)
@
(
@
1
1;
2)
22 ).
2
2
(
1
+
2
2
+r
2)
(
1
=z
2
+
2)
1;
2) ;
1
(
1
1) :
11
2q
2
2q
2
(2r 2 + q 2 )
q 2 + 4r 2
(2r 2 + q 2 )
q 2 + 4r 2
1
;
(13)
2
:
(14)
we must have
2
+
(
Applying the envelope theorem, we
+r
2
1
Note that at any internal optimum (
@
11 ;
2
+r
2
2)
2)
r
2
2 (2r
+ q2) (
q 2 + 4r
2)
1
2
=0
or equivalently
(
1
2)
To satisfy this for the case that
1+
z
1
( 1+
1
6=
2
=z
2)
2;
r
2q
2
+ 2r
2
q + 4r
we need z (
1
1+
2
2
2)
q 2 + 4r 2
=: :
r 2 (q 2 + 2r 2 )
29
= 0:
r
2 q 2 +2r
q 2 +4r
2
2
= 0 or
(15)
Hence, in any asymmetric maximum, 1 + 2 must be equal to the constant . Now,
we show that such a maximum cannot exist.
The second derivatives of the objective functions are
@2
@2
(
@
(
@
2 22
( 1 + 2 )3
2 21
1; 2)
=
z
2
( 1 + 2 )3
2
1;
2
1
2)
and
@2
@
1@
2q
r
2z
=
(
2
2
2
+ 2r
2
q + 4r
q 2 + 2r
r 2 2
q + 4r
= z
1
1
2
3
2)
+
2
=
2
2
=
2
2
2
z
2
2
1
2
z
2
1
(16)
1
(17)
< 0:
Thus, if an interior extremum exists, it must be a maximum. However, for (
be a maximum, we additionally need that
@2
(
@
1;
2
1
2)
@2
2
2
2
2
(
@
1;
2
2
2
1
2
1
@2
2)
1
2
0>(
@
4
2
2)
1
(
1;
1@
2 2
1 2
4
1;
2)
to
2
2)
>0,
2
>0,
;
but this can never be the case.
Proof of Proposition 2: We can solve for the optimal symmetric contract by substi2
2
tuting 1 = 2 =
and ii = qq2+2r
into the principal’s optimization problem to
+4r 2
obtain
q 2 +2r 2
q 2 +4r 2
max q 2
+z
1
2
q 2 +2r 2
q 2 +4r 2
2
q2 z
2
r
2
q 2 +2r 2
q 2 +4r 2
2
2
q 2 +2r 2
q 2 +4r 2
+
:
(18)
The …rst derivative of the objective function is
2
q 2 q22r
+4r
2
q 2 +2r
(q 2 +4r
2
2 )2
2r
2 2
q
z
2
q 2 +2r
(q 2 +4r
2
2 )2
4r2
4
+ 2r
2
q 2 + 2r
2
q 2 2r 2
(q 2 +4r
q2
2 )2
:
Note that the second derivative is negative. Hence, if there is an internal optimum, this
30
is characterized by the …rst-order condition, which yields
=
This is strictly positive i¤
4r 2 q 2
q 2 +4r 2
4r
2 2
q
4r
z (q 2 + 4r 2 )
:
2 (q 2 + 2r 2 )
> z.
Proof of Proposition 4:
First note from (13) and (14) that the …rst derivative of the principal’s pro…t w.r.t.
to i always becomes negative when i is su¢ ciently large. Hence, only one of the
following two contracts is optimal: either a symmetric internal contract with in‡uence
e¤ort, or an asymmetric contract preventing in‡uence e¤ort, in which for one agent
ii = ij = 0. Expected pro…ts under the optimal contract with in‡uence e¤ort can
be computed by inserting the optimal incentive parameters from Proposition 2 into the
expected pro…t function:
sym
=
2q 2 z
2 (2r 2 + q 2 )
2q 2 z
q2 + z
2
2
2 (2r + q )
1
r
2
2
2
2q 2 z
2 (2r 2 + q 2 )
2
+2
2
q2
2
z
4r
2
2
4r 2 q 2 z (4r 2 + q 2 )
z
8r 2 (2r 2 + q 2 )
!
2
=
(2q 2 z)
z2
+
8 (q 2 + 2r 2 ) 16r
2
(19)
+ z;
2 2
q
which –as we know from Proposition 2 –can only be pro…t maximizing if z < q4r
2 +4r 2 .
