Von Hippel 2011

OrganizationScienceVol. 22, No. 6, NovemberDecember 2011, pp. 13991417issn 1047-7039 eissn 1526-5455 11 2206 1399 http://dx.doi.org/10.1287/orsc.1100.0618

2011 INFORMS

Modeling a Paradigm Shift: From Producer Innovationto User and Open Collaborative Innovation

Carliss BaldwinHarvard Business School, Boston, Massachusetts 02163, [email protected]

Eric von HippelSloan School of Management, Massachusetts Institute of Technology, Cambridge, Massachusetts 02141, [email protected]

In this paper, we assess the economic viability of innovation by producers relative to two increasingly important alternativemodels: innovations by single-user individuals or firms and open collaborative innovation. We analyze the design costsand architectures and communication costs associated with each model. We conclude that both innovation by individualusers and open collaborative innovation increasingly compete with and may displace producer innovation in many partsof the economy. We explain why this represents a paradigm shift with respect to innovation research, policy making, andpractice. We discuss important implications and offer suggestions for further research.

Key words : user innovation; modularity; open innovation; collaborative innovationHistory : Published online in Articles in Advance March 11, 2011.

1. Introduction and OverviewEver since Schumpeter (1934) promulgated his theoryof innovation, entrepreneurship, and economic develop-ment, economists, policy makers, and business managershave assumed that the dominant mode of innovation isa producers model; that is, it has been assumed thatthe most important designs for innovations would orig-inate from producers and be supplied to consumers viagoods and services that were for sale. This view seemedreasonable on the face of itproducer innovators gener-ally profit from many users, each purchasing and using asingle producer-developed design. Individual-user inno-vators, in contrast, depend only on their own in-houseuse of their design to recoup their innovation-relatedinvestments. Presumably, therefore, a producer servingmany customers can afford to invest more in an inno-vation design than can any single user. From this it hasbeen generally assumed that producer-developed designsshould dominate user-developed designs in most parts ofthe economy.

This long-held view of innovation has, in turn, ledto public policies based on a theory of producer incen-tives. Producers, it is argued, are motivated to innovateby the expectation of profits. These profits will disappearif anyone can simply copy producers innovations, andtherefore producers must be granted subsidies or intellec-tual property rights that give them exclusive control overtheir innovations for some period of time (Machlup andPenrose 1951, Teece 1986, Gallini and Scotchmer 2002).

However, the producers model is only one mode ofinnovation. Two increasingly important additional mod-els are innovations by single-user firms or individuals

and open collaborative innovation. Each of these threeforms represents a different way to organize human effortand investments aimed at generating valuable new inno-vations. In the body of this paper, we will analyze thesethree models in terms of their technological propertiesspecifically, their design costs and architecture, and theircommunication requirements. In these two technologicaldimensions, each model has a different profile that givesit economic advantages under some conditions and dis-advantages in others. Each has a valuable role to play inthe economy.

Our modeling of design costs, design architectures,and communication costs allows us to place bounds onthe contexts in which each model will be economicallyviable. Our analysis will lead us to conclude that inno-vation by individual users and user firms as well as opencollaborative innovation are modes of innovating thatincreasingly compete with and may displace producerinnovation in many parts of the economy. This shift isbeing driven by new technologiesspecifically, the tran-sition to increasingly digitized and modularized designand production practices, coupled with the availabilityof very-low-cost, Internet-based communication.

We will show that under the right technological andarchitectural conditions, users seeking direct value candisintermediate producers seeking profit. Our analysistherefore challenges the assumption that the profits arethe only economically important motive causing inno-vators to create new designs. Profits to innovation arisebecause users are willing to pay for a new or superiorproduct or process. Thus profit itself is derived fromusers willingness to pay. We are not suggesting that the

1399

Baldwin and von Hippel: From Producer Innovation to User and Open Collaborative Innovation1400 Organization Science 22(6), pp. 13991417, 2011 INFORMS

profit motive is not a major stimulus to investment ininnovation, but value to users is both a necessary con-dition for profits to exist and an alternate motive forinvesting in innovative designs.

We will argue that, taken in combination, the patternsand findings we describe create a significant change inthe problem field in innovation research, policy mak-ing, and practice, and thus represent a paradigm shift inthese fields (Kuhn 1962).

In 2 we review relevant literature. In 3 we presentand explain conditions under which each of the threeeconomic models of innovation we describe is viable.In 4 we discuss some broader patterns related to ourmodels and also offer suggestions for further research.

2. Literature ReviewIn this section we briefly review the literature on userinnovation, openness of intellectual property, and collab-orative innovation and modular designs.

2.1. Innovation by UsersUsers, as we define the term, are firms or individualconsumers that expect to benefit from using a design,a product, or a service. In contrast, producers expectto benefit from selling a design, a product, or a ser-vice. Innovation user and innovation producer are thustwo general functional relationships between innova-tor and innovation. Users are unique in that they alonebenefit directly from innovations. Producers must sellinnovation-related products or services to users; hencethe value of innovation to producers is derived fromusers willingness to pay. Inventors creating knowledgeto sell rather than use are producers, as are those thatinnovate to manufacture and sell goods embodying orcomplementing the innovation.

Qualitative observations have long indicated that im-portant process improvements are developed by employ-ees working for firms that use them. Smith (1937, p. 17)pointed out that a great part of the machines madeuse of in those manufactures in which labor is mostsubdivided, were originally the inventions of commonworkmen, who, being each of them employed in somevery simple operation, naturally turned their thoughtstowards finding out easier and readier methods of per-forming it. Rosenberg (1976) studied the history of theU.S. machine tool industry and found that important andbasic machine types like lathes and milling machineswere first developed and built by user firms having astrong need for them. Textile manufacturing firms, gunmanufacturers, and sewing machine manufacturers wereimportant early user developers of machine tools.

Quantitative studies of user innovation documentthat many of the most important and novel productsand processes commercialized in a range of fields aredeveloped by users for in-house use. Thus, Enos (1962)

reported that nearly all the most important innovationsin oil refining were developed by user firms. von Hippel(1976, p. 1977) found that users were the developersof approximately 80% of the most important scientificinstrument innovations as well as the developers of mostof the major innovations in semiconductor processing.Pavitt (1984) found that a considerable fraction of inven-tion by British firms was for in-house use. Shah (2000)and Hienerth (2006) found that the most commerciallyimportant product innovations in several sporting fieldstended to be developed by individual users.

Empirical studies also show that many usersfrom6% to nearly 40%engage in developing or modify-ing products. This has been documented in the case ofseveral specific types of industrial products and con-sumer products (Urban and von Hippel 1988; Herstattand von Hippel 1992; Morrison et al. 2000; Lthje et al.2002; Lthje 2003, 2004; Franke and von Hippel 2003;Franke and Shah 2003). It has also been documented inlarge-scale, multi-industry surveys of firms developingprocess innovations for in-house use in both Canada andthe Netherlands (Arundel and Sonntag 1999, Gault andvon Hippel 2009, de Jong and von Hippel 2009). Finally,via a survey of a representative sample of consumers inthe United Kingdom, it has been found that 6.2% of theUK populationabout 3 million peoplehave recentlydeveloped or modified consumer products to better servetheir personal needs (von Hippel et al. 2010).

When taken together, the findings of these empiricalstudies make it very clear that users have long beendoing and continue to do a lot of commercially signifi-cant product and process development and modificationin many fields.

2.2. Innovation OpennessAn innovation is open in our terminology when allinformation related to the innovation is a public goodnonrivalrous and nonexcludable. This usage is closelyrelated to the meaning of open in the terms opensource software (Raymond 1999) and open science(Dasgupta and David 1994). It differs fundamentallyfrom the recent use of the term to refer to organizationalpermeabilityan organizations openness to the acqui-sition of new ideas, patents, products, etc., from outsideits boundaries, often via licensing protected intellectualproperty (Chesbrough 2003).

