The Impact of Modularity on Intellectual Property and ... fileMODULARITY, IP AND VALUE APPROPRIATION 2 The Impact of Modularity on Intellectual Property and Value Appropriation Carliss

Copyright © 2011 by Carliss Y. Baldwin and Joachim Henkel

Working papers are in draft form. This working paper is distributed for purposes of comment and discussion only. It may not be reproduced without permission of the copyright holder. Copies of working papers are available from the author.

The Impact of Modularity on Intellectual Property and Value Appropriation Carliss Y. Baldwin Joachim Henkel

Working Paper

12-040 December 08, 2011

MODULARITY, IP AND VALUE APPROPRIATION

2

The Impact of Modularity on Intellectual Property and Value Appropriation

Carliss Y. Baldwin,1 Joachim Henkel2

December 08, 2011

Distributed innovation in open systems is an important trend in the modern global economy. In general, distributed innovation is made possible by the modularity of the underlying product or process. But despite the documented technical benefits of modularity, history shows that it is not always straightforward for firms to capture value in a modular system. This paper brings together the theory of modularity from the engineering and management literatures with the modern economic theory of property rights and relational contracts to address the question of value appropriation. It defines three generic threats to intellectual property (IP) and models the interactive impact of modularity and state-sanctioned IP rights on these threats. It identifies strategies for capturing value in so-called “open systems” in which IP is distributed among several parties. It shows why open systems should be designed as modular systems. Finally, it analyzes in detail the strategy of capturing value by maintaining exclusive control of an essential module in an open system. Keywords: Modularity, value appropriation, intellectual property, open innovation, design

1 Harvard Business School, Soldiers Field, Boston, MA 02163, USA, [email protected]. 2 Technische Universität München, Arcisstr. 21, 80333 Munich, Germany, [email protected].


3

The Impact of Modularity on Intellectual Property and Value Appropriation

INTRODUCTION

Distributed innovation in open systems is an important trend in the modern global economy. As

education levels rise and communication costs fall, more people have the means and motivation to

innovate. Supply chains now stretch around the world as firms outsource production to innovative

suppliers (Sturgeon, 2002). At the same time, many firms have structured their products as open

systems in which users and complementors are invited to innovate (Gawer and Cusumano, 2002,

Adner and Kapoor, 2010).

In general, distributed innovation in open systems is made possible by the modularity of the

underlying product or process. Modular systems are made up of components that are highly

interdependent within sub-blocks, called modules, and largely independent across those sub-blocks

(Simon, 1962; Baldwin and Clark 1997, 2000, Schilling, 2000). Independence between modules

means that changes within a module do not affect the rest of the system, a property that is known as

“information hiding” (Parnas, 1972a,b). Information hiding reduces the risk that small changes in the

environment will cause the whole system to fail, and makes it easier for the overall system to be

adapted and evolve towards higher levels of performance (Simon, 1962; Baldwin and Clark, 2000).

Despite the technical benefits of modularity, history shows that it is not always straightforward

for firms to capture value in a modular system. For example, IBM failed to capture value in the case

of the highly modular IBM PC. IBM’s managers consciously leveraged the PC’s modularity by

outsourcing most of its hardware and software components, but retained control of critical code called

the BIOS (Basic Input Output System). However, in 1983, Compaq and Phoenix Technologies

independently managed to replicate the BIOS code, thus enabling “clones” that were fully compatible

with IBM PCs. By 1992, IBM was struggling as cheap clones proliferated and replaced larger

computers in many applications (Cringely, 1992; Ferguson and Morris, 1993).


4

Success with a modular product is illustrated by the case of Valve Software, which released the

game “Half-Life” in 1998. Code for the game was divided into two modules: the source engine and

the game code (Jeppesen, 2004). Valve kept the source engine proprietary, but published the game

code and granted users a broad license to modify and share it. Within eight months of release, users

had built a modified game, “Counter-Strike,” which became far more popular than the original game.

However, to play Counter-Strike, players had to license the source engine from Valve, and thus

Counter-Strike increased total demand for Valve’s product.

On the surface, IBM’s and Valve’s strategies were not too different. Both created modular

systems, and both opened up their systems to outside innovators. But over time, value slipped away

from IBM, while (as of this writing), Valve appears to be profitable3 and still controls the technical

evolution of the system it created. In this paper, we will argue that the crucial difference between the

two firms lies in the way they managed their intellectual property in conjunction with modularity.

We define intellectual property (IP) as knowledge that is exclusively controlled by a particular

firm and thus can serve as a source of economic rent. Such property includes the classic legal forms

of IP—patents, copyrights, and trade secrets—but also includes confidential information known to the

firm’s employees and suppliers. Consistent with the property rights literature, we consider such

knowledge to be the property of a particular firm if the firm can exclude others from using it (Hart

and Moore, 1990). Our analysis will be concerned with how modularity affects the owner’s ability to

maintain exclusivity and capture value from the system.

Our analysis rests on two assumptions, which we believe are uncontroversial. First, knowledge is

a source of economic value. If unique knowledge is needed to create a valuable product or process,

then control of that knowledge can be translated into a monopoly with a corresponding flow of

monopoly rents. The ability to exclude others turns knowledge into property (Hart and Moore, 1990).

Second, knowledge is divisible. Humans have cognitive limitations (Simon, 1957), hence knowledge

is divided into domains of specialization. A key problem in the design of a complex product or

3 The company is privately owned, hence does not publish financial statements.


5

process is to partition the design tasks and related knowledge into a set of sub-problems that can be

solved by specific people who communicate and share knowledge in particular ways (Parnas,

1972a,b; Clark, 1985; von Hippel, 1990).

The “architecture” of the system in turn determines the way in which the sub-problems are

managed (Henderson and Clark, 1990; Whitney et al., 2004). In a “one-module” architecture, all sub-

problems are inter-related. As a result, every designer must know what the others are doing, and each

must be able to share his or her knowledge and reasoning with all others. This high degree of

information sharing is not necessary in a modular architecture. Here the sub-problems are partitioned

into independent modules, where “every module ... is characterized by its knowledge of a design

decision which it hides from all others” (Parnas, 1972b). The module designers do need access to a

common body of design rules, but if they obey these rules, the separate modules will operate together

as a system (Mead and Conway, 1980; Baldwin and Clark, 2000).

Thus modularity is a technical means of dividing and controlling access to knowledge. For this

reason, it can be used to preserve IP. However, if the owner of valuable knowledge already has

perfect and costlessly enforceable state-sanctioned IP rights, then he can use his rights and the powers

of the state to exclude any or all others from using his knowledge. In such cases, using modularity to

preserve exclusivity is unnecessary and redundant. We capture this reasoning in our first proposition,

which defines the scope of our analysis:

Proposition 1. In a world of perfect, costlessly enforceable state-sanctioned IP rights, it is

unnecessary to use the modularity of a product or process to protect IP.

Conversely, modularity can be a useful strategic tool if IP rights are imperfect and costly to

enforce. But using modularity to protect IP is a complex undertaking for two reasons. First, as we will

show, there are several threats to IP, and actions that increase protection against one threat may

reduce protection against others. In addition, modularity makes it technically feasible to share

knowledge about some modules, while closing off access to others. As in the case of Counter-Strike,


6

opening up some modules can instigate external innovation and increase demand for the system as a

whole, thereby benefiting the original owner. Thus, for profit-seeking designers of complex systems,

a key question is how to gain the benefits of openness and yet maintain enough exclusivity to sustain

monopoly rents.

This paper makes four distinct contributions to the theoretical literatures on modularity and

intellectual property. (1) It defines three generic threats to IP and models the interactive impact of

modularity and state-sanctioned IP rights on these threats. (2) It identifies strategies for capturing

value in so-called “open systems” in which IP is distributed among several parties. (3) It shows why

and how open systems should be designed as modular systems. (4) It analyzes in detail the strategy of

capturing value by maintaining exclusive control of an essential module in an open system.

To support our theoretical arguments, we have compiled a set of illustrative examples. Readers

will note that the majority of our examples involve digital technologies. Digital systems are

susceptible to being modularized at low cost and in many different ways (Whitney, 2004), and it is

not surprising that examples of the strategic use of modularity are clustered in situations where

modular options are numerous and cheap. Yet, as we show, examples involving other technologies

exist as well. Nevertheless, as digital technologies spread throughout the economy, we expect

opportunities to make use of modularity strategically for the purpose of value capture to increase.

The rest of the paper is organized as follows. In the next section, we describe our intellectual

roots. Fundamentally, this paper brings together the theory of modularity from the engineering and

management literatures with the modern economic theory of property rights and relational contracts.

In the section after, we begin our formal analysis by identifying three generic threats to knowledge

exclusivity and showing how modularity, in conjunction with weak or strong IP rights, affects each

threat. In this part of the paper, we focus on “closed” systems, in which a firm (or individual) owns

the knowledge underlying the system, contracts with others to realize the value of this knowledge, but

does not partition or share its own IP. We then shift our focus to “open” systems, in which several

firms (or individuals) have property rights to knowledge. We discuss how the owners of core IP can


7

capture value in open modular systems. We conclude the paper proper by describing the limitations of

our analysis, implications for scholars and managers, and directions for future work.

BACKGROUND

This paper seeks to unite two separate strands of literature, the theory of modularity, which

originated with Herbert Simon (1962), and the modern theory of property rights and relational

contracts which originated with Grossman and Hart (1986) and Hart and Moore (1990) and was

extended by Baker, Gibbons and Murphy (2002). These bodies of theory serve as the foundation of

our analysis. In addition, we have taken specific concepts and ideas from other literatures as described

below.

The Theory of Modularity

For our purposes, four key concepts from the large literature on modularity are essential. First, the

modular structure of a technical system is a choice made by the system architects (Ulrich and

Eppinger, 1994; Whitney et al. 2004). The architects’ choice is constrained by the laws of physics and

the limits of their knowledge, but most complex technical systems can be designed to be more or less

modular, and module boundaries can be located in different places (Mead and Conway, 1980;

Hennessy and Patterson, 1990; Ulrich and Eppinger, 1994; Whitney et al., 2004; Baldwin, 2008;

Fixson and Park, 2008,).

Second, the technique of modularization involves partitioning design decisions into discrete

subsets and then creating a body of design rules (also known as standards) that specify how the

resultant modules will interoperate (Mead and Conway, 1980; Baldwin and Clark, 2000). If the

separation of modules is done properly, then design decisions taken with respect to one module will


8

not affect decisions taken in other modules. Design tasks can then be allocated to different

organizational units or firms (Langlois and Robertson, 1992; Sanchez and Mahoney, 1996).4

Third, just as modules can be separated in terms of their underlying design decisions, knowledge

about modules can likewise be separated. As long as they can access the design rules, Module A’s

designers do not need to have specific knowledge about Module B’s structure. Thus the designers of

each module have (potentially) exclusive knowledge. Conversely, designers working within a module

cannot fail to share knowledge without jeopardizing the success of their efforts. It follows that

modularization is a technical means of creating non-overlapping, exclusive bodies of knowledge.