The expected pro…ts under the optimal contract without in‡uence e¤ort are obtained
31
by using (10) and (11) instead, such that
q2
asym
=
q 2 +2r 2
q2
2
1
r
2
=
q4
4 (q 2 + 2r
q2
+
q 2 +4r 2
2)
+
q
+ 2r
q4
4 (q 2 + 4r
q 2 +4r 2
2
2
+2
2
2)
q2
q
+ 4r
2
q2
q2
4
2
2
q2
q 2 +2r 2
+z
2
2
2
q2
q2
2
q2
! 4
(20)
+ z:
Expression (19) exceeds (20) i¤
2
(2q 2 z)
z2
q4
q4
+
>
+
8 (q 2 + 2r 2 ) 16r 2
4 (q 2 + 2r 2 ) 4 (q 2 + 4r
8q 2 r 2
8q 4 r2 4
,
z2 z 2
>
q + 4r 2
(q 2 + 4r 2 )2
z
4q 2 r 2
q 2 + 4r 2
This condition is satis…ed i¤
s
8q 4 r2
4q 2 r 2
>
z
q 2 + 4r 2
(q 2 + 4r
2
8q 4 r2
(q 2 + 4r
>
4
or z
2 )2
2)
4
:
2 )2
4q 2 r 2
<
q 2 + 4r 2
s
8q 4 r2
(q 2 + 4r
4
2 )2
or equivalently i¤
p
2 + 2 2q 2 r
z>
q 2 + 4r 2
2
2
or z <
p
2 2q 2 r
q 2 + 4r 2
2
:
2 2
q
But a symmetric internal optimal contract exists only if z < z = q4r
2 +4r 2 ; which is
p
p
(2+ 2)2q2 r 2
(2 2)2q2 r 2
strictly smaller than
but
strictly
larger
than
. Hence, such a
2
2
q +4r
q 2 +4r 2
symmetric contract is superior i¤
z<
2
p
2 2q 2 r
q 2 + 4r 2
32
2
=: zb;
which completes the proof.
Proof of Corollary 1:
For the relevant parameter values of z; straightforward calculations yield that in the
optimal symmetric contract, the agents will in expectation produce:
Qsym
=
i
z q 2 2q 2 z
+
:
2
4 q 2 + 2r 2
With the optimal asymmetric contract, we …nd (using
Qasym
=z+
1
q
q2
2
2 q + 2r
; Qasym
=
2
2
1
> 0;
q2
q
2
2 q + 2r
2
= 0) :
2
;
which implies that the di¤erence in total expected output between the symmetric and
the asymmetric contract equals:
1
2
which is positive i¤ z <
q4
q 2 + 4r
q 4 zq 2
q 2 + 2r 2
2
;
2q 2 r 2
.
q 2 +4r 2
Proof of Corollary 2:
The principal will still prefer an asymmetric solution if
asym
sym
> S:
sym
But we know from the proof of Proposition 4 that asym
is strictly positive if
z is su¢ ciently large. Hence, indeed a cut-o¤ for S exists such that the principal still
prefers an asymmetric solution for su¢ ciently small values of S.
33
References
Allen, R. W., D. L. Madison, L. Porter, P. A. Renwick, and B. Mayes
(1979): “Organizational politics,”California Management Review, 22(1), 77–83.
Baer, M., K. T. Dirks, and J. A. Nickerson (2013): “Microfoundations of strategic problem formulation,”Strategic Management Journal, 34(2), 197–214.
Barney, J. (1991): “Firm resources and sustained competitive advantage,” Journal
of Management, 17(1), 99–120.
Bidwell, M. J. (2012): “Politics and Firm Boundaries: How Organizational Structure, Group Interest, and Ressources A¤ect Outsourcing,” Organization Science,
23(6), 1622–1642.
Burgelman, R. A. (1991): “Intraorganizational ecology of strategy making and organizational adaptation: Theory and …eld research,” Organization Science, 2(3), 239–
262.
Burgelman, R. A., and A. S. Grove (2007): “Let chaos reign, then rein in
chaos°Urepeatedly: Managing strategic dynamics for corporate longevity,” Strategic
Management Journal, 28(10), 965–979.
Campbell-Hunt, C. (2000): “What have we learned about generic competitive strategy? A meta-analysis,”Strategic Management Journal, 21(2), 127–154.
Chandler, A. D. (1962): Strategy and Structure: Chapters in the History of the
Industrial Enterprise. M.I.T. Press, Cambridge.
Congleton, R., K. Konrad, and A. Hillman (2008): 40 Years of Research on
Rent Seeking (Vol. 1 and Vol. 2). Springer, Heidelberg.
Corts, K. (2000): “Focused …rms and the incentive to innovate,”Journal of Economics
& Management Strategy, 9(3), 339–362.
Crook, T., D. Ketchen Jr, J. Combs, and S. Todd (2008): “Strategic resources
and performance: a meta-analysis,” Strategic Management Journal, 29(11), 1141–
1154.
Eisenhardt, K. M., and M. J. Zbaracki (1992): “Strategic decision making,”
Strategic management journal, 13(S2), 17–37.
34
Floyd, S. W., and P. J. Lane (2000): “Strategizing throughout the organization:
Managing role con‡ict in strategic renewal,”Academy of Management Review, 25(1),
154–177.
Floyd, S. W., and B. Wooldridge (1997): “Middle managementŠs strategic in‡uence and organizational performance,” Journal of Management Studies, 34(3), 465–
485.
Ghoshal, S., and N. Nohria (1989): “Internal di¤erentiation within multinational
corporations,”Strategic Management Journal, 10(4), 323–337.
Hall, B. J. (2004): “Incentives within Organization,” Harvard Business School Case
9-904-043.