Economic theorists have long thought that inno-vation openness is undesirablethat uncompensatedspillovers of proprietary innovation-related knowledgedeveloped by private investment will reduce innova-tors expected profits from innovation investments andso reduce their willingness to invest. Accordingly, manynations have long offered intellectual property rightsgrants that afford inventors some level of temporarymonopoly control over their inventions. The assump-tion has been that losses incurred because of intellectual

Baldwin and von Hippel: From Producer Innovation to User and Open Collaborative InnovationOrganization Science 22(6), pp. 13991417, 2011 INFORMS 1401

property rights grants will be more than offset by gainsto society from related increases in innovation invest-ment, and increased disclosure of information that wouldotherwise be kept hidden as trade secrets (Machlup andPenrose 1950, Penrose 1951, Foray 2004, Heald 2005).

Given this argument, empirical research should showinnovators striving to keep information on their inno-vations from being freely diffused. However, researchinstead shows that both individuals and firms often vol-untarily freely reveal what they have developed. Whenwe say that an innovator freely reveals information aboutan innovation, we mean that exclusive intellectual prop-erty rights to that information are voluntarily given up bythe innovator, and all interested parties are given accessto itthe information becomes a public good (Harhoffet al. 2003). (Intellectual property rights may still beused to protect the developers of these public goods fromliability and to prevent expropriation of their innovationsby third parties; see OMahony 2003.)

The practices visible in open source software develop-ment were important in bringing the phenomenon of freerevealing to general awareness. In these projects it wasclear policy that project contributors would routinely andsystematically freely reveal code they had developed atprivate expense (Raymond 1999). However, free reveal-ing of innovations has a history that began long beforethe advent of open source software. Allen (1983) andNuvolari (2004) described and discussed 19th-centuryexamples. Contemporary free revealing by users hasbeen documented by von Hippel and Finkelstein (1979)for medical equipment, by Lim (2009) for semiconduc-tor process equipment, by Morrison et al. (2000) forlibrary information systems, and by Franke and Shah(2003) for sporting equipment. Gault and von Hippel(2009) and de Jong and von Hippel (2009) have shownin multi-industry studies in Canada and the Netherlandsthat user firms developing process equipment oftentransfer their innovations to process equipment suppli-ers without charge. In the case of consumer products,several studies have shown that it is quite rare for con-sumers to attempt to protect or restrict access to inno-vations they have developed (Shah 2000, Raasch et al.2008, von Hippel et al. 2010).

Evidence has now accumulated that innovators whoelect to freely reveal their innovations can gain sig-nificant private benefitsand also avoid some privatecosts. With respect to private benefits, innovators thatfreely reveal their new designs often find that othersthen improve or suggest improvements to the innova-tion, to mutual benefit (Raymond 1999). Freely revealingusers also may benefit from enhancement of reputation,from positive network effects as a result of increaseddiffusion of their innovation, and from other factorssuch as obtaining a source of supply for their inno-vation that is cheaper than in-house production (Allen1983, Lerner and Tirole 2002, Harhoff et al. 2003,

Lakhani and Wolf 2005, von Hippel and von Krogh2003). With regard to cost, protecting design informa-tion is generally expensive, requiring security walls andrestricted access or the enforcement of intellectual prop-erty rights (Blaxill and Eckardt 2009). For this rea-son, preventing others from viewing and using a newdesign may be significantly more costly than leaving thedesign open for inspection or use by any interested party(Baldwin 2008).

Not surprisingly, the incentive to freely revealdecreases if the agents compete with one anotherforexample, if they are firms making the same end productor individuals competing in a sport (Franke and Shah2003, Baldwin et al. 2006). Selective openness strate-gies illustrate this point nicely. Thus, Henkel (2003) hasdocumented selective free revealing among producers inthe case of embedded Linux software. The producerspartition their code into open modules on which theycollaborate and closed modules on which they compete(Henkel and Baldwin 2010).

2.3. Collaboration and ModularityCollaboration is a well-known attribute of online, mul-ticontributor projects such as open source softwareprojects and Wikipedia (Raymond 1999, Benkler 2002).Franke and Shah (2003) studied user innovators in foursporting communities and found that all had receivedassistance in their development efforts by at least oneother user from their communities. The average numberassisting each user innovator was three to five. Finally,a study of process equipment innovations by high-techsmall and medium enterprises in the Netherlands con-ducted by de Jong and von Hippel (2009) found that24% of 364 user firms drawn from a wide range ofindustries had received assistance in their innovationdevelopment work from other process equipment users.

Modular design architectures are an important aidto collaborative work. A modular system is one inwhich the elements, which may be decisions, tasks, orcomponents, are partitioned into subsets called mod-ules. Within each module, elements of the system aredensely interdependent: changing any one will requirechanges in many others. Across modules, however, ele-ments are independent or nearly so; a change in onemodule by definition does not require changes in oth-ers (Baldwin and Clark 2000). Modular systems canbe easily broken apart: Herbert Simon called such sys-tems near-decomposable (Simon 1962). Furthermore,given appropriate knowledge, a nonmodular system canbe made modular (or near decomposable) by creat-ing a set of coordinating design rules that establishinterfaces and regulate the interactions of the mod-ules (Mead and Conway 1980, Baldwin and Clark2000, Brusoni and Prencipe 2006). Most design-relevantknowledge and information does not need to cross mod-ule boundaries. This is the property of information hid-ing (Parnas 1972).


Modularity is important for collaboration in designbecause separate modules can be worked on indepen-dently and in parallel, without intense ongoing commu-nication across modules. Designers working on differentmodules in a large system do not have to be colocatedbut can still create a system in which the parts can beintegrated and will function together as a whole. In smallprojects or within modules, designers can utilize action-able transparency rather than modularity to achievecoordination. When the targeted design is small, eachdesigners activities will be transparent to his or hercollaborators. Each contributor can then take separateaction to improve the design, building on the transpar-ently visible contributions of the others. In open collab-orative projects, modularity and actionable transparencygenerally go hand in hand, with both factors contributingto the divisibility of tasks (Colfer and Baldwin 2010).

Building on the arguments of Ghosh (1998), Raymond(1999), and von Hippel and von Krogh (2003), Baldwinand Clark (2006b) showed formally that if communi-cation costs are low relative to design costs, then anydegree of modularity suffices to cause rational inno-vators that do not compete with respect to the designbeing developed to prefer open collaborative innovationover independent innovation. This result hinges on thefact that the innovative design itself is a nonrival good:each participant in a collaborative effort gets the valueof the whole design but incurs only a fraction of thedesign cost.

3. Where Is Each Model Viable?Previous work has demonstrated the existence of thethree basic ways of organizing innovation activity andhas elucidated their characteristics. However, to ourknowledge, there has been no systematic thinking aboutthe conditions under which each model is likely toappear and whether each is expanding or contracting rel-ative to the other two. To make progress on these ques-tions, it is necessary to develop a theoretical frameworkthat locates all three models in a more general space ofattributes. That is our aim in this section.

Our methodology is that of comparative institutionalanalysis. In this diverse literature, laws, social customs,modes of governance, organizational forms, and indus-try structures are compared in terms of their incentives,economic consequences, and ability to survive and growin a given historical setting or technological context. Inthe particular branch we are most concerned with, orga-nizational forms and industry structures are taken to beendogenous and historically contingent (Chandler 1977;Woodward 1965; Williamson 1985, 1991; Nelson andWinter 1982; Aoki 1988, 2001; Langlois 1986b, 2002;Baldwin and Clark 2000; Jacobides 2005). Differentforms may be selected to suit different environmentsand then adaptively modified. Thus, organizational forms

emerge in history and recede as technologies and pref-erences change.

Our approach is modeled after Williamsons (1985,1991) analysis of transaction costs and especially afterFama and Jensens (1983a, b) account of how agencycosts affect organizational forms. However, in contrastto this prior work, we will not attempt to determinewhich model is most efficient in terms of minimizingtransaction or agency costs, but we instead will establishbounds on the viability of each model. When more thanone form is viable, we do not expect to see one formdrive out the other (as is the common assumption) butrather expect to see creative combinations of the formsto take advantage of what each one does best.

Finally, in contrast to virtually all prior work exceptfor Chandler (1977) and Woodward (1965), we take anexplicitly technological approach to the question of via-bility. We fundamentally assume that in a free economy,the organizational forms that survive are ones with ben-efits exceeding their costs (Fama and Jensen 1983a, b).Costs in turn are determined by technology and changeover time. Thus, Chandler (1977) argued that the mod-ern corporation became a viable form of organization(and the dominant form in some sectors) as a conse-quence of the (partly endogenous) decline in productioncosts for high-flow-through technologies, together with(exogenous) declines in transportation and energy costs.Adopting Chandlers (1977) logic, we should expect aparticular organizational form to be prevalent when itstechnologically determined costs are low and to growrelative to other forms when its costs are declining rela-tive to the costs of other forms.

Today, design costs and communication costs are de-clining rapidly, and modular design architectures arebecoming more common for many products and pro-cesses. In the rest of this section, we argue thatthese largely exogenous technological trends are makingsingle-user innovation and open collaborative innovationviable across a wider range of innovation activities thanwas the case before the arrival of technologies such aspersonal computers and the Internet. We have seen, andexpect to continue to see, single-user innovation andopen collaborative innovation growing in importance rel-ative to producer innovation in most sectors of the econ-omy. We do not believe that producer innovation willdisappear, but we do expect it to become less pervasiveand ubiquitous than was the case during most of the20th century, and to be combined with user and opencollaborative innovation in many settings.

3.1. DefinitionsA single-user innovator is a single firm or individual thatcreates an innovation to use it. Examples are a singlefirm creating a process machine to use it, a surgeon cre-ating a new medical device to use it, and an individual


consumer creating a new piece of sporting equipment touse it.

A producer innovator is a single noncollaboratingfirm. Producers anticipate profiting from their design byselling it to users or others: by definition, they obtainno direct use value from a new design. We assume thatthrough secrecy or intellectual property rights a producerinnovator has exclusive access and control over the inno-vation, and so is a monopolist with respect to its design.Examples of producer innovators are (1) a firm or indi-vidual that patents an invention and licenses it to others,(2) a firm that develops a new process machine to sell toits customers, and (3) a firm that develops an enhancedservice to offer its clients.

An open collaborative innovation project involvescontributors who share the work of generating a designand also reveal the outputs from their individual andcollective design efforts openly for anyone to use. Thedefining properties of this model are twofold: (1) theparticipants are not rivals with respect to the inno-vative design (otherwise, they would not collaborate),and (2) they do not individually or collectively planto sell products or services incorporating the innova-tion or intellectual property rights related to it. Many,but not all, open source software projects have thesecharacteristics.

A design is a set of instructions that specify how toproduce a novel product or service (Simon 1981; Romer1990; Suh 1990; Baldwin and Clark 2000, 2006a). Theseinstructions can be thought of as a recipe for accom-plishing the functional requirements of the design (Suh1990, Winter 2010, Dosi and Nelson 2010). In the caseof products or services that themselves consist of infor-mation such as software, a design for an innovation canbe virtually identical to the usable product itself. In thecase of a physical product such as a wrench or a car,the design recipe must be converted into a physical formbefore it can be used.

A given mode of innovation is viable with respect toa particular innovation opportunity if the innovator oreach participant in a group of innovators finds it worth-while to incur the requisite costs to gain the anticipatedvalue of the innovation. By focusing on anticipated ben-efits and costs, we assume that potential innovators arerational actors who can forecast the likely effects of theirdesign effort and choose whether to expend the effort(Simon 1981, Langlois 1986a, Jensen and Meckling1994, Scott 2001). Our definition of viability is relatedto the contracting view of economic organizations, theconcept of solvency in finance, and the concept of equi-librium in institutional game theory.

In the contracting literature, firms and other organi-zations are viewed as a nexus of contracts, that is, aset of voluntary agreements (Alchian and Demsetz 1972;Jensen and Meckling 1976; Fama and Jensen 1983a, b;

Demsetz 1988; Hart 1995). For the firm or organiza-tion to continue in existence, each party must perceivehimself or herself to be better off within the contractingrelationship than outside of it.

In finance, a firm assembles resources by issuingclaims (contracts) in the form of debt and equity. It usesthe proceeds to purchase assets and to bridge the gapbetween cash outflows and inflows. A firm is solvent aslong as it can pay off or refinance all its debt claims andhave something left over. If this condition is not met, thefirm ceases to be a going concern and must be liquidatedor reorganized.

In institutional game theory, an institution is definedas the equilibrium of a game with self-confirming beliefs(Aoki 2001, Greif 2006). Within the institutional frame-work, participants join or contribute resources in theexpectation that other parties will enact their respec-tive roles. If all behave as the others expect, everyonesinitial beliefs are confirmed: the pattern of action thenbecomes a self-perpetuating institution. When the par-ticipants in the institution are rational actors, one oftheir self-confirming beliefs must be that I am betteroff participating in this institutional arrangement thanwithdrawing from it. In this view, a stable nexus of con-tracts, a solvent firm, and an active open collaborativeinnovation project are all special cases of institutionalequilibria.

We define an innovation opportunity as the opportu-nity to create a new design. With respect to a particularinnovation opportunity, each of the three models of inno-vation may be viable or not, depending on the benefitsand costs flowing to the actors.

In terms of benefits, we define the value of an inno-vation 4V 5 as the benefit that a party expects to gainfrom converting an innovation opportunity into a newdesignthe recipeand then turning the design into auseful product, process, or service. Different individualsand organizations may benefit in different ways. By def-inition, users benefit from direct use of the product, pro-cess, or service specified by the new design. Producersbenefit from profitable sales, which may take the formof sales of intellectual property (a patent or license) orsales of products or services that embody the design.Ultimately, however, a producers benefit (hence, value)derives from the users willingness to pay for the inno-vative design.

Each innovation opportunity has four generic costs:design cost, communication cost, production cost, andtransaction cost. Consistent with our assumption thatinnovators are rational actors, we assume that these costs(as well as benefits) are known ex ante to potential inno-vators, although there may be uncertainty in their assess-ments. As with value, the costs may differ both acrossindividuals and across the three models of innovation.

Design cost (d) is the cost of creating the design foran innovationthe instructions that when implemented


will bring the innovation into reality. Following Simon(1981), these costs include (1) the cost of identifyingthe functional requirements (that is, what the design issupposed to do); (2) the cost of dividing the overall prob-lem into subproblems, which can be solved separately;(3) the cost of solving the subproblems; and (4) the costof recombining the subproblems solutions into a func-tioning whole. Many of these tasks require calculations,and thus design cost is strongly affected by the cost ofcomputation.

Communication cost (c) is the cost of transferringdesign-related information among participants in differ-ent organizations during the design process. Under thisdefinition, single-user innovators, because they are inthe same organization, incur no communication cost.(Of course, there can be intraorganization costs of com-munication. However, for our purposes it is sufficientif the costs of communication are less within an orga-nization than across organizational boundaries.) Pro-ducer innovators and innovators collaborating in an openproject must communicate across organizations and thusincur communication costs.

Production cost (u) is the cost of carrying out thedesign instructions to produce the specified good or ser-vice. The input is the design instructions (the recipe)plus the materials, energy, and human effort specifiedin those instructions; the output is a good (the designconverted into usable form).

Transaction cost (t) is the cost of establishing prop-erty rights and engaging in compensated exchanges ofproperty. For an innovation, transaction cost includesthe cost of creating exclusive rights to the design, bykeeping it secret or by obtaining a patent or copy-right. It also includes the cost of controlling opportunis-tic behavior (Williamson 1985), writing contracts (Hart1995), and accounting for transfers and compensation(Baldwin 2008).

3.2. Bounds on ViabilityEvery innovation opportunity, that is, every potentialnew design, can be characterized in terms of its valueand the four dimensions of cost described above. Thecriterion of viability can thus be specified mathemati-cally as follows.

Bounds on Viability 1. For a given innovation op-portunity, a particular model of innovation is viable ifand only if for each necessary contributor to the model

Vi >di + ci + ui + ti0 (1)(The subscripts indicate that the benefits and costs mayvary by contributor and across models.)

For single-user innovators and producer innovators,there is only one contributor to be considered. (Producerinnovators may employ many people, but the producerscontracts with employees are subsumed in the producers

costs.) In open collaborative innovation projects, how-ever, there are several or many contributors, and theinequality must hold for each one individually. Noticewe have defined the criterion as a strict inequality: weassume that the actors must anticipate a strictly posi-tive gain to undertake the effort and cost of innovation.We do not rule out the possibility that the activities ofdesign, communication, production, or exchange mightbe pleasurable for some agents: if this is the case, therelevant cost would be negative for those agents. How-ever, the cases of interest here are those for which thesum of costs is positivethat is to say, the innovation isnot a free good.

As indicated in the introduction, design costs andcommunication costs have declined and are continuingto decline very rapidly because of the advent of personalcomputers and the Internet. We believe these largelyexogenous technological trends are the main causes ofthe increasing importance of single-user and open col-laborative innovation models in the economy at large.To make this argument as clear as possible, we will firstfocus our analysis on these costs alone, holding produc-tion costs and transaction costs constant across all threeeconomic models. Once we have established bounds onviability for the three models with respect to design andcommunication costs, we will reintroduce the other twodimensions of cost and show how they affect the results.

To simplify our notation in the next few sections, wedefine v as the value of an innovation opportunity netof production and transaction costs. Because it subtractsout production and transaction costs, v can be thoughtof as the (expected) value of the design alone, before itis put up for sale or converted into a useful thing. Thebounds of viability can then be restated as follows.

Bounds on Viability 2. For a given innovation op-portunity, if production and transaction costs are con-stant across models, a particular model of innovation isviable if and only if for each necessary contributor tothe model

vi >di + ci0 (2)With this simplifying assumption, we can now rep-

resent innovation opportunities with different costs aspoints in a graph with design cost and communicationcost on the x and y axes, respectively. We next ask thequestion, for what combinations of design and commu-nication cost will each model be viable?

3.3. Single-User InnovationConsider first a single-user innovatoran individual or afirmcontemplating investment in a design whose valueto her is vs . This measure of benefits includes all aspectsof the innovation that the user values. It may reflect, forexample, improved performance, lower cost, lower envi-ronmental impact, greater flexibility, and/or enhancedcapabilities. The effort of innovation is worthwhile (for


Figure 1 Bound on Single-User Innovation

Design cost (d)

Com

mun

icat

ion

costs

(c) Large-scalesingle-user

innovatorviable

d


Figure 2 Bound on Producer Innovation

Com

mun

icat

ion

costs

(c)

Design cost (d)

Producerinnovators

only

X not viable

X viableopportunity

Producersbound:

p*Q*=d+c

Single-user

onlyinnovators

Single-userinnovators

and

innovatorscoexist

producer

for the same opportunity. As we show in the figure, thedesign costs are higher than the value of the innovationto a single user; hence the single-user innovation modelis not viable for this design. We then show two possibleoutcomes for the producer. In the first case, communica-tion costs are low so that the sum of design and commu-nication costs falls below the producers bound. In thesecond case, the sum falls above the bound. Producerinnovation is a viable model for the first combination ofcosts but not for the second.

From this analysis, we learn three things. First,through the demand curve, a producers profit is deter-mined by its customers willingness to pay. Producersincentives are derived from and depend on users valu-ations. As a general rule in a free market, without usevalue there can be no profit. (Products whose purchase ismandated by the state or some other authority constituteexceptions to this rule.)

Second, like large-scale single-user innovators, pro-ducer innovators are affected by the size of the marketfor their goods. In large markets, the producer will havemany customers; thus its net revenue (after productionand transaction costs) will generally be far in excess ofthe value of the product to any one customer. Becausethe producer can aggregate demand, it can invest inboth product and process innovations that its own cus-tomers (acting as single-user innovators) would not findattractive.

Finally, the need to communicate differentiates pro-ducer innovators from single-user innovators. As wasmentioned earlier, to learn about which designs andgoods will be profitable to sell, producers must com-municate with and learn from their potential customersvia marketing research. To sell goods, a firms potentialcustomers have to be told about the innovative and mer-itorious features of the firms products and services. Thepercentage of cost devoted to marketing expenses is

a rough measure of what a given producer spends com-municating with customers. In effect, a producer innova-tor must split its (net) revenue between design cost andcommunication cost and still have something left over.Thus, if communication costs fall because of technolog-ical progress, a producer innovator may become viableeven if design cost stays the same.

3.5. Open Collaborative InnovationConsider finally the model of open collaborative innova-tion. Recall that open collaborative innovation projectsinvolve users and others who share the work of generat-ing a design and also reveal the outputs from their indi-vidual and collective design efforts openly for anyone touse. In such projects, some participants benefit from thedesign itselfdirectly in the case of users, indirectly inthe case of suppliers of complements that are increasedin value by that design. Each of these incurs the cost ofdoing some fraction of the work but obtains the valueof the entire design, including additions and improve-ments generated by others. Other participants obtain pri-vate benefits such as learning, reputation, fun, etc., thatare not related to the projects innovation outputs.

For ease of exposition, we will derive the bounds ofthe model for user innovators first and then consider theimpact of other participants on those bounds.

For the contributing user innovators, the key advan-tage of open collaborative innovation is that each con-tributor can undertake some of the work but rely onothers to do the rest (von Hippel and von Krogh 2003,Baldwin and Clark 2006b). The ability to divide updesign tasks via modular design architectures eliminatesthe design cost bound, d < vs , that made large-scaleinnovations infeasible for single-user innovators.

Creating a modular architecture that supports distri-buted innovation across geographical and organizationalboundaries will add to the upfront cost of a new designif detailed and comprehensive modularity is designed infrom the start. This was famously the case, for example,for IBMs System/360 and Pirellis modular integratedrobotized system for tire production (Baldwin and Clark2000, Brusoni and Prencipe 2006). However, modulararchitectures need not add significantly to costs if, asis typical modern practice, modularity is largely emer-gent. In such cases, the original architect designs a smallfunctional core and a set of interfaces. New modulescan then be attached to the system even if they werenot originally envisaged by the architects, as long asthey comply with the interface specifications. Examplesof this type of architecture include the IBM personalcomputer and the Linux operating system (Fergusonand Morris 1993, Raymond 1999). Emergent modular-ity is compatible with the constraints of open collabora-tive projects, where the design cost of each contributor(including project architects) needs to be relatively low.


Communication costs are a major concern for opencollaborative innovation projects. To divide their workeffectively and then to put it back together to form acomplete design, contributors must communicate withone another rapidly and repeatedly. This means that lowcommunication costs, as recently enabled by the Inter-net, are critical to the viability of the open collaborativeinnovation model.

User innovators will choose to participate in an opencollaborative innovation project if the increased commu-nication cost each incurs by joining the project is morethan offset by the value of designs obtained from others.To formalize this idea, assume that a large-scale inno-vation opportunity is perceived by a group of N com-municating designers. As rational actors, each memberof the group (indexed by i) will estimate the value ofthe large design and parse it into two subsets: (1) thepart, valued at vsi, that the focal individual can completehimself at a reasonable cost (by definition, vsi > dsi5;and (2) the part, valued at voi, that would be nice tohave but that he cannot complete at a reasonable costgiven his skills and other sticky information on hand (bydefinition, voi doi).

We assume that member i has the option to com-municate his portion of the design to other membersand receive their feedback and complementary designsat a cost ci. It makes sense for i to share his designsif he expects to receive more value from others thanhis communication cost. His expected benefit fromcommunicating can be parsed into (1) the probability jthat member j will respond in kind, (2) the fraction jof the remaining design that member j can provide, and(3) the total value voi that i can obtain from others contri-butions. As a rational actor, member i will communicatehis design to the other members of the group if

N1j 6=i

j j voi > ci0 (4)

This is the first bound on the open collaborative inno-vation model. It establishes the importance of communi-cation cost and technology for the viability of the opencollaborative model of innovation. The lower the costof communicating with the group, the lower the thresh-old other members contributions must meet to justifyan attempt to collaborate. Higher communication costsaffect inequality (4) in two ways: they increase the directcost of contributing (ci), and they reduce the probabilitythat others will reciprocate (j ). It follows that if com-munication costs are high, an open collaborative projectcannot get off the ground. However, if communicationcosts are low for everyone, it is rational for each mem-ber of the group to contribute designs to the general pooland expect that others will contribute complementarydesigns or improve on his own design. This is in fact thepattern observed in successful open source projects and

other forums of open collaborative innovation (Raymond1999, Franke and Shah 2003, Baldwin et al. 2006).

Note that open collaborative projects do not as a gen-eral rule discourage or prevent free riding. This featureof their makeup can be understood in light of the factthat the cost of screening or other protective measures toexclude free riders would raise communication costs andthus shrink the pool of potential contributors and, hence,the overall scale of the project. The network propertiesof the open collaborative model (the fact that the valueto everyone increases as the total number of contribu-tors increases) mean that this reduction in the contributorpool would reduce the value of the project to the con-tributors that remain as well as to free riders (Raymond1999, Baldwin and Clark 2006b, Baldwin 2008).

The second bound determines the maximum scale ofthe design. If there are N members of the group andeach contributes his or her own part, the total designinvestment will be the sum of their individual designcosts. The upper bound on design cost is then

Ni=1

dsi da + ca0 (4)Other things equal, this bound is more likely to be sat-

isfied if the ancillary contributors communication costsare low. Thus, communication costs constrain nonuserparticipants as well as users.

The presence of ancillary contributors further relaxesthe upper bound on the scale of the design. If thereare M ancillary contributors in addition to N users, theupper bound on total design value is

Ni=1

dsi +Ma=1

da


where va ca is value of participating (net of communi-cation cost) to the average ancillary contributor. Thus thescale of an open collaborative project is expandedandmay be greatly expandedby attracting ancillary con-tributors who value learning, fun, reputation, etc., morethan using the design itself.

All in all, the two bounds indicate both the limitationsand the possibilities associated with the open collabo-rative innovation model. The first bound (Equations (4)and (4)) shows that this mode of innovation is severelyrestricted by communication costs. If the value of theother part of the system is low or the expectation thatothers will actually contribute is low relative to the costof communication, single-user innovators will stick totheir knitting and not attempt to collaborate, and ancil-lary contributors will find some other outlet for theirtalents. But if communication costs are low enough toclear these hurdles, then the second bound (Equations (5)and (5)) shows that, using a modular design architec-ture as a means of coordinating their work, a collab-orative group can develop an innovative design that ismany times larger in scale than any single member ofthe group could manage alone.

Figure 3 places the three models of innovationsingle-user, producer, and open collaborativein thesame figure. The shadings and text in the figure indicateareas in which one, two, or all three models are viable.Basically, single-user innovation is viable when designcosts are low for any level of communication cost. Opencollaborative innovation is viable when communicationcosts are low for high levels of design cost, as long asthe design can be divided into modules that one or afew contributors can work on independently. Producerinnovation is viable when the sum of design and com-munication costs falls below the producers expected netrevenue as indicated by the negative 45 line.

3.6. Bringing Back Production andTransaction Costs

At the beginning of this section, to focus on the con-trasting effects of design and communication costs onthe three models of innovation, we made the simplifyingassumption that production costs and transaction costswere similar across all three and thus had no effect onany models viability relative to the other two. We didthis by defining the value of the design (vi) as the totalvalue of the innovation to the innovator (Vi) minus thecosts of production and transactions:

vi Vi ui ti0 (6)(Subscripts indicate that values and costs may differacross individuals and models.)

From this definition it is clear that if production costsor transaction costs are systematically higher for a par-ticular model of innovation, then for the same willing-ness to pay (Vi) there will be less value in the design (vi)

Figure 3 Bounds of Viability for All Three Innovation Models

Producerinnovators

only

innovatorsSingle-user

Com

mun

icat

ion

costs

(c)

Design cost (d)

only

innovatorsand

producerinnovators

Single-user

coexist

All threetypes are

viableOpen collaborative

innovators and producerinnovators compete

Open collaborativeinnovators only

to cover the upstream costs of design and communi-cation. The range of viability for the model with highercosts is then reduced. In terms of the bounds derivedabove, the single-user innovators bound would moveto the left, the producers bound would move towardthe origin, and the open collaborative projects boundswould move both down and left.

We now consider whether there are systematic differ-ences in production or transaction costs across the threemodels.

3.6.1. Production Costs. At the start of this section,we explained that a design is the information requiredto produce a novel product or servicethe recipe.For products that themselves consist of information suchas software, production costs are simply the cost ofcopying and instantiating the design. For digitized prod-ucts and services, these costs are now very low. In thecase of a physical products, however, the design recipemust be converted into a physical form before it can beused. In such cases, the input is the design instructions(the recipe) plus the materials, energy, and human effortspecified in those instructions; the output is a good (thedesign converted into usable form).

One of the major advantages producers have histori-cally had over single-user innovators and open collab-orative innovation projects is economies of scale withrespect to mass production technologies. Mass produc-tion, which became widespread in the early 20th century,is a set of techniques whereby certain physical productscan be turned out in very high volumes at very low unitcost (Chandler 1977, Hounshell 1985). The economiesof scale in mass production generally depend on usinga single design (or a small number of designs) over andover again. In classic mass production, changing designsinterrupts the flow of products and causes setup andswitching costs, which reduce the overall efficiency of


the process. There is no room for variety, as indicatedby Henry Fords famous quote, [A] customer can havea car painted any color 0 0 0 so long as it is black (Ford1922, Chapter 4, p. 34).

Can single-user innovators or open collaborative inno-vation projects convert their various designs into physi-cal products that will be economically competitive withthe products of mass producers? Increasingly, the answeris yes. Consider that, today, modularization is affect-ing the interface between design and production as wellas the interfaces between design tasks. This means thatmass producers can design their production technolo-gies to be independent of many of the specifics of thedesigns they produce. Such processes are said to pro-vide mass customization (Pine 1993, Tseng and Piller2003). When mass customization is possiblethat is,when particular designs are no longer for technical rea-sons tied to production technologiesproducers can inprinciple make their low-cost, high-throughput factoriesavailable for the production of designs created by singleusers and collaborative open projects.

Some producers might resist this idea, wanting to cap-ture profits from proprietary product designs as well asproprietary production capabilities. But if there is com-petition among producers, some will be willing to pro-duce outside designs as well as their own and forgo therents they formerly obtained from proprietary designs.Indeed, this possibility is manifest in many industrieswhere toll production is common. For example, sil-icon fabs produce custom designs to order via verysophisticated and expensive production processes, as doproducers of specialty chemicals and contract manufac-turers of consumer electronic goods (Sturgeon 2002).

Nevertheless, for a long time to come, there willcontinue to be instances where economies in massproduction significantly depend on careful and subtlecodesign of products and product-specific productionsystems. In such instances, we expect producer innova-tors to continue to have an advantage in designing andproducing goods and services for mass markets.

3.6.2. Transaction Costs. If producer innovatorshave a production cost advantage for some (but not all)production technologies, free-revealing single-user andopen collaborative innovators have an advantage withrespect to transaction costs. As indicated, the transac-tion costs of innovation include the cost of establishingexclusive rights over the innovative design, for example,through secrecy or by obtaining a patent. Also includedare the costs of protecting the design from theft, forexample, by restricting access, and enforcing noncom-pete agreements (Teece 2000, Marx et al. 2009). Finally,transaction costs include the costs of legally transfer-ring rights to the good or service embodying the innova-tion, receiving compensation, and protecting both sidesagainst opportunism (Williamson 1985, Baldwin 2008).

Producer innovators must incur transaction costs. Bydefinition, they obtain revenue and resources from com-pensated exchanges with users, employees, suppliers,and investors. A considerable amount of analysis in thefields of economics, management, and strategy consid-ers how to minimize transaction costs by rearranging theboundaries of firms or the structure of products and pro-cesses. (For reviews of this literature, see Williamson2000, Lafontaine and Slade 2007.) The bottom line isthat for producer innovators, transaction costs are aninevitable cost of doing business.

Single-user innovators, including process innovators,incur transaction costs when they seek to assert exclu-sive rights over their innovative designs. Patents on inter-nal processes and equipment, the enforcement of secrecyand need-to-know policies within a firm, and non-compete agreements with key employees are all visibleevidence of transaction costs that single-user innova-tors incur to protect valuable intellectual capital. In suchcases, as rational actors, single-user innovators wouldhave to find a net gain after subtracting both design andtransaction costs from the expected value of an innova-tive design to themselves.

However, single-user innovators have a choice as towhich innovations are worth protecting and which arenot. As discussed in the literature review, empiricalresearch suggests that single-user innovators generallydo not treat all or even most of their innovations as valu-able property that must be sequestered within their walls.They often find it more practical and profitable to freelyreveal their designs to achieve network effects, reputa-tional advantages, and other benefits and/or to avoid thecost of protecting their innovations. For example, manyuser firms that develop process equipment innovationsfor in-house use freely reveal them to get an outsidesource of equipment production and improvements fromothers (Harhoff et al. 2003, de Jong and von Hippel2009). This free revealing decreases as the importance ofprocess innovation as a source of competitive advantageincreases (Franke and Shah 2003, Raasch et al. 2008).By definition, when single-user innovators freely revealan innovation, they do not incur transaction costs, andthe region of viability for the innovation opportunity isthereby expanded.

Open collaborative innovation projects do not sellproducts, nor do they pay members for their contribu-tions. In this respect, they do not incur transaction costs.However, when an open collaborative project becomeslarge and successful, its members generally find thatthey must incur costs to protect the now-valuable designfrom malfeasance and expropriation. For example, vir-tually all large open source projects have a system ofhierarchical access that prevents anyone from changingthe master copy of the source code without authoriza-tion by a trusted member of the project. The General


Public License (GPL) was explicitly designed to pro-tect the rights of users to view, modify, and distributecode derived from the licensed code (Stallman 2002,OMahony 2003). The costs of restricting access andenforcing the GPL are like classic transaction costs inthat they assert and enforce property rights to preventvandalism and theft.

Notwithstanding these necessary expenditures, opencollaborative innovation projects do avoid the mun-dane transaction costs of defining, counting, and payingfor goods in formal legal transactions (Baldwin 2008).Their contributors do not have to figure out what tosell, how much to charge, or how to collect paymentcostly activities that producers must perform in the nor-mal course of business. In this respect, free-revealingsingle-user innovators and open collaborative innovationprojects have a transaction cost advantage over producerinnovators.

Regulation can be viewed as a form of transactioncost imposed by the government on all three innovationmodels. Drugs, commercial aircraft, and automobiles areamong the product types that must meet heavy safety-related regulatory burdens before being allowed to enterthe marketplace. Regulation in the form of standard set-ting affects many other industries, such as telecommuni-cations. Within our theoretical framework, regulation andstandard setting tend to decrease the value of innovationopportunities, thus shrinking the bounds of viability. InFigure 3, all three bounds will move down and to theright, and the areas of viability will become smaller.

4. DiscussionAs we said at the start of this paper, there is a widespreadand longstanding perception among academics, policymakers, and practitioners that innovation by producers isthe primary mode of innovation in market economies. Inthis view, innovations are undertaken by firms that canaggregate demand or not at all. In the 1930s, Schumpeter(1934, p. 65) placed producers at the center of his theoryof economic development, saying, It is 0 0 0 the producerwho as a rule initiates economic change, and consumersare educated by him if necessary. Sixty years later,Teece (1996, p. 193) echoed Schumpeter: In marketeconomies, the business firm is clearly the leading playerin the development and commercialization of new prod-ucts and processes (see also Teece 2002, p. 36). Atabout the same time, Romer (1990, p. S74) made thepredominance of producer innovation the basis for hismodel of endogenous growth: The vast majority ofdesigns result from the research and development activ-ities of private, profit-maximizing firms. And Baumol(2002, p. 35) placed producer innovation at the center ofhis theory of oligopolistic competition: In major sec-tors of U.S. industry, innovation has increasingly grownin relative importance as a instrument used by firms

to battle their competitors. However, like all humanendeavors, the organizations and institutions that createinnovations are historically contingent. They are solu-tions to the problems of a specific time and place usingthe technologies of that time and place. It is the casethat throughout most of the 20th century, single-user andopen collaborative innovation were extant, but central-ized R&D, product development, and process engineer-ing groups within firms were the most economical wayto design mass-produced products and related produc-tion processes (Chandler 1977, Hounshell 1985).

Four technological factors contributed to the preem-inence of mass-produced products, and thus the dom-inance of producer and large-scale process innovatorsin technologically advanced economies in the early andmiddle parts of the 20th century. First, computationalresources were scarce, and therefore the cost of creatingindividual designs was quite high. Second, as discussedabove, there was generally a close tie between designof items to be produced and the complex requirementsof process technologies. Third, modular design methodswere not well understood and seldom implemented. Andfourth, cheap, rapid communication enabling distributeddesign among widely separated participants in a designprocess was not technically possible. Taken together,these factors made it cheaper to design standardized anduniform products centrally and in conjunction with theirmanufacturing processes. Given these conditions, it isreasonable to speculate that Schumpeter and later Teece,Romer, and Baumol were simply observing the most vis-ible innovation processes of their times when they statedthat producers (business firms) were the leading devel-opers of innovation in market economies.

Today, as was mentioned earlier, conditions facingwould-be innovators are changing rapidly and radically.Just as the rise of producer innovation was enabled by in-terdependencies between centralized product design andthe technologies of mass production, today the rapidgrowth of single-user and open collaborative innovationis being assisted by technologies that both enhance thecapabilities of individual designers and support distri-buted, collaborative design projects. These technologiesinclude powerful personal computers; standard designlanguages, representations, and tools; the digitization ofdesign information; modular design architectures; andlow-cost any-to-any and any-to-all communication viathe Internet. Of course, we should remember that theinstitutions of single-user and open collaborative inno-vation have long existed (Rosenberg 1976, von Hippel1976, Shah 2005). However, they are growing moreprominent today because of the largely exogenous tech-nological developments just mentioned.

Technological trends suggest that both design costsand communication costs will be further reduced overtime. To visualize this effect, imagine Figure 3 beingpopulated with numerous points each representing an


innovation opportunity. As design and communicationcosts fall, each point would move down and to the left.As a result of this general movement, some points wouldcross the thresholds of viability for single-user and opencollaborative innovation. Conversion of some designsfrom small-scale to mass production would cause somepoints to move in the opposite direction, against the gen-eral trend. But for the most part, technological progressalong both dimensions of cost will have the effect ofmoving whole classes of innovation opportunities fromthe regions where only innovation by firms able to aggre-gate demand is viable to regions where single-user inno-vation or collaborative innovation are also viable. Inthese cases, what was previously a dominant modelthe only feasible way to cover the costs of innovationbecomes subject to competition from other, newly viablemodels. This means that producer innovators increas-ingly must contend with single-user innovators and opencollaborative innovation projects as alternative sourcesof innovative products, processes, and services.

The declining cost of computation most directlyaffects the cost of designs that can be developed andtested on computersand today, one is hard-pressed tothink of a type of design that does not fall into thisdomain. Of course, computer-based design is central tothe design of software. Increasingly, it is also centralto the development of hardware-embodied designs rang-ing from consumer products to buildings. Design pro-cesses for both software and hardware generally requiredesigners to represent different states of the design usingtext, pictures, and models; store and analyze data; com-pute rates, bounds, and tolerances; and simulate behav-ior under various conditions. Technologies that automatethe tasks of writing, drawing, modeling, data storage andanalysis, computation, and simulation all have the effectof reducing design cost.

The declining cost of communication most directlyaffects the design of products and services where designis informed by community trial-and-error learning. Thedesign of physical games and the equipment used forthem are an example, because the final shape of suc-cessful games only emerges via collaborative play andrelated redesign. Wikis like Wikipedia are an example ofopen collaborative innovation in the world of text: theycoexist and compete with producer-designed compendia.Social networking sites like Facebook also have featuresof both single-user innovation (each person designs herpage) and open collaborative innovation (she and herfriends contribute content to each others pages).

Although not all designs are equally affected, webelieve declining computation and communication costsare having enough of an impact across the economyto change the relative importance of the three differentmodels of innovation discussed previously.

4.1. Interactions Between the Three ModelsFrom Figure 3 it is evident that for some combinationsof design and communication costs, two or even all threemodels of innovation will be viable. How will the pres-ence of one influence the other(s)? In other words, howwill the models interact? Prior research allows us toelaborate on this basic matter in several interesting ways,as we discuss next.

When single-user innovation and producer innovationare both viable, the single-user innovators must eval-uate an innovation opportunity, not only in relation totheir design cost but also in relation to the producersproduct and price. If the producer offers a good-enoughproduct at a low-enough price, purchasing the innovationmay dominate developing it in-house, and some poten-tial single-user innovators may switch to becoming cus-tomers of the producer. (This happens regularly whencompanies switch from custom software developed byan in-house IT department to off-the-shelf, purchasedsoftware.) However, to attract users who can innovateon their own, the producers price must be less thanthe users design cost, which by definition is less thanthe users value: p < ds < vs . Given differentiated users,rational producers are likely to target as customers userswith high design costs and leave those with low designcosts to work out their own solutions.

Indeed, because of their distinct roles, producer inno-vators and single-user innovators may develop a sym-biotic relationship. Empirical studies have shown thatmost single-user innovation is done by a subset of allusers called lead users that are ahead of the bulk ofthe market with respect to an important trend and alsohave a high incentive to innovate to solve the needsthey encounter at the leading edge (von Hippel 1986).Some of these lead users have no interest in commer-cializing their innovations. However, their innovationsmay nonetheless serve as an attractive feedstock of field-tested product prototypes for producers. By monitoringand incorporating lead-user innovations into their ownofferings, producer innovators may enhance their prod-uct and service offerings while at the same time reducingtheir design costs and increasing their likelihood of suc-cess in the marketplace (Lilien et al. 2002, von Hippel2005). In other instances, individual lead users mayfound companies for the purpose of commercializingtheir designs. Such firms, which generally serve marketniches not large enough to attract established firms, areoften the first to introduce new products into the econ-omy (Baldwin et al. 2006, Shah and Tripsas 2007).

Open collaborative innovation projects are more likelyto pose a threat to producer innovators than are single-user innovators. In the first place, by attracting effort anddividing up tasks, collaborative projects can in princi-ple match the scale of a producers design. Second, asa matter of policy and to reduce transaction costs, opencollaborative projects make their designs available to all


comers at little or no charge. The existence of a freedesign, even if it must be adapted by end users, putsprice pressure on a producer innovator. Indeed, usershave incentives to collaborate for this purpose preciselyin those cases where the producer innovator can deterother producers from entering its market. If the price ofa design will collapse on entry of a second contender,no profit-seeking producer will find the second-in oppor-tunity attractive. In contrast, users directly gain fromany price drop and hence will benefit by supporting acollaborative project aimed at breaking the producersmonopoly (Baldwin and Clark 2006b). Supporting thislogic, some open source projects have been founded withthe aim of preventing or breaking producers monopo-lies. The most famous is the GNU Project, begun byRichard Stallman in 1984 for the express purpose of pro-viding a free alternative to commercially owned software(Stallman 2002).

However, the openness and modularity of open col-laborative projects make pure head-to-head competitionwith producers an unlikely end result. Once an open col-laborative project has been started, producer innovatorscan adapt their own strategies in response. The monop-olist challenged by the project may withdraw to parts ofthe market that are locked in, are not price sensitive, ordemand high levels of service (Casadesus-Masanell andGhemawat 2006). Alternatively, producers may becomecontributors to open collaborative projects for whichthey supply complements. Thus, in 1999, IBM becamean important supporter of and contributor of code toLinux. IBM sells products and services that complementLinux, ranging from computers to proprietary softwareto consulting services. Similar examples are legion, suchas Googles support for an open source software stackfor mobile devices (Android).

4.2. Hybrid Innovation ModelsA hybrid innovation model combines elements ofthe three polar models analyzed in previous sections.Hybrids of the three basic models thrive in the realworld. This is because the architecture of a design toachieve a given function can often take a number offorms suited to development by combinations of ourthree basic models. For example, producers or users canchoose to modularize a product architecture into a mix oflarge components that can today best be created by pro-ducers, plus many smaller components suited for devel-opment by single-user innovators or open collaborativeinnovation projects. Thus, Intel develops expensive andcomplex central processing unit chips for computers.Complementary software and hardware designs are thendeveloped by for-profit producers, by single-user innova-tors, and by open collaborative projects. Another exam-ple is the development of software engines for com-puter games by producer firms upon which platforms

individual gamers or groups of gamers acting collabora-tively develop mods (Jeppesen 2004).

Large indivisible design projects, which have tradi-tionally been in the producer-only zone of Figure 3, maybecome hybrids as a result of the rearchitecting of tradi-tional, producer-centered design approaches. For exam-ple, drug development costs are commonly argued tobe so high that only a producer innovator, buttressedby strong intellectual property protection for drugs, cansucceed. Increasingly, however, we are learning how tosubdivide drug trialsa large cost traditionally borneby drug producersinto elements suitable for voluntary,unpaid participation by users acting within a collabo-rative open innovation framework. This possibility hasrecently been illustrated in a trial of the effects of lithiumon amyotrophic lateral sclerosis (ALS; Lou Gehrigs dis-ease) carried out by ALS patients themselves with thesupport of a toolkit and website developed by the firmPatientsLikeMe.

Innovation platforms and the innovations appended tothem are often a hybrid of single-user innovation, pro-ducer innovation, and/or collaborative open innovation.Innovation platforms are components that provide a sta-ble framework or binding surface that serves to sup-port and organize the innovation contributions of manycomplementors (Gawer and Cusumano 2002, Gawerand Henderson 2007, Baldwin and Woodard 2009,Gawer 2009). Platforms can range from interface stan-dards such as an application programming interface or ascrew thread specification, to open source software plat-forms like Apache or Linux, to social networking siteslike Facebook. In the case of Apache, the platform isan open, collaboratively built one, and appended innova-tions are developed by innovators representing all threeof our polar models. In contrast, Facebook and YouTubeare producer-built and producer-owned platforms, andappended creative content is generated primarily by indi-vidual users. Innovation platforms can be a great sourceof profit for their owners when entry costs are low andnetwork effects are strong. Indeed, under such conditionsplatform owners often face a winner-take-all situationand so vie fiercely to attract free content to leveragetheir internal resources, attain category leadership, andthereby improve their financial performance (Gawer andCusumano 2002).

Closed collaborative innovation, often termed crowd-sourcing, is a hybrid of open collaborative innovationand producer innovation. In this hybrid model, a pro-ducer innovator poses a problem, solicits proposed solu-tions from numerous third parties (the crowd) and thenselects the best solution or combination. Members of thecrowd do not see nor do they have rights to use theproposed solutions: the outputs are closed and ownedby the sponsor (Howe 2006, Pisano and Verganti 2008,Jeppesen and Lakhani 2010).


To understand closed collaborative innovation, con-sider that there are two reasons to open up a collabo-rative project. In the first place, open access to the out-puts of a collaborative project acts as an inducementto single-user innovators to join and contribute insteadof innovating on their own (Baldwin and Clark 2006b).In the second place, when effective problem solvingrequires contributors to know and understand the solu-tion being developed, open access is the low-cost defaultsolution.

Sponsors of collaborative projects can close and ownthe innovative output of a collaborative project if theycan escape these two constraints. To escape the first,sponsors can create incentives that will attract nonusercontributors to their project. For example, they can offerpayment or process-related rewards such as learning orfun (Raymond 1999, Lerner and Tirole 2002, Lakhaniand Wolf 2005, Benkler 2006). To escape the secondconstraint, project sponsors can employ an extreme formof modularity in which no participant knows (or needsto know) what the others are doing, and only the spon-sor sees everything. Examples of closed collaborativeprojects are design contests such as the Netflix Prize orthe Cisco I-Prize, in which many contestants competefor a monetary prize and reputational gains (Jeppesenand Lakhani 2010, Boudreau et al. 2011).

Like open collaborative projects, hybrids generallyrely on an underlying modular architecture to providethe basis for splitting up design tasks among vari-ous participants. The modular architecture also makesit possible to create a structure of intellectual prop-erty rights so that producer innovation, single-user inno-vation, and open collaborative innovation can all takeplace within the same technical framework (Henkel andBaldwin 2010).

4.3. Suggestions for Further ResearchIn this paper, we have argued that a paradigm shiftis occurring in our understanding of innovation. Kuhn(1962, p. 103) writes,

[Paradigms] are the source of the methods, problem-field,and standards of solution accepted by any mature scien-tific community at any given time. As a result, the recep-tion of a new paradigm often necessitates a redefinitionof the corresponding science. Some old problems may berelegated to another science or declared entirely unsci-entific. Others that were previously nonexistent or trivialmay, with a new paradigm, become the very archetypesof significant scientific achievement.

Taken in combination, we think the patterns and find-ings we have described represent a significant change inthe problem field of innovation research, policy mak-ing, and practice. In brief recapitulation, we have foundthat technological progress is moving whole classes ofinnovation opportunities from the region where only

producer innovation is viable to regions where single-user innovation or open collaborative innovation is alsoviable. In these cases, what was previously a domi-nant modelthe only feasible way to cover the costs ofinnovationbecomes subject to competition from other,newly viable models.

In the course of explaining the viability of these addi-tional innovation models, we have also uncovered a chal-lenge to two fundamental assumptions in prior work oninnovation. Recall that, since the time of Schumpeter,the preeminence of producer innovation as well asthe need for intellectual property rights to enable pro-ducer innovators to protect their rents have gone largelyunquestioned by scholars and policy makers alike. Thesetwo assumptions have deeply permeated academic schol-arship, policy making, and practice in many fields. Bothassumptions are now challenged by the viability of thesingle-user and open collaborative innovation models wehave described in this paper. Users are now seen to bean important source of innovation, and value to usersis seen as an alternate motive to profits for investing ininnovative designs. The net result is, we think, to greatlychange the problem field in innovation research, policymaking, and practice.

With respect to research, consider that one of the sig-nal accomplishments in economics in the 1990s wasto incorporate rent-seeking investment into macroeco-nomic growth models, thereby endogenizing technolog-ical change at the level of the economy (Romer 1990,Aghion and Howitt 1998). In its most common inter-pretation, endogenous growth theory supports politi-cal and legal moves to strengthen intellectual propertyrights. The basic argument is that without the abilityto exclude others from using new designs, incentives toinvest in innovation would disappear, and in the absenceof such investments, technological progress and eco-nomic growth would not occur.

Recall that our analysis suggests that this view is toostark. We have shown that users working alone, or inthe context of open collaborative projects and in con-junction with ancillary contributors, have the potential tosupply innovations for their own direct benefit withoutthe inducement of excludability or monopoly. Such userscannot themselves be strong rivals, but the nonrivalrousnature of consumers is a longstanding implicit assump-tion in most of economics. (If a firm can sell goods withthe same design to many users, by definition, those usersdo not need to exclude other users from access to thedesign. Such users are excellent prospects for forminga collaborative project to supply the design without theintermediation of a monopolistic producer innovator.)

Thus, one avenue for future theoretical research is todevelop micro- and macroeconomic models that allowfor the possibility of nonrivalrous user innovation as wellas producer innovation. Another is to explore further the


technological and institutional factors needed for col-laborative nonrivalrous innovation to become an impor-tant contributor to technological progress and economicdevelopment.

With respect to policy making, an area that clearlymust be reviewedand that would strongly benefit fromresearch along the lines described aboveis intellec-tual property rights. From the time of the Enlighten-ment, many have held the view that providing inventorswith incentives in the form of property rights to theirwritings and discoveries would induce them to investin the creation of useful new ideas, i.e., innovations.Of course, it was also known that grants of intellectualproperty rights would create undesirable monopolies.Producers create deadweight losses when they exploitintellectual property rights to reap monopoly profitsand spend money to protect or extend their monopolypositions (Machlup and Penrose 1950, Penrose 1951,MacLeod 2007). Value in use as an alternative incentivefor single-user innovators and participants in open col-laborations indicates that there can be ways to supportrobust innovation without the devils bargain inherentin the granting of intellectual property rights.

Today, essentially all national governments supportcostly intellectual property rights infrastructures to sup-port inventors who wish to restrict access to their inno-vations. At the same time, governments have done verylittle to create an infrastructure to support inventorsand innovators who may wish to practice open inno-vation. The result is that open innovators are forcedto operate within an framework of intellectual prop-erty rights designed for closed innovators (Strandburg2008). This framework imposes significant costs on openinnovation: because innovation-related information canbe owned, all developers and users of such informa-tion must conduct extensive and expensive searches forpotential ownersand still run the risk of litigation fromundiscovered ownersbefore they can freely utilize orreveal what they know. In effect, public policy makingtoday has created a nonlevel playing field between openand closed innovation (Dreyfuss 2010).

Research is urgently needed into the basic matter ofthe social welfare effects of closed versus open inno-vation to