Fourth and finally, as compared with the technological and organizational consequences of

modularity, the strategic consequences—i.e., how modularity affects competition among firms—have

not been widely studied. A notable exception is Pil and Cohen (2006) who look at modularity through

the lens of the resource-based view of the firm. They argue that modularity poses a strategic trade-off

for firms: on the one hand, it makes a firm’s products easier to imitate, but on the other hand, it allows

the focal firm to innovate faster and thus stay ahead of would-be imitators. In their conclusion, Pil and

Cohen (2006: p. 1006) identify “IP in modular systems” and “open vs. closed technology systems” as

important areas for future research. These topics are the central focus of this paper.

The Theory of Property Rights and Relational Contracts

The economic theory of the firm is concerned with the design of incentives within firms and the

location of boundaries between firms (Coase, 1937). Building on an earlier theory of property rights

(Demsetz, 1967, 1988; Klein, Crawford and Alchian, 1978), Grossman and Hart (1986) and Hart and

Moore (1990)—hereafter referred to as “Grossman, Hart and Moore”—unified and extended prior

theories of the firm based on agency and transaction costs (Williamson, 1985; Jensen and Meckling,

1986; Holmstrom and Milgrom, 1994). They noted that contracts cannot be written in sufficient detail

to cover all contingencies, and in any case, can only specify behavior that can be verified by a third

4 For a survey of the extensive literature of the impact of modularity on organizations and industry structure, see Colfer and Baldwin (2010).


9

party. They then defined “property rights” as the residual rights of control over assets used in

production, and developed a theory of the optimal allocation of property rights.

Grossman, Hart and Moore framed their argument in terms of ownership to physical assets and

data (such as a customer list). Brynjolffson (1994) extended their argument to information assets. We

follow Brynjolffson in focusing on intangible, not physical assets, and we follow Hart and Moore

(1990) in defining “property” as the ability to exclude others from using an asset. We differ from

these prior works, however, in that we do not consider property rights to be secure. Indeed the first

half of our analysis will focus on threats to IP and actions that can be taken to protect it.

Baker et al. (2002) extended Grossman, Hart and Moore’s theoretical framework to include so-

called “relational contracts.” In a relational contract, deviations from cooperative behavior are

punished by terminating the relationship. As long as the near-term reward to deviation is less than the

long-term continuation value of the relationship, parties to the contract will cooperate without

enforcement by the state (Greif, 1998; 2006). Relational contracts are thus said to be “self-enforcing”

(Telser, 1980; Baldwin, 1983; Bull, 1987; Greif, 1998). They can be modeled as repeated games, a

practice we adopt below (Bull, 1987; Baker et al., 2002).

Other Sources

From Barney (1991) and the resource-based view of the firm, we take the idea that knowledge

may be a source of sustained competitive advantage, but only so long as it cannot be imitated or

substituted. From technology strategy, especially Teece (1986, 2000), we take the ideas that firms

must actively and dynamically manage their knowledge resources (“intellectual capital”) and that

profits from innovation often flow to the owners of complementary assets. Also from Teece (1986)

and from the literature on cross-country property rights (La Porta et al., 1997, 1998; Rajan and

Zingales, 1995, 2001; Maskus, 2000; Zhao, 2006; Branstetter et al., 2011), we take the idea that

property rights, especially IP rights, vary by jurisdiction and may be weak or strong. Our strategy for

modeling imperfect IP rights is adapted from Antràs, Desai and Foley (2009).


10

From the literature on information technology and networks (Arthur, 1989; Farrell and Saloner,

1986; Farrell and Shapiro, 1989; Katz and Shapiro, 1992; Shapiro and Varian, 1999) we take the

concepts of increasing returns and network externalities. From the literatures on platforms, open

source software (Gawer and Cusumano, 2002; Casadesus-Masanell and Ghemawat, 2006;

Economides and Katsamakas, 2006; Eisenmann, Parker and Van Alstyne, 2006, 2011a,b; Evans,

Hagiu and Schmalensee, 2006; Henkel, 2006; Baldwin and Woodard, 2010; Casadesus-Masanell and

Llanes, 2011), we take concepts of “open” and “closed” systems and modules. However, we differ

from the prior theoretical literature on platforms and open vs. closed systems in that our analytic

approach rests on property rights rather than price theory. We believe that property rights theory can

unify the large number of special cases previously modeled using price theory.

Finally from the literature on markets for technology, we take the idea that there are great hurdles

to setting up efficient markets for knowledge (Arrow, 1962; Arora, Fosfuri and Gambardella, 2003;

Gans and Stern, 2003, 2010). In particular, we extend the analysis of Gans and Stern (2010) by

modeling specific threats to IP. We then introduce modularity as a strategic option that can be used to

protect IP and attract outside parties to innovate within a particular system.

THE IMPACT OF MODULARITY ON THREATS TO THE VALUE OF KNOWLEDGE

Consistent with our fundamental assumptions, we stipulate that a firm has unique knowledge that

can be used to design a valuable new product or process. If the firm can maintain exclusive control of

its knowledge and has access to any complementary external knowledge, it will obtain a stream of

monopoly rents. These conditions imply three generic threats to the value of knowledge: (1)

unauthorized use of knowledge by the firm’s own agents; (2) imitation or substitution by third parties;

and (3) withdrawal of the right to use complementary knowledge owned by others.

In subsections below, we explain these threats and construct a formal model to investigate the

impact of modularity on each. A model is needed because the threats interact with each other and

with the legal system in non-obvious ways. Our core model is a principal-agent model based on the


11

concept of self-enforcing or relational contracts (Telser, 1980; Baldwin, 1983; Bull, 1987; Greif,

1998, 2006; Baker et al., 2002). For each threat, we model the value of the system with and without

modularity and with imperfectly enforceable property rights. Each model builds upon the previous

one, thus at the end of this section, we will be able to see how modularity interacts with the legal

system in the presence of all three threats.

The Threat of Unauthorized Use of Knowledge by Agents

We begin with the threat of unauthorized use by agents of the firm. In this and all subsequent

sections, we refer to the original owner of valuable knowledge as the “principal.” We assume that to

realize the value of his knowledge, the principal must employ individuals and contract with suppliers

who will turn the knowledge into a working product or process. The principal must reveal his

valuable knowledge to these agents, subject to the modular architecture of system. Those agents in

turn reveal the knowledge to competitors or set up a rival establishment. This threat is well-known in

law and economics, and has been discussed by Teece (1986), Liebeskind (1997), Rajan and Zingales

(2001), and (specifically discussing the limitations of non-compete agreements) Marx, Strumsky and

Fleming (2009).

One-module systems, with no enforceable property rights or contracts

Consider the relational contract applicable to a one-module system. As discussed above, this

architecture has the property that each design decision is related to all other decisions. To address

such interdependencies, designers working on the system must have unrestricted access to all relevant

knowledge.

Initially, we assume that property rights or contracts over knowledge are not enforceable within

the principal’s legal system. (This assumption will be relaxed in short order.) However, the principal

can set up a relational contract with his agents to protect his monopoly. Following Bull (1987) and

Baker et al. (2002), we understand the relational contract between principal and agents as a repeated


12

game in which the principal pays the agents not to defect. For simplicity, we assume all parties are

risk neutral, although this assumption is not essential to the results.

For time consistency, the principal must design the contract as a series of payments whose present

value to each agent is always greater than or equal to the agent’s expected payoff to defecting. Given

geometric time preference on the part of the agents and a constant (or zero) probability of dying in

each time period, the payments can be structured as an annuity that terminates on an agent’s death or

the dissolution of the monopoly, whichever comes first. Since payments will end on dissolution of the

monopoly, agents have incentives not to defect. But if payments stop, the agents can defect, hence the

principal has incentives to continue making payments. Thus an incentive-compatible, subgame-

perfect relational contract between the principal and agents is theoretically feasible.5

Let the total number of agents with access to the principal’s knowledge be denoted N. The agents

fall into two types. The first type, called “trustworthy” will under no circumstances defect. The

second type, called “untrustworthy” will defect if it is in their own interest to do so. Each agent

knows his or her own type, but not the types of other agents. The probability that any given agent is

untrustworthy is denoted u, and is known to both the principal and all agents. We assume that

untrustworthy agents decide independently whether to defect or not.6

Let v denote the flow of profits (rents) from the monopoly, and V ! v / r denote the capitalized

value of the rents in perpetuity. In what follows, we will use lower-case letters to denote cash flows,

and the related upper-case letters to denote the present value of the cash flows. For simplicity, we

assume a single discount rate, r, is applicable throughout.

As indicated, agents with access to the principal’s knowledge may defect to competitors. They

will then receive an aggregate reward, rxX /≡ , greater than zero. In the event of defection, the

principal will lose his monopoly and his establishment will also be worth X. We assume that the

5 Note that if one agent defects, the principal has no incentive to continue making payments to the others. 6 The timing of moves is as follows. Each period is divided into two sub-periods. In the first sub-period, agents simultaneously and independently decide whether to defect and the defectors leave. In the second sub-period, the principal learns if any have defected and pay the agents accordingly. The defectors, if any, collect and split their reward. Then, conditional on no defections, the game is repeated. There is no last period of the game, although it may end probabilistically as a result of exogenous events (see below).


13

aggregate value of the resulting duopoly is below that of the monopoly, thus 0 < 2X < V. (Otherwise

the principal would want to set up the second establishment himself.) We also assume that if several

agents defect they will band together and split the reward equally, while the principal still receives X

(this assumption simplifies the argument but is not essential).

The principal pays each agent a salary, specified in a contract, with a present value of Z if no-one

defects and zero otherwise. The minimum salary is affected by the principal’s need to make the

contract self-enforcing. Specifically, if Z < X then “defect” is the dominant strategy for each agent: if

all others stay, then defecting increases the respective agent’s payoff from Z to X > Z; if n-1 other

agents defect, then a switch from “stay” to “defect” increases the defector’s payoff from 0 to X/n > 0.

Thus, if Z < X then the unique Nash equilibrium is characterized by all agents defecting.7

To bring about an “All Stay” equilibrium, the principal must pay every untrustworthy agent an

amount whose value is equal to the maximum reward, X. And since (by assumption) the principal

cannot distinguish untrustworthy agents from trustworthy ones in the population, all agents must

receive a stream of payments whose value equals X. Thus the total cost of protecting the principal’s

knowledge against unauthorized use by agents is NX. And if the (incremental) value of the monopoly,

V–X, is less than NX, the monopoly is not worth protecting, and the principal will be content with X.

It follows that the value of the monopoly to the principal is the maximum of 0 and V–(N+1)X. We

encapsulate this reasoning in our second proposition:

Proposition 2. If property rights and contracts are not enforceable, the principal cannot

distinguish between trustworthy and untrustworthy agents, and the total reward to all defectors is

X > 0, then to protect his monopoly, the principal must pay each agent an annuity worth X. The

total cost of protecting the monopoly is NX, where N is the number of agents with access to the

principal’s knowledge. The value of the monopoly to the principal is max[0,V–(N+1)X].

7 The game is a (multi-player) prisoner’s dilemma if Z is larger than the payoff in an “All Defect” situation. This will be true if the (expected) number of untrustworthy agents is relatively high.


14

An implication of Proposition 2 is that, when X is large relative to V, the only way to sustain a

monopoly is to keep N small. Interestingly, it does not help to decrease u, the fraction of

untrustworthy agents, unless the principal can tell who is untrustworthy and who is not.

Proposition 2 has implications for the existence of markets for technology (Arrow, 1962; Arora et

al., 2001; Gans and Stern, 2010). To illustrate the problem, suppose the principal, after establishing

the monopoly, desires to sell it and pursue other interests. From the buyer’s perspective, the principal

has the knowledge and after one sale could sell it again to another party. To prevent this, the buyer

must include the erstwhile principal in a (new) relational contract, and pay him (on an ongoing basis)

not to defect. Thus, if we define a “clean sale” of property as one in which the two parties do not have

a continuing relationship, then, under the conditions stipulated in Proposition 2, there can be no clean

sale of knowledge. In addition, the most attractive buyer, from the principal’s perspective, is an

existing agent, for selling to an outsider increases N by 1, whilst selling to an agent leaves N

unchanged. These points are captured in the following corollaries:

Corollary 2A. Under the conditions set forth in Proposition 2, the principal cannot sell his

knowledge-based monopoly without becoming himself an agent who receives ongoing payments

under the relational contract.

Corollary 2B. Other things equal, a buyer who is already an agent of the principal can afford to

pay more for the knowledge-based monopoly than an outsider.8

The argument we are advancing is different from the Arrow (1962) Information Paradox, which

states that in the process of educating a buyer about the value of information, the seller may need to

disclose it, at which point the buyer can simply take the information without payment. Instead, as in

Anton and Yao (1994), the threat is that the seller will continue to have the knowledge after the sale,

hence must be given incentives (via a relational contract) not to sell it again.9 However, both the

8 Of course, it is theoretically possible for X > V–(N+1)X. In this case, no agent would be willing to exchange places with the principal, no outsider would pay anything, and the market would fail altogether. 9 In the context of a one-shot game, Anton and Yao (1994) show that a seller can use the threat of resale to elicit value from a buyer even if the buyer can costlessly expropriate the knowledge. In a multi-period model, we show that the value


15

Arrow Paradox and the “no-clean-sale” result depend on the fact that knowledge is a non-rival or

“non-subtractible” good (Romer, 1990; Ostrom, 2005).

In addition to employees and suppliers, customers may receive valuable knowledge from the

principal. Gans and Stern (2010) describe a threat related to the Arrow Paradox, which they call “user

reproducibility.” They observe that, if users have access to cheap and accurate copying technologies,

any customer may be able to re-create and transfer copies of the principal’s good, thereby breaking

his monopoly. In terms of our model, under conditions of (cheap) user reproducibility, N would

include not only employees and suppliers of the principal, but also all customers. In theory, a limited

number of customers could be made party to the relational contract, but they would have to receive a

stream of payments, perhaps in the form of follow-on services, equal in value to the defection reward

X. In a large market, the principal’s monopoly will be unsustainable, unless protected by state-

sanctioned property rights.

Impact of Modularity

As discussed above, systems can be made modular by separating design sub-problems, hiding

information, and setting up design rules to coordinate the modules (Parnas, 1972; Mead and Conway,

1980; Baldwin and Clark, 2000). We assume that the principal’s original product or process can be

divided into M modules (with corresponding module monopolies). For simplicity, we consider a

symmetric modularization: the original N agents are split into M groups with N/M agents per group.

To implement a modular architecture, the principal must create M+1 separate bodies of knowledge:

one for each module and one set of design rules spanning all modules. We assume the principal

conveys the design rules to each group of module designers, and each group then works separately

and independently of the others, without communicating.10

transferred from buyer to seller must be structured as the continuation benefit of a relational contract. In other words, for incentive compatibility, the seller must be paid each period not to defect, hence there is “no clean sale.” 10 This is the same network structure that is sometimes used in clandestine and revolutionary organizations. The logic behind the two designs is the same—to hide information and thus reduce the impact and rewards to defection.


16

We first consider the impact of a “value-neutral” modularization, i.e., one that does not change

the value of the system. If we assume that value is evenly distributed over the modules and the

system’s value equals the sum of the stand-alone values of all the modules, then the value of each

module monopoly is V/M and the reward to defecting is X/M. Then, by the same logic as was used in

Proposition 2, to prevent defections, the principal must pay each agent an amount equal in value to

the group’s reward, X/M. Summing these payments over all agents, the total cost of protecting the

monopoly under a modular architecture is NX/M < NX. Thus modularity decreases payments to

agents under the relational contract and increases the value of the principal’s monopoly.

The reduction in payments under a modular architecture is even greater if the modules are

functional complements. Suppose, following Milgrom and Roberts (1990), the value of the whole

system is greater than the sum of the stand-alone values of all the modules:

X > XA + XB +…+ XM ; (1)

where XA … XM are the defection rewards for each module. Multiplying both sides of this expression

by N/M gives us total agent payments for a system with additive module values (on the left) and one

with complementary module values (on the right). The total cost of the relational contract is lower in

a system with complementary modules.

Finally, the principal might use a modular architecture to concentrate the application of his

knowledge within a particular subset of the larger system. Then even if the per-person reward to

defection stays the same, the number of people with access to the knowledge, hence total payments

under the relational contract will decline. We summarize our reasoning in:

Proposition 3. A value-neutral modularization reduces the cost of protecting IP from

unauthorized use by agents by reducing the average defection reward per person and/or the

number of people with access to valuable knowledge.

Some modularizations increase the total value of the system. Value-increasing modularity tends

to occur when one or more components have great “technical potential,” defined as the capacity for


17

improvement in response to experimentation with new designs. For example, Baldwin and Clark

(2000) estimated that the modular architecture of IBM’s System/360 may have increased its value by

25x over previous non-modular computer systems. Value-enhancing modularizations can be

incorporated into our analysis in a straightforward way. Let the ratio of the value of the modular

system to the one-module system be denoted α (which is greater than 1), and assume that defection

rewards are proportional to value. Then in a symmetric modularization without complementarities the

defection reward per person will be αX/M, and total payments under the relational contract will then

be NαX/M. Depending on whether α/Μ is greater than or less than one, total payments under the

relational contract may be higher or lower in the modular system.

Impact of a Legal System

Up to this point, we have assumed that the principal cannot enforce IP rights or contracts, and

thus must rely on relational contracts to prevent agents from defecting with his knowledge. Weak IP

rights are characteristic of many developing countries (Branstetter et al., 2011), and thus we expect

information-hiding modularity to be useful in such jurisdictions. And even in developed countries

there is some uncertainty about a patent’s enforceability as well as the scope of copyright and trade

secrets protection (Lemley and Shapiro, 2005, 2007). However, even an imperfect legal system can

reduce payments under the relational contract, hence be useful to the principal.

Adapting the approach of Antràs et al. (2009), we model an imperfect legal system in a simple

way. Let the parameter φ denote the weakness of IP rights to the principal’s knowledge. If φ = 1, the

principal has minimal property rights, while if φ = 0, his property rights are strong enough to make

defection rewards zero. The actual value of φ depends on the surrounding legal regime and the nature

of the principal’s knowledge. The defection reward is then defined as φX for the system as a whole

and φXA, …, φXM for the modules of the system.11

11 Modules may be heterogeneous with respect to their legal protection: for example, some modules may incorporate novel technology subject to patents or creative ideas subject to copyright, while other modules may use only widely available


18

Obviously, if φ = 0, then the legal system alone will be a deterrent, and (by Proposition 1)

modularity will be irrelevant to the protection of IP.12 In contrast, if φ is positive, then agents can

expect a positive reward to defection, even in the presence of state-sanctioned IP rights (i.e., if φ < 1).

In that case, by the logic of Proposition 3, modularity can be used to reduce payments to the agents,

thus increasing the value of the monopoly to the principal.

Legal systems are not only imperfect but costly to use. We will address the cost of obtaining IP

rights below, after discussing the threat of imitation or substitution by third parties.

Three examples show how modularity and the legal system interact to protect valuable IP from

misappropriation by employees and suppliers. The examples are based on 18th Century porcelain

technology, automobile brakes, and R&D projects by multinational corporations.

Example 1—Porcelain

In the eighteenth century, Frederick Augustus II, Elector of Saxony, obtained a monopoly on

European porcelain by the simple expedient of imprisoning the inventor in a fortress in Meissen and

paying him (quite well) to work there. However, neighboring rulers were constantly offering

inducements to Meissen employees to defect to their kingdoms. Thus when the original inventor was

close to death, Augustus ordered him to divide his knowledge between two successors. One man was

told the formula for porcelain paste; the other learned the secrets of making porcelain glaze. After the

inventor died, no one person could replicate the Meissen porcelain-making process. (Gleeson, 1998).

Augustus had to rely on agents—chemists and artists—to carry out the porcelain-making process.

Furthermore, there was no perfect legal system that could effectively enforce his IP rights—a defector

had only to ride as far as the nearest border (a relatively short distance) to escape his jurisdiction. In

the beginning, Augustus managed to keep all the essential knowledge in the head of one man (N=1)

whose movements he controlled by force. He could have done the same with a single successor, but

technologies and mundane ideas. Differences in the legal status of modules can be an important consideration in practice, but to simplify notation, we suppress that complexity. 12 Mathematically, changing the cost of protecting the monopoly from NφX to NφX/M makes no difference if φ = 0.


19

he could do better. Augustus arranged to modularize knowledge about the porcelain-making process

(M=2) in such a way that X > XA + XB. This inequality reflects complementarity between the porcelain

paste and glaze modules in the sense of Milgrom and Roberts (1990): glazed porcelain products were

much more valuable than either unglazed porcelain or glazed pottery. Thus, even though N had

increased from 1 to 2, dividing up the agents’ knowledge about the process reduced their individual

rewards from defection.

Example 2—Auto Brake Stability System

In the early 2000s, an auto manufacturer developed a brake stability control system which relied

heavily on certain features of its supplier’s antilock braking system. From a strictly technical

perspective, it would have been optimal to integrate the supplier’s and the automaker’s knowledge

into a one-module system. The automaker instead developed its own stability system as a separate

module, which was added to the braking system during assembly. In the words of an interviewee

from the auto company:

Sometimes we need to re-segment both hardware and software modules, or the modularity of the system, based more on the commercial needs of, say, protecting an in-house algorithm, than on just the most efficient design. (Interview with R&D manager, 11 July 2008.)

The automaker might have taken a technically more efficient route and developed an integrated

braking-and-stability system together with its supplier. It could then have relied on IP rights to protect

its knowledge from leaking to competitors and/or paid the supplier a bonus as long as the information

did not leak out. However, the automaker clearly did not think either its IP rights or its contract with

the supplier would protect its knowledge from misappropriation. By modularizing the brake system,

the automaker suffered some technical inefficiencies, but at the same time reduced the number of

agents who needed access to its knowledge, and restricted those agents to its own employees. In

effect, the company reduced N. It may also have reduced X if its own employees’ defection rewards

were lower than the supplier’s employees’ rewards.


20

Example 3—Protecting the Value of R&D in Countries with Weak IP Rights

Zhao (2006) and Quan and Chesbrough (2010) have studied modularization to protect IP in the

context of multinational companies’ (MNCs) research and development (R&D) across international

boundaries. Knowledge created through R&D cannot be protected effectively in countries with weak

IP rights. According to Zhao, multinationals address this problem by assigning projects to such

countries whose results are strongly complementary with other projects conducted in the United

States. She presents evidence from patent citations that patents obtained by MNC subsidiaries in

countries with weak IP rights have more value inside the MNC parent company than outside of it. In a

series of case studies and interviews, Quan and Chesbrough (2010) found that MNC managers

modularized the R&D process and located projects with little stand-alone value in China because of

concerns about weak IP protection in that country.

The MNCs modularized their R&D projects so that the knowledge obtained in countries with

weak IP rights had relatively little stand-alone value. In this fashion, they reduced the potential

rewards to defectors in those countries. The multinationals could thus take advantage of the lower

cost of conducting research in developing countries, and still appropriate most of the value of their

R&D investments. Without the modularization, payments to prevent defection in countries with weak

IP rights might exceeded (or in any case reduced) the wage advantage.

Threats of Imitation or Substitution by Third Parties

We now consider the impact of modularity on threats of imitation or substitution by third parties.

People unknown to the principal may be able to imitate a product or design a substitute without

having access to the principal’s unique knowledge. If imitation or substitution by third parties is

likely, the value of the monopoly and the rewards to defection will both go down.

We model imitation and substitution by third parties using a hazard model. Let !s denote the

probability of imitation or substitution in any time period. As in the previous section, φ measures the

weakness of IP rights in the legal system, and can take on values between zero and one. If φ=0,


21

property rights are strong enough to deter all attempts at imitation or substitution.13 The parameter s

captures all other determinants of the probability of imitation or substitution. Consistent with our

assumptions in the previous section, we assume that if imitation or substitution occurs, the principal’s

per-period cash flow will drop to x and his establishment will be worth X. Thus the principal obtains

surplus cash flow of v–x as long as the monopoly endures.

Under these assumptions, the probability of the monopoly surviving from t to t+1 is ( )sφ−1 .

Using the perpetuity formula with a positive hazard rate, the value of the monopoly under this threat

(after subtracting payments to agents under the relational contract) is:

( ) ( )xNxvsq φφ −−⋅ , (2)

where:

( )srssq

φφφ

+−≡ 1 . (3)

(Here we are assuming that the monopoly is worth more than zero, even after payments under the

relational contract.)

Equation (2) shows that the value of the monopoly can be decomposed into two parts: (1) excess

cash flows (v – x – Nφ x) that continue as long as the monopoly endures; and (2) a capitalization

factor, q(φ s), that takes into account the probability (φ s) that the monopoly will end in any time

period. Obviously, a positive probability of imitation or substitution ( 0>sφ ) reduces the value of

the monopoly to the principal. Payments under the relational contract last only as long as the

monopoly endures, thus their value goes down as well, although per-period payments remain the

same.

13 As before, φ may vary across systems and modules. In addition, imitation and substitution are separate events which have different probabilities and likely different φs. In particular, IP rights are typically more effective against imitation than against substitution. We suppress these complexities in the interest of notational simplicity.


22

The Impact of Modularity

What is the impact of modularity on the probability of imitation or substitution by third parties?

Pil and Cohen (2006) argued that modularity decreases the difficulty of imitation and we agree with

their logic. Basically, by eliminating interdependencies between design elements in different

modules, modularity decreases the complexity of individual components. This reduction in the

cognitive complexity of modular systems was first recognized by Simon (1962), who considered it a

virtue. But when one is trying to maintain the exclusivity of knowledge, simplicity makes the design

more transparent and easier to imitate.

Modularity also decreases the difficulty of substitution, although the mechanism is somewhat

different. A modularization, by definition, reduces dependencies between particular components and

the rest of the system. Designers can then focus their resources on module-level experiments,

improving the designs of modular components without changing other parts (Baldwin and Clark,

2000). Ease of experimentation in turn increases rewards to potential substitutors, and the likelihood

of successful substitution.

Thus, other things equal, modularity operates to increase s, the probability of imitation or

substitution by third parties net of φ. However, the overall impact of a given modularization must

balance its effect on agent payments against the hazards of imitation and substitution. This tradeoff

must be evaluated module-by-module. Using the subscript “m” to refer to a particular module and

assuming, for simplicity, additive (rather than complementary) module values, the value of the

corresponding module monopoly, if positive, can then be written as:

( ) ( ) ( )mmmmmmmm xNxvsqXsV φφφ −−⋅=−, (4)

(The relative strength of property rights can also vary by module, but we suppress this in the interest

of simplicity.)

We have argued that the probability of imitation or substitution is higher in every module of the

modular system than in the corresponding one-module system, thus ssm > . But the probabilities will


23

not shift equally: large, complex modules and modules with low technical potential will not attract

third-party effort, hence will have relatively low sm. Conversely, small modules and those with high

technical potential will be targets for third parties and their sm will be high. At the same time,

aggregate payments to agents under the relational contract will decrease post-modularization, thus

xNxN mmmφφ <∑ . Thus in assessing the impact of a particular modularization on his IP, the

principal must trade off lower payments to agents under the relational contract against a shorter

expected lifetime of the asset because of higher probabilities of imitation or substitution by third

parties. Whether a particular modularization increases or decreases the total value of the monopoly

depends on how these countervailing mechanisms operate module-by-module and aggregate to

determine the sum of value of the module monopolies.

We capture this reasoning in the following:

Proposition 4. In the absence of a relational contract, a value-neutral modularization decreases

the overall value of monopoly to the principal by increasing the probability of imitation or

substitution by third parties. In the presence of a relational contract, the impact of

modularization depends on the balance of countervailing effects, hence is indeterminate.

Two examples, both involving IBM, show how modularity in conjunction with an imperfect legal

system affects third parties’ incentives to imitate or substitute.

Example 4—IBM PC Cloning (Threat of Imitation).

As discussed in the introduction, the IBM PC, introduced in 1981, was designed as a highly

modular system. In order to bring the PC to market quickly, IBM outsourced almost all components,

peripheral devices and software, but kept control of an essential software program called the BIOS

(Ferguson and Morris, 1993). The BIOS code in turn was protected by copyright (IBM, 1981).

However, copyrighted software can be legally imitated using a technique called “clean room reverse

engineering” (Cringely, 1992; Ferguson and Morris, 1993). In this process, designers who have never

seen the artifact are given detailed information about its behavior, and they create a new artifact that


24

exactly mimics that behavior. The new design, by definition, is not a copy, since its creators never

saw the original.

Compaq and Phoenix Technologies reverse engineered the BIOS using clean room techniques,

and Phoenix licensed its BIOS widely to clone-makers. With the advent of clones, IBM’s market

share and profits began to drop, even as sales of IBM-compatible machines and related hardware and

software took off.

IBM relied on IP rights, specifically copyright, to maintain exclusive control over the BIOS.

However, its protection turned out to be imperfect since copyright on software protects only the

concrete realization but not the functionality. Notably, the high degree of modularity of the PC system

and the small size of the BIOS assisted third-party imitators by reducing the cost of clean room

reverse engineering. A less modular system with a larger BIOS would have been less vulnerable to

this threat. Indeed, one expert has noted:

[The Macintosh] BIOS is very large and complex and is essentially part of the OS [operating system], unlike the much simpler and more easily duplicated BIOS found on PCs. The greater complexity and integration has allowed both the Mac BIOS and OS to escape any clean-room duplication efforts. (Mueller, 2003, p. 28)

Example 5—System/360 and Plug-Compatible Peripherals (Threat of Substitution)

IBM’s System/360 was the first “truly modular” computer (Ferguson and Morris, 1993). The

system was split into approximately twenty-five modules with a shared set of design rules and

standard interfaces (Baldwin and Clark, 2000). Peripheral devices, such as disk drives, tape drives,

and printers, that complied with the design rules, could be added to an existing system without

difficulty. Soon after the introduction of System/360, hundreds of new firms making peripheral

devices entered the market in competition with IBM. IBM’s top managers were surprised and

annoyed by this competition but were unable to prevent it. (Pugh et al., 1991.)

Plug-compatible firms could not legally sell pure copies of IBM equipment, because these

products were protected by IP rights, including patents, copyrights, and trade secrets law. However,

the plug-compatible devices were not copies, but new and better models that performed the same


25

functions. Hence they did not infringe on IBM’s IP. System/360’s design rules—the critical interface

standards—also did not qualify for IP protection because they lacked novelty or were obvious (Bell

and Newell, 1971).

Thus System/360’s modularity facilitated substitution by third parties. Modular substitution was

most evident in components like disk drives with high technical potential and well-codified

interfaces. However, in contrast to the IBM PC, the value of System/360 to IBM did not depend on

controlling a single module like the BIOS. Instead, the company’s profits were spread over numerous

modules, many of which had no substitutes. Thus in spite of competition from plug-compatible

peripheral firms, most of the value of System/360 was captured by IBM (Baldwin and Clark, 2000).

Legal Protection as a Modular Option

Intellectual property law generally prohibits direct imitation and some forms of substitution for

goods covered by patents, copyrights and trade secrets. Even an imperfect legal system thus reduces

the rewards to third-party imitators and substitutors just as it reduces rewards to defection by agents.

Thus in our model, lower φ, denoting stronger property rights, reduces the probability of imitation or

substitution.

However, legal systems are not only imperfect, but costly to use. Ex ante legal costs include the

cost of acquiring property rights such as patents or copyrights, and the costs of drawing up

employment contracts, non-disclosure agreements, and licenses based on IP rights. Ex post legal costs

include the costs of monitoring, litigation and enforcement. Let the value of expected legal costs (both

present and future) be denoted Lm, where again m refers to a particular module. The principal will

assert IP rights over a given module only if it pays to do so. In other words, the joint deterrent impact

of the legal system on third-party imitators and substitutors and potential defectors must outweigh the

costs of using it. This leads to:


26

Proposition 5. For a particular module, m, the principal will acquire imperfect state-sanctioned

property rights if and only if:

( ) ( )[ ] ( )[ ]mmmmmmmmm xNvsqLxNvsq 1)(1 +−⋅≥−+−⋅ φφ . (5)

(Here we assume ties are resolved in favor of the legal system.)

In general, the attractiveness of using the legal system will vary across modules. Four kinds of

modules may be observed:

1. Modules protected by agent payments only. These modules have some combination of a small number of knowledgeable agents (low Nm), a low probability of imitation or substitution (low sm), and ineffective or high-cost legal protection (high φ and/or high Lm).

2. Modules protected by the legal system only. For such modules, φ equals zero, i.e, there

are no risks of imitation or substitution and no payments under the relational contract.

3. Modules protected by a combination of the legal system and agent payments. Such cases arise when the legal system is only moderately effective (0 < φ < 1), but also not very costly to access (low L).

4. Modules not protected in any way. In these cases, both the right- and the left-hand sides

of Equation (5) are negative. Such cases are most likely to arise when the number of agents with access to knowledge is high; the defection reward is high; and the legal system is ineffective and/or expensive.

Proposition 5 makes it clear that the principal’s strategy for IP protection must be devised

module-by-module. There is no “one-size-fits-all” solution applicable to all modules. Furthermore, if

the principal chooses to use the legal system for a given module, he will set payments to agents under

the relational contract high enough to prevent agent defections. In this case, assuming no mistakes,

the principal will never have occasion to sue his own agents, but he may need to sue imitators and

substitutors who are not part of the relational contract. Thus even if most IP lawsuits are pursued

against third parties, that in itself does not mean that the principal’s own agents are not an equal or

greater threat. Agents operate “in the shadow of the law,” and the presence of an effective legal

system can reduce their incentives to defect, hence the principal’s payments under the relational

contract.


27

We now consider the last threat to the value of knowledge: the threat of withdrawal by another

owner. Obviously this threat arises only in the presence of a legal system that recognizes and protects

IP rights, albeit imperfectly.

The Threat of Withdrawal of the Right to Use Knowledge

Given a legal system that recognizes property rights in knowledge, the principal must be

concerned about where his knowledge comes from. If he uses knowledge owned by someone else,

then the other party may demand compensation. We model this threat in the following way. Suppose

the principal is presented with a demand to share his rents or face the withdrawal of the right to use a

certain body of knowledge. Consistent with the modern property rights literature (Grossman and Hart,

1986; Hart and Moore, 1990; Baker et al., 2002), we assume that the principal and the outside owner

will reach a settlement that is ex post efficient.

Given the outside owners demand, the principal faces four options (cf. Reitzig et al., 2007): (a) to

cease production altogether; (b) to “design around” the external IP, at cost Y (Gallini, 1992;

Scotchmer, 2006; Golden, 2007); (c) to use the IP in question, thus risking a lawsuit; and (d) to

negotiate with the outside owner in the shadow of options (a) to (c). For simplicity, we assume that

the design around cost, Y, is less than the total value of the monopoly for the one-module system, thus

we exclude option (a). If the external owner takes the conflict to court (c), it will prevail with

probability ! and then be awarded damages D.14 The costs of legal proceedings are C, which are split

evenly among the parties.15 By settling out of court, the parties can avoid this deadweight cost. The

Nash bargaining solution in this negotiation is for the principal to pay royalties of Dθ . However, if

DY θ≤ , the principal can also credibly threaten to design around the focal IP. By licensing, the

parties avoid the cost of designing around and will split this surplus evenly, and so the principal will

14 Note that θ is negatively correlated with ϕ, but the correlation does not have to be exact. 15 This assumption is typically correct in the United States. The alternative assumption that the loser pays C is valid in many other countries. It implies that the threshold as well as the royalty in the lower line of Equation (6) (see below) become

CD )2/1( −+ θθ , but leaves our argument qualitatively unchanged.


28

pay a royalty of Y/2. If, in contrast, DY θ> then designing around is not a credible threat and the

principal ends up paying Dθ . Thus, the cost of the threat of withdrawal, denoted W, is:

⎩⎨⎧

>≤

=DYDDYY

Wθθθ

::2/

(6)

Note that the parties settle in any case: since we assume full information and zero transaction cost of

negotiating, they reach an efficient outcome. Ironically, weaker property rights

( )DYDY /)2/( <<θ can be more beneficial to the outside owner than stronger ones ( )DY />θ ,

since in the latter case designing around becomes a credible threat.

The Impact of Modularity

Just as modularity can be used to reduce payments to agents under a relational contract, it can be

used to reduce payments to external owners of knowledge. To employ modularity for this purpose,

the principal must identify where the externally owned knowledge is used in the system and

encapsulate those areas in separate modules. Encapsulation reduces the cost of designing around the

externally owned knowledge. (A “design around” is essentially a substitution, and we have already

observed that modularity makes substitution less costly.)

The cost of a design around, Y, affects both the threshold between the two cases in Equation (6)

and the payoff in the upper line. Modularity works via both effects to reduce the cost W of the threat

of withdrawal. We capture this reasoning in:

Proposition 6. A value-neutral modularization that encapsulates knowledge owned by others

reduces the cost of designing around the knowledge, hence the cost of the threat of withdrawal.

Two examples from the software industry illustrate how modularity can mitigate the threat of

withdrawal:

Example 6—Web Server Platform

LaMantia et al. (2008) describe a software company that sells web-based applications. The

company’s entire product family depended on a single platform component, which contained both the


29

company’s own code and code licensed-in from another software vendor. The platform was designed

as a one-module system with many dependencies between the company’s own and licensed code. It

would have been very difficult to separate the licensed-in code from the rest on short notice, and thus

the firm expected to pay a high cost to renew the license.

Anticipating this threat, the firm re-designed its platform, encapsulating the licensed-in code in a

separate module. In effect, the firm prepared for designing around the licensed-in code before they

faced the actual threat of withdrawal. After the modularization, the cost of substituting other code for

the licensed-in module was greatly reduced. Indeed soon after the redesign, the company began to

offer products that used third-party substitutes for the previously licensed-in components.

Example 7—The GPL and Proprietary Licenses

Software developers today obtain code from both commercial vendors, under proprietary

licenses, and open source repositories, where code is often under the so-called General Public License

(GPL). Proprietary licenses generally prohibit passing on human-readable source code to any third

party, while the GPL explicitly states that any user of a program derived from GPL code is entitled to

its source code (Free Software Foundation, 1991). Interweaving the commercially licensed code and

the open source code thus leads to conflicting legal requirements. These conflicts can be resolved by

placing GPL and proprietary code in separate modules.16 Through modularization, the source code of

GPL modules can be revealed and the source code of proprietary modules remain unpublished.

The Value of a Module Monopoly Net of IP Protection

We can now write an expression for the value of a module monopoly (in the case of additive

modules values) net of the cost of protecting the underlying IP. As before, let the subscript “m”

denote a particular module. The value of the module monopoly (cf. Equations (4), (5)) is then:

( ) ( )[ ] ( ) ( )[ ]{ } WxNvsqLxNvsqXV mmmmmmmmmmm −+−⋅−+−⋅=− 1;1max φφ . (7)

16 Provided this separation is clear enough to ensure that the proprietary code is not to be considered as “derivative work” in the sense of the GPL (Free Software Foundation, 1991).


30

Equation (7) implies that, for each module, the principal must decide whether to seek legal

protection for his own IP in the module: this is determined by the comparison in the “max” function.

From this amount, he must subtract royalty payments to any outside owners of IP in the module, as

determined by equation (6).17 And, again assuming additive module values,18 the value of the system

as a whole is simply the sum of the values of the individual module monopolies.

APPROPRIATING VALUE IN OPEN MODULAR SYSTEMS

In the cases analyzed so far, the principal conveyed valuable knowledge to employees and

suppliers (and perhaps customers) and used knowledge owned by outside parties. However, the

principal did not voluntarily transfer property rights to his own knowledge to anyone else: in this

sense, the system was “closed.” In contrast, when we speak of an “open” system, we are referring to

cases where the principal shares some of his own IP to encourage others to innovate within the

system.19 Innovators within the system do not simply work under contract to the principal, but have

their own IP and the power to exclude others from using it.

The framework we use is, as before, the property rights framework of Grossman, Hart and Moore

(1986; 1990). Their basic argument is that many value-creating actions are “non-contractible.” For

such actions, inputs (e.g., effort) are unobservable (at least by third parties), and outputs are uncertain,

thus it is impossible to structure an enforceable contract that will elicit such actions. Self-interested

agents will take non-contractible actions only if they can claim some part of the value created by their

actions. Property rights, defined as the ability to exclude others, give agents incentives to take value-

creating, non-contractible actions.20

17 If there are several outside owners of IP, then we understand W to be the sum of payments to all of them. 18 Value additivity holds if modules are not complements in the sense of Milgrom and Roberts (1990). In particular, it holds for all modules that are not essential to a given system. We discuss essential modules in the next section. 19 Practitioners and academics use the term “open” in many different, often inconsistent ways. For example, “open source,” “open standards,” “open science,” and “open innovation” are common phrases that have different, but related meanings. See West (2007), Eisenmann et al. (2011), Boudreau and Hagiu (2011), Greenstein (2011), and Schilling (2011) for complementary discussions of what it means to be “open.” Our definition is grounded in property rights theory and corresponds closely to what practitioners call “open standards.” 20 However, agents with property rights also have incentives to take value-destroying actions to improve their bargaining positions in an ex post negotiation over distribution of the surplus (Grossman and Hart, 1986).


31

To make the partitioning of IP attractive to the principal, the external agents must be able to do

something the principal cannot do via contracts alone. For systems subject to increasing returns to

scale or network externalities, being adopted by a critical mass of users may be the difference

between survival and extinction (Farrell and Saloner, 1986; Farrell and Shapiro, 1989; Katz and

Shapiro, 1992; Shapiro and Varian, 1999). Complementary goods and services can cause the market

to “tip” in favor of the principal’s system, but the principal may not have the ability to supply the

complements himself, at least in the short run. The principal can increase the external agents’

incentives to supply complements quickly, by giving them a share of the system value created by their

own (non-contractible) effort. Thus we expect to see open systems with distributed IP created as a

competitive move in markets that are subject to “tipping” behavior.

The Indivisibility of Modules

We now argue that open systems should be designed as modular systems. Recall that by

definition, modules have highly interdependent interior structures. Within a module, every design

decision potentially depends on every other, and relevant knowledge must be shared by everyone.

It is theoretically feasible to give different agents exclusive rights to make different decisions

within a module. For example, the principal might give Alice the right to make half the design

decisions in a particular module, and Bob the right to make the other half. But if decision rights are

allocated this way, then Alice and Bob will have to figure out how to deal with each other (and how

any surplus will be split between them) before they can begin to work. The cost of coordination and

the fact that the surplus must be split reduce their incentives to invest. From this we have:

Proposition 7. In the allocation of property rights, every module should be treated as an

indivisible unit, in the sense that any agent with the right to make one design decision in a given

module should have the right to make all of them.

Proposition 7 has three immediate corollaries:


32

Corollary 7A. All open systems should be modular systems.

Corollary 7B. Systems containing only one module should not be open.

Corollary 7C. The boundaries of property in a modular system should be co-terminous with the

system’s module boundaries. In other words, property rights should shift only at module

boundaries where the dependencies between two distinct sets of design decisions are sparse.

It is important to recognize that Proposition 7 and its corollaries are normative statements that

describe best practice, but may not be descriptive of all cases. A misguided principal might split the

rights to a module between two or more parties and leave them to solve the resulting governance

problems. But this is not a smart design choice because it reduces the external agents’ incentives to

invest with no corresponding benefit to the principal.

In what follows, modules over which external agents exercise decision rights will be called “open

modules,” while those over which the principal exercises decision rights will be called “closed

modules.” The critical difference between open and closed modules lies in the answer to the question,

“does the module designer have the right to withhold her design or the output based on her design

from the principal?” If the answer to this question is “yes,” the module is open; if it is “no,” the

module is closed.21

How the Principal Can Capture Value in an Open Modular System

It is all very well to increase external agents’ incentives to invest in the system, but a profit-

seeking principal would also like to obtain some of the extra value created for himself. In general,

there are three ways for the principal to capture part of the value created in open modules. First, the

principal can charge the open modules a fee or a royalty for the use of system design rules or

standards. Second, he can purchase the output of open module designers, negotiating with them ex

21 These definitions run in parallel with those of Baker et al. (2002). What we are calling “closed modules” are produced via what they call “employment”, while “open modules” are produced via “outsourcing.”


33

post to split the value they have created through their non-contractible investments. Lastly, he can sell

a closed module that is an essential complement of the open modules.

Each of these strategies is observed in practice. The first—charge a fee—has been dealt with at

length in the literature on optimal contracts and tariffs, and we have little to add to this discussion.

Examples of companies that capture value in this way are Qualcomm, which owns numerous patents

on standards related to mobile telephony and Texas Instruments, which in 1999 reportedly earned on

the order of $800 million in patent royalties (Moeller, 2011; Rivette and Kline, 2000). A fee-based

strategy requires an effective legal system that recognizes system design rules and standards as IP. It

also requires the principal to know the identities of the providers of open modules, and have some

method of charging them. Fees are also a deterrent to some types of external innovators, in particular

user innovators who do not plan to commercialize their innovations, but are willing to reveal them

freely to others (Harhoff, Henkel and von Hippel, 2003; von Hippel, 2005).22

The second strategy—purchase the designs or products from the external innovators—is based on

Teece’s (1986) observation that profits from innovation may flow to the owners of unique

complementary assets. This strategy appears in three different guises in practice. First, the principal

may be a systems integrator as defined by Brusoni, Prencipe and Pavitt (2001). A systems integrator

controls what goes into the final assembled product, hence can demand a cut of the value created by

its suppliers (Pisano and Teece, 2007). Alternatively, the principal may control a retail channel

through which open module innovators sell their products to end users. As with systems integrators,

the owner of a retail channel can to some extent control what goes into the end users’ systems, thus

can demand a cut of the value of popular modules. For example, Apple Computer is both a systems

integrator for the iPhone handset and iPad, and also controls the retail channel of its complementors

through the iTunes Store and the Apps Store.

22 For this reason, it is not unusual for the owners of IP to provide separate licenses for commercial and non-commercial innovators.


34

Last but not least, the principal may stand ready to acquire the companies of successful open

module innovators, and split the value with them via the price paid for their shares. Cisco Systems is

known for this strategy, reportedly obtaining one-third of its IP from acquisitions (Gawer and

Cusumano, 2002; Mayer and Kenney, 2004).

The third strategy—sell a closed module that is a complement to the open modules—has the

virtue that the principal does not have to know, track, or transact with the open module innovators. As

external parties innovate on open modules, demand for the complementary closed module increases,

and the principal enjoys increased profits as a result.

To claim a share of the surplus generated by every open module innovation, the principal must

retain a monopoly over one or more modules that are complementary with all others. Following Hart

and Moore (1990), we define Module A as “essential” if all groups of modules that exclude A are

worth no more than the sum of their parts, and some groups that include A are worth much more than

the sum of their parts.23 Anyone with a monopoly over an essential module can claim a share of the

total surplus created by the system.

The “Ideal” Essential Module

Of course, the value of the essential module monopoly depends on maintaining exclusive control

of knowledge needed to make the module. That knowledge in turn is exposed to the threats of

misappropriation, imitation, substitution, and withdrawal analyzed above. To reduce the impact of

these threats, an “ideal” essential module (1) requires knowledge to be shared with only a few agents

(low N); (2) is difficult for third parties to imitate or substitute (low s); (3) has effective and low-cost

23 In mathematical notation, for all combinations, S, of modules excluding A:

V (S) = Pi

i!S"

; and for some S:

V (S! A) >> PA + Pi

i"S#

; where PA and Pi respectively denote the outside values of A and each element in the set S. Under this definition, there may be systems including A that are not functional (hence are only worth the sum of their parts), and there may also be other essential modules.


35

legal protection (low φ and L); and finally, (5) does not use knowledge owned by others (low W). In

other words, the ideal essential module should be tightly circumscribed, original, and technically

advanced. It should also have relatively complex behavior to avoid reverse engineering, and be

immune to claims of infringement by outside owners of IP. By these criteria, the IBM PC BIOS was

not an ideal essential module because it was too small and easy to imitate. The web server platform in

Example 6 was also not ideal because it mixed the firm’s own IP with licensed-in IP, hence was

vulnerable to the threat of withdrawal.

From these observations, it follows that the ideal essential module should be modular but not too

modular. As we saw in the previous section, modularity facilitates tighter control of knowledge

among the principal’s agents and encapsulation of other owners’ IP. But it also makes it easier to

legally imitate the module’s behavior and to design substitutes. Thus to protect critical IP in an

essential module, the principal may rationally make the module more opaque and more complex in

both internal structure and external behavior than it needs to be from a strictly technical standpoint.

(In fact, this is a criticism open source software advocates often levy against proprietary software

code. Cf. Raymond, 1999.)

A system can have more than one essential module and anyone with a monopoly over an essential

module can claim a share of the rents generated by all modules. Thus if some third party (not the

principal) obtains a monopoly over an essential module, the principal’s share of the surplus and the

open module innovators’ incentives to innovate will both diminish.

However, open source licenses, specifically the General Public License (GPL) and its derivatives,

allow the principal to open up essential modules to third-party innovation without creating a new

claimant to system rents. These licenses give external agents decision rights (i.e., the right to modify

a module) and use rights, but no rights to commercial profits.24

This leads to:

24 Merges (2004) calls the strategy of placing essential modules under open source license or in the public domain “property pre-empting investment.”


36

Proposition 8. When partitioning the system’s IP, the principal should either maintain direct

control of every essential module, or license an essential module’s IP under an open source

license (such as the GPL) that prevents later versions from becoming someone else’s property.

Our two final examples illustrate how the strategy of controlling essential modules works in

practice.

Example 8—Valve Software and Counter-Strike

As discussed in the introduction, Valve Software designed its game “Half-Life” in two parts: the

source engine and the complementary game code (Jeppesen, 2004). The engine was given a

proprietary license and distributed only in a machine-readable (binary) format, while the game code

was distributed as human-readable source code, and users were granted broad license to modify it. A

user-created modified game, “Counter-Strike,” became far more popular than the original game.

However, Counter-Strike players had to license and use Valve’s core engine, and thus Counterstrike

became an important driver of Valve’s growth and profitability.25

Valve’s insight was to see that games players could be an important source of creative talent, if

they could be convinced to invest their own time and effort in the process. Creative effort is a classic

form of non-contractible investment, where the quality of inputs is unobservable and the quality of

output is uncertain (Baker et al., 2002). In addition, potential game designers were located all over the

world, and their identities unknown to Valve.

Valve could attract non-contractible investments from unknown parties by giving them property

rights within the system. Thus Valve unilaterally transferred both knowledge and property rights to its

users. However, they did so selectively, publishing only a software development kit (SDK) and

restricting it to non-commercial use:

You may use, reproduce and modify the SDK on a non-commercial basis solely to develop a modified game (a "Mod") for Half-Life 2 or other Valve products compatible with and using the Source Engine. … [The modified game must be] made publicly available and distributed without charge on a

25 http://planethalflife.gamespy.com/View.php?view=Articles.Detail&id=121 (accessed 11/12/11)


37

non-commercial basis.26

In a legal system with strong, enforceable property rights, Valve might have published all of its

code and relied on copyright and license restrictions to protect its IP. Instead they implemented a

mixed strategy using both modularity and the state-sanctioned property rights. Also Valve’s managers

were aware that a truly popular game, such as Counter-Strike, had the potential to be viewed by

players as an essential module on a par with the game engine. To avoid splitting system rents with

another for-profit firm, their license stipulated that any modified game must be made “publicly

available and distributed without charge.”

Example 9—Wintel

Proposition 8 states that systems in which two or more firms have control over essential modules

are sub-optimal both from the standpoint of the principal and in terms of total value created. Thus it is

not surprising that the most famous “dual system”—the Windows/Intel (“Wintel”) architecture for

IBM-compatible PCs—came about against the express intentions of the principal, IBM.

IBM did not originally intend to convey module monopolies, much less essential module

monopolies to Intel and Microsoft. Indeed, its initial contract with Intel stipulated that the chipmaker

would share its designs with Advanced Micro Devices (AMD) so that Intel would not have a

monopoly over the supply of central processors. However, in 1986, Intel unilaterally refused to

provide AMD with masks for the new 386 family of chips,27 and thus Intel then became the sole

supplier of 386 and later chips. In effect, IBM’s contractual rights were not strong enough to protect

its second-source agreement.

IBM also intended to develop its second-generation operating system (OS/2) in partnership with

Microsoft. The two firms would then share IP rights, hence neither would have a monopoly on the

operating system. Without informing IBM, however, Microsoft developed its own incompatible

26 Valve Software, Subscriber Agreement, http://www.steampowered.com/v/index.php?area=subscriber_agreement, accessed 11/12/11). 27 Advanced Micro Devices, Inc. v. Intel Corp. (1994) 9 Cal. 4th 362 [36 Cal.Rptr.2d 581; 885 P.2d 994].


38

Windows operating system in parallel with OS/2, and then, in an acrimonious breakup, abandoned

OS/2 development to promote Windows exclusively (Ferguson and Morris, 1993).

Intel and Microsoft respectively seized essential module monopolies within the PC architecture.

Having earlier lost exclusive control of the BIOS to imitators (Example 4 above), IBM was left with

no essential module, hence could no longer claim a share of total system rents. In effect, it was frozen

out of the rent stream its modular architecture had created.

CONCLUSION

At its core, this paper seeks to understand the newly important phenomenon of distributed

innovation in open modular systems. We are specifically interested in the question, how can firms

that create open modular systems appropriate value from them? In the presence of sufficiently strong

state-sanctioned IP rights, the answer to this question is trivial. However, we showed that, when IP

rights are relatively weak, modularity interacts with imperfect IP rights to determine a given module’s

vulnerability to various IP threats. The threats we considered were misappropriation of IP by agents

of the original owner, imitation and substitution by third parties, and withdrawal of rights by another

owner of IP. After systematically analyzing the impact of modularity in conjunction with the legal

system on these threats, we were able to characterize the value of a module monopoly net of the cost

of protecting its IP. We then turned to consider “open” systems where the original owner shares or

even gives away IP to attract outside innovators. We showed why open systems should be modular

and presented a comprehensive analysis of how firms can appropriate value in such systems. Thus the

main contribution of this paper is to provide a systematic analysis of value appropriation in closed

and open modular systems.

In addition, this paper makes four contributions which may be of interest to scholars. First, we

defined three generic threats to the value of knowledge and showed how these could be modeled

within a single framework. Second, we believe we are first to show how the threat of

misappropriation of IP by a firm’s agents can be mitigated by a relational contract and to derive the


39

properties of that contract including its potential to become a multi-agent Prisoners Dilemma. Third,

our “no clean sale” result (Corollaries 2a and 2b) extends Anton and Yao’s (1994) analysis of the sale

of inventions to a multi-period context and shows that, if IP rights are weak, the seller and buyer of a

piece of IP will be bound together indefinitely in a sub-game-perfect relational contract. In other

words, to guarantee the buyer’s monopoly, the seller must become a stakeholder—perhaps a

shareholder—in the buyer’s enterprise. Fourth, we derived a formal expression (Equation 7) for the

value of a module monopoly net of the cost of protecting IP and showed how this value depends on

module-level decisions about accessing the legal system.

Our analysis has various implications for managers. First and foremost, strategies for capturing

value in an open, modular system must be formulated at the module level. The IP related to some

modules can and should be given away, although care must be taken not to let essential modules fall

into the wrong hands. Other modules can be protected via state-sanctioned IP rights and/or agent

payments under a relational contract. Finally, the modular architecture of the system should not be

cast in stone until its IP dimensions are understood. After the IP issues have been analyzed, some

modules may need to be split further to concentrate agents’ knowledge or reduce payments to outside

owners of knowledge. Others may need to be made larger to make imitation and substitution more

difficult. And always, special attention should be paid to essential modules, which have the capacity

to capture a portion of total system value.

Our analysis has many limitations, hence there are many opportunities to extend it in different

directions. From a theoretical perspective, how does opening up a system affect agents’ incentives to

defect? And what is the optimal replacement or upgrade cycle for an essential module? Questions like

these lie beyond the scope of this paper. There are also a number of interesting empirical questions to

be investigated. The most promising avenue, we think, is to look across and within large systems to

see if IP protection varies systematically between essential and non-essential modules. For example,

we expect essential modules controlled by for-profit firms to have more—and more expensive—state-

sanctioned IP protection than non-essential modules. We also expect such modules to be larger and


40

have more complex behavior than dictated by purely technical considerations. Finally we expect for-

profit architects of large systems to be averse to including IP owned by others in essential modules,

but to be willing to source essential IP from open source communities.

Modularity is not a single strategy: it is rather a large set of strategic options and related tactics

that can be deployed in different ways in different places. Time and again, our theoretical analysis

and empirical examples have shown there is no “one best way” to be modular: instead it seems

inescapable that the best use of modularity depends on an interplay of countervailing forces.

However, we hope we have convinced our readers that, in a world of distributed open innovation,

firms can make strategic use of modularity to capture value.

ACKNOWLEDGMENTS

We are grateful to Ron Adner, Christina Raasch, and Venkat Kuppuswamy as well as participants

in seminars and workshops at the Tuck School of Business, Harvard Business School, and UNC

Kenan-Flagler Business School for comments that led to significant improvements of this paper.

Errors and omissions are ours alone.


41

REFERENCES

Adner R, Kapoor R. 2010a. Value Creation in Innovation Ecosystems: How the Structure of Technological Interdependence Affects Firm Performance in New Technology Generations. Strategic Management Journal 31: 306-333.

Anton JJ, Yao D. 1994. Expropriation and Inventions: Appropriable Rents in the Absence of Property Rights, The American Economics Review, 84(1):190-209.

Antràs P, Desai M, Foley CF. 2009. Multinational Firms, FDI Flows, and Imperfect Capital Markets, The Quarterly Journal of Economics, 124 (3):1171-1219

Arora A, Fosfuri A, Gambardella A. 2001. Markets for Technology: The Economics of Innovation and Corporate Strategy, MIT Press, Cambridge, MA.

Arrow, KJ. 1962. Economics of Welfare and the Allocation of Resources for Invention, in The Rate and Direction of Inventive Activity, (R. R. Nelson, ed.) Princeton, NJ: Princeton University Press, pp. 609-625.

Arthur, WB. 1989. Competing Technologies, Increasing Returns, and Lock-in by Historical Events, Economic Journal, 99(397): 116-131.

Baker G, Gibbons R, Murphy KJ. 2002. Relational Contracts and the Theory of the Firm, The Quarterly Journal of Economics 117:39-84.

Baldwin, Carliss Y. 1983. Productivity and Labor Unions: An Application of the Theory of Self- Enforcing Contracts, The Journal of Business, 56(2):155-185

Baldwin CY. 2008. Where do transactions come from? Modularity, transactions, and the boundaries of firms. Industrial and Corporate Change 17: 155–195.

Baldwin, Carliss Y. and Kim B. Clark 1997. Managing in the Age of Modularity, Harvard Business Review Sept/Oct: 81-93.

Baldwin CY, Clark KB. 2000. Design rules, Volume 1: The power of modularity. MIT Press: Cambridge, MA.

Baldwin CY, Clark KB. 2006. The architecture of participation: Does code architecture mitigate free riding in the open source development model? Management Science 52: 1116–1127.

Baldwin CY, Woodard CJ. 2011. The architecture of platforms: A unified view. In Platforms, Markets, and Innovation, A Gawer (ed). Edward Elgar: Cheltenham, UK; 19-44.

Barney JB. 1991. Firm resources and sustained competitive advantage. Journal of Management 17: 99–120.

Bell, C. Gordon and Allen Newell 1971.. Computer Structures: Readings and Examples, New York, NY: McGraw-Hill.

Boudreau, Kevin J. and Andrei Hagiu 2011. Platform Rules: Multi-Sided Platforms As Regulators, in Platforms, Markets and Innovation, Annabelle Gawer, ed. (London: Edward Elgar)

Branstetter, Lee G., Raymond Fisman, C. Fritz Foley and Kamal Saggi 2011. Does Intellectual Property Reform Spur Industrial Development, Journal of International Economics, 83(2011): 27-36.

Brusoni S, Prencipe A, Pavitt K. 2001. Knowledge specialization, organizational coupling, and the boundaries of the firm: Why do firms know more than they make? Administrative Science Quarterly 46: 597–621.


42

Brynjolfsson, Erik 1994. Information Assets, Technology and Organization, Management Science, 40(12):1645-1662.

Bull, Clive 1987. The Existence of Self-Enforcing Implicit Contracts, The Quarterly Journal of Economics, 102 (1):147-159.

Casadesus-Masanell, Ramon and Gaston Llanes 2011. Mixed Source, Management Science, forthcoming.

Casadesus-Masanell, Ramon, and Pankaj Ghemawat 2006. Dynamic Mixed Duopoly: A Model Motivated by Linux vs. Windows. Management Science 52(7): 1072-1084.

Coase RH. 1937. The nature of the firm. Economica 4(16): 386-405.

Colfer, Lyra J. and Carliss Y. Baldwin 2010. The Mirroring Hypothesis: Theory, Evidence and Exceptions, Harvard Business School Working Paper No. 10-058, January 2010 (revised June 2010), available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1539592 .

Cringely, Robert X. 1992. Accidental Empires, Reading, MA: Addison-Wesley.

Eisenmann, Thomas, Geoffrey Parker and Marshall Van Alstyne 2006. Strategies for Two-Sided Markets, Harvard Business Review 84(10): 92-101.

Eisenmann, Thomas, Geoffrey Parker and Marshall Van Alstyne 2011. Platform Envelopment, Strategic Management Journal, forthcoming (doi: 10.1002/smj.935).

Eisenmann, Thomas, Geoffrey Parker and Marshall Van Alstyne 2011. Opening Platforms: How, When and Why? In Platforms, Markets and Innovation, Annabelle Gawer, ed., London: Edward Elgar.

Evans, David S., Andrei Hagiu and Richard Schmalensee 2006. Invisible Engines: How Software Platforms Drive Innovation and Transform Industries, Cambridge, MA: MIT Press.

Farrell, Joseph and Carl Shapiro 1989. Optimal Contracts with Lock-in, American Economic Review 79 (1): 51-68.

Farrell, Joseph and Garth Saloner 1986. Economic Issues in Standardization, Telecommunications and Equity: Policy Research Issues (New York, NY: North-Holland).

Ferguson CH, Morris CR. 1993. Computer wars: How the West can win in a post-IBM world. Times Books: New York.

Free Software Foundation. 1991. The GNU General Public License (GPL) – Version 2, June 1991. http://www.opensource.org/licenses/gpl-2.0.php.

Gallini, Nancy T. 1992. Patent Policy and Costly Imitation, Rand Journal of Economics, 23(1):52-63.

Gans, Joshua S. and Scott Stern 2003. The Product Market and the Market for "Ideas": Commercialization Strategies for Technology Entrepreneurs, Research Policy, 32(2003):333-350.

Gans, Joshua S. and Scott Stern 2010. Is There a Market for Ideas? Industrial and Corporate Change, 19(3):805-837.

Gans, Joshua S., David H. Hsu and Scott Stern 2008. The Impact of Uncertain Intellectual Property Rights on the Market for Ideas: Evidence from Patent Grant Delays, Management Science, 54(5):982-997.

Gawer A, Cusumano MA. 2002. Platform leadership: How Intel, Microsoft, and Cisco drive industry innovation. Harvard Business School Press: Boston, MA.


43

Gleeson J. 1998. The arcanum: The extraordinary true story. Warner Books: New York.

Greif, Avner 1998. Self-Enforcing Political Systems and Economic Growth: Late Medieval Genoa, in Analytic Narratives, (ed. Bates et al) Princeton University Press, Princeton, NJ.

Greif, Avner 2006. Institutions and the Path to the Modern Economy: Lessons from Medieval Trade, New York, NY: Cambridge University Press.

Grossman, Sanford J. and Oliver D. Hart 1986. The Costs and Benefits of Ownership: A Theory of Vertical and Lateral Integration, Journal of Political Economy, 94(4):691-719.

Harhoff D, Henkel J, von Hippel E. 2003. Profiting from voluntary information spillovers: how users benefit by freely revealing their innovations. Research Policy 32: 1753–1769.

Hart, Oliver and John Moore 1990. Property Rights and the Nature of the Firm, Journal of Political Economy, 98(6):1119-1158.

Henderson RM, Clark KB. 1990. Architectural innovation: The reconfiguration of existing product technologies and the failure of established firms. Administrative Science Quarterly 35: 9-30.

Henkel, Joachim 2006. Selective Revealing in Open Innovation Processes: The Case of Embedded Linux, Research Policy, 35(7):953-969.

Henkel J, Reitzig M. 2007. Patent sharks and the sustainability of value destruction strategies. Working paper: available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=985602.

Henkel J, Reitzig M. 2008. Patent sharks. Harvard Business Review 86: 129–133.

Hennessy, John L. and David A. Patterson 1990. Computer Architecture: A Quantitative Approach, San Mateo, CA: Morgan Kaufmann.

IBM 1981. IBM Personal Computer Technical Reference Manual, Boca Raton, FL: IBM.

Jensen MC, Meckling WH. 1976. Theory of the Firm: Managerial Behavior, Agency Costs, and Ownership Structure, Journal of Financial Economics 3(4): 305-360.

Jeppesen LB. 2004. Profiting from innovative user communities: How firms organize the production of user modifications in the computer games industry. Working paper No. 2004-03. Copenhagen Business School: http://ep.lib.cbs.dk/download/ISBN/ 8778690978.pdf

Katz, Michael L. and Carl Shapiro 1992. Product Introduction with Network Externalities, Journal of Industrial Economics 40 (1): 55-83.

La Porta, Rafael, Florencio Lopez-De-Silanes, Andrei Shleifer, Robert W. Vishny 1997. Legal Determinants of External Finance, The Journal of Finance, 52(3):1131-1150.

La Porta, Rafael, Florencio Lopez-De-Silanes, Andrei Shleifer, Robert W. Vishny 1998. Law and Finance, The Journal of Political Economy, 106(6): 1113-1155.

LaMantia MJ, Cai Y, MacCormack AD, Rusnak J. 2008. Analyzing the evolution of large-scale software systems using design structure matrices and design rule theory: Two exploratory cases. Seventh Working IEEE/IFIP Conference on Software Architecture: 83–92.

Langlois RN, Robertson PL. 1992. Networks and innovation in a modular system: Lessons from the microcomputer and stereo component industries. Research Policy 21: 297–313.

Lemley MA, Shapiro C. 2005. Probabilistic patents. Journal of Economic Perspectives 19(2): 75–98.

Lemley MA, Shapiro C. 2007. Patent holdup and royalty stacking. Texas Law Review 85: 1991–2049.

Liebeskind JP. 1997. Keeping organizational secrets: Protective institutional mechanisms and their costs. Industrial and Corporate Change 6: 623-663.


44

Marx, Matt, Deborah Strumsky, and Lee Fleming 2009. Mobility, Skills, and the Michigan Non-Compete Experiment, Management Science, 55(6) 875-889.

Maskus, Keith E. 2000. International Property Rights in the Global Economy, Washington DC: Institute for International Economics.

Mayer, David and Martin Kenney 2004. Economic Action Does Not Take Place in a Vacuum: Understanding Cisco's Acquisition and Development Strategy, Industry and Innovation, 11(4): 299–325.

Mead, Carver and Lynn Conway 1980. Introduction to VLSI Systems, Addison-Wesley, Reading, MA.

Merges, Robert P. 2004. A New Dynamism in the Public Domain, University of Chicago Law Review, 71(1):183-203.

Milgrom, Paul and John Roberts 1990. The Economics of Manufacturing: Technology, Strategy and Organization, American Economic Review 80 (3): 511-28.

Moeller, James H. 2011. An Estimate of the Value of Qualcomm's Patents, available at http://www.moellerventures.com/blog/36-moeller-ventures-blog/56-value-estimate-of-qualcomms-patents.html (viewed 9/18/11).

Mueller, Scott 2003., Upgrading and Repairing PCs, 15th ed., Indianapolis: Que Publishing.

Ostrom, Elinor 2005. Understanding Institutional Diversity, Princeton, NJ: Princeton University Press.

Parnas DL. 1972a. A technique for software module specification with examples. Communications of the ACM 15: 330–336.

Parnas, David L. 1972b. On the Criteria to Be Used in Decomposing Systems into Modules, Communications of the ACM 15: 1053-58.

Pil, Frits K. and Susan K. Cohen 2006. Modularity: Implications for Imitation, Innovation, and Sustained Competitive Advantage, Academy of Management Review, 31(4):995-1011.

Pisano, Gary P. and David J. Teece 2007. How to Capture Value from Innovation: Shaping Intellectual Property and Industry Architecture, California Management Review, 50(1):278-296

Pugh, Emerson W., Lyle R. Johnson, and John H. Palmer 1991. IBM's 360 and Early 370 Systems, Cambridge, MA: MIT Press.

Quan, Xiaohong and Henry Chesbrough 2010. Hierarchical Segmentation of R&D Process and Intellectual Property Protection: Evidence from Multinational R&D Laboratories in China, IEEE Transactions in Engineering Management, 57(1):9-21.

Rajan, Raghu and Luigi Zingales 1995. What Do We Know about Capital Structure? Some Evidence from International Data, The Journal of Finance, 50(5):1421-1460.

Rajan, Raghu and Luigi Zingales 2001. The Firm as a Dedicated Hierarchy: A Theory of the Origin and Growth of Firms, Quarterly Journal of Economics, 116(1):805-851.

Raymond, Eric S. 1999. The Cathedral and the Bazaar, Sebastopol, CA: O’Reilly & Associates, Inc.

Reitzig M, Henkel J, Heath CH. 2007. On sharks, trolls, and their patent prey – Unrealistic damage awards and firms’ strategies of ‘being infringed.’ Research Policy 36: 134–154.

Rivette, Kevin and David Kline 2000. Rembrandts in the Attic: Unlocking the Hidden Value of Patents, Boston MA: Harvard Business School Press.


45

Romer, Paul M. 1990. Endogenous Technological Change, Journal of Political Economy, 98(5):S71-S102

Sanchez R, Mahoney JT. 1996. Modularity, flexibility, and knowledge management in product and organizational design. Strategic Management Journal, Winter Special Issue 17: 63–76.

Schilling MA. 2000. Toward a general modular systems theory and its application to interfirm product modularity. Academy of Management Review 25: 312–334.

Schilling, Melissa A. 2011. Protecting or Diffusing a Technology Platform: Tradeoffs in Appropriability, Network Externalities, and Architectural Control, in Platforms, Markets and Innovation, Annabelle Gawer, ed., London: Edward Elgar.

Scotchmer, Suzanne 2006. Innovation and Incentives, Cambridge, MA: MIT Press.

Shapiro, Carl and Hal R. Varian 1999.. Information Rules: A Strategic Guide to the Network Economy, Boston, MA: Harvard Business School Press.

Simon HA. 1962. The architecture of complexity. Proceedings of the American Philosophical Society 106: 467–482.

Simon, Herbert A. 1957. A Behavioral Model of Rational Choice, in Models of Man, Social and Rational: Mathematical Essays on Rational Human Behavior in a Social Setting; New York: Wiley.

Sturgeon T. 2002. Modular production networks: A new model of industrial organization. Industrial and Corporate Change 11: 451–496.

Teece DJ. 1986. Profiting from technological innovation: Implications for integration, collaboration, licensing and public policy. Research Policy 15: 285–305.

Teece DJ. 2000. Managing intellectual capital: Organizational, strategic, and policy dimensions. Oxford University Press: Oxford, UK.

Telser, L. G. 1980. A Theory of Self-Enforcing Agreements, The Journal of Business, 53(1): 27-44

von Hippel, Eric 1990. Task Partitioning: An Innovation Process Variable, Research Policy 19: 407-18.

von Hippel, Eric 2005. Democratizing Innovation, Cambridge, MA: MIT Press.

West, Joel 2007. The Economic Realities of Open Standards: Black, White and Many Shades of Gray, in Shane Greenstein and Victor Stango, eds., Standards and Public Policy, Cambridge: Cambridge University Press.

Whitney, Daniel E. 2004. Physical Limits to Modularity, http://esd.mit.edu/symposium/pdfs/papers/whitney.pdf, viewed July 21, 2005.

Williamson OE. 1979. Transaction-cost economics: The governance of contractual relations. Journal of Law and Economics 22: 233–261.

Williamson OE. 1985. The Economic Institutions of Capitalism. Free Press: New York, NY.

Zhao M. 2006. Conducting R&D in countries with weak intellectual property rights protection. Management Science 52: 1185–1199.

Documents

The Impact of Modularity on Intellectual Property and ... fileMODULARITY, IP AND VALUE APPROPRIATION 2 The Impact of Modularity on Intellectual Property and Value Appropriation Carliss