Hall, D. J., and M. A. Saias (1980): “Strategy follows structure!,” Strategic Management Journal, 1(2), 149–163.
Holmström, B., and P. Milgrom (1987): “Aggregation and Linearity in the Provision of Intertemporal Incentives,”Econometrica, 55, 303–28.
Holmström, B., and P. Milgrom (1991): “Multitask Principal-Agent Analyses:
Incentive Contracts, Asset Ownership and Job Design,” Journal of Law, Economics
and Organization, 7, 24–52.
Inderst, R., H. Müller, and K. Wärneryd (2007): “Distributional con‡ict in
organizations,”European Economic Review, 51(2), 385–402.
Konrad, K. (2009): Strategy and dynamics in contests. Oxford University Press.
Kretschmer, T., and P. Puranam (2008): “Integration through incentives within
di¤erentiated organizations,”Organization Science, 19(6), 860–875.
Kumar, M. S. (2013): “The Costs of Related Diversi…cation: The Impact of the Core
Business on the Productivity of Related Segments,” Organization Science, 24(6),
1827–1846.
Lawrence, P., and J. Lorsch (1967a): “Di¤erentiation and integration in complex
organizations,”Administrative Science Quarterly, pp. 1–47.
(1967b): Organization and environment: Managing di¤erentiation and integration. Harvard University Press, Cambridge, MA.
35
Meyer, M., P. Milgrom, and J. Roberts (1992): “Organizational prospects, in‡uence costs, and ownership changes,”Journal of Economics & Management Strategy,
1(1), 9–35.
Milgrom, P., and J. Roberts (1988): “An economic approach to in‡uence activities
in organizations,”American Journal of Sociology, 94(1), 154–179.
(1992): Economics, Organization and Management. Prentice-Hall, Englewood
Cli¤s.
Mintzberg, H. (1980): “Structure in 5’s: A Synthesis of the Research on Organization
Design,”Management Science, 26(3), 322–341.
Mintzberg, H. (1990): “The Design School: Reconsidering the Basic Premises of
Strategic Management,”Strategic Management Journal, 11(3), 171–195.
Münster, J., and K. Staal (2010): “How organizational structure can reduce rentseeking,”Public Choice, pp. 1–16.
Porter, M. (1985): Competitive advantage. Free Press New York.
Porter, M. (1996): “What Is Strategy?,”Harvard Business Review, 74(6), 61 –78.
Rivkin, J., and N. Siggelkow (2003): “Balancing search and stability: Interdependencies among elements of organizational design,” Management Science, 49(3),
290–311.
Rotemberg, J., and G. Saloner (1995): “Overt interfunctional con‡ict (and its
reduction through business strategy),”RAND Journal of Economics, 26(4), 630–653.
(2000): “Visionaries, managers, and strategic direction,” RAND Journal of
Economics, 31(4), 693–716.
Rotemberg, J. J., and G. Saloner (1994): “Bene…ts of Narrow Business Strategies,”American Economic Review, 84, 1330–49.
Schaefer, S. (1998): “In‡uence costs, structural inertia, and organizational change,”
Journal of Economics & Management Strategy, 7(2), 237–263.
Scharfstein, D., and J. Stein (2000): “The dark side of internal capital markets:
Divisional rent-seeking and ine¢ cient investment,” The Journal of Finance, 55(6),
2537–2564.
36
Siggelkow, N., and D. Levinthal (2003): “Temporarily divide to conquer: Centralized, decentralized, and reintegrated organizational approaches to exploration and
adaptation,”Organization Science, 14(6), 650–669.
Simon, H. (1962): “The architecture of complexity,” Proceedings of the American
Philosophical Society, pp. 467–482.
Teece, D. J., G. Pisano, and A. Shuen (1997): “Dynamic capabilities and strategic
management,”Strategic Management Journal, 18(7), 509–533.
Thornhill, S., and R. White (2007): “Strategic purity: A multi-industry evaluation
of pure vs. hybrid business strategies,” Strategic Management Journal, 28(5), 553–
561.
Treacy, M., and F. Wiersema (1993): “Customer Intimacy and Other Value Disciplines,”Harvard Business Review, 71(1), 84 –93.
Tullock, G. (1980): “E¢ cient rent seeking,” in Toward a theory of the rent-seeking
society, ed. by J. Buchanan, R. Tollison, and G. Tullock, vol. 97, p. 112. Texas A &
M Univ Pr.
Van den Steen, E. (2005): “Organizational beliefs and managerial vision,” Journal
of Law, Economics, and Organization, 21(1), 256.
Van Maanen, J., and E. Schein (1977): “Toward a theory of organizational socialization,” in Research in Organizational Behavior, ed. by B. M. Staw, vol. 1, Greenwich, CT. JAI Press.
Wageman, R., and G. Baker (1999): “Incentives and cooperation: The joint e¤ects
of task and reward interdependence on group performance,” Journal of Organizational Behavior, 18(2), 139–158.
Wolfstetter, E. (2002): Topics in Microeconomics. Cambridge University Press.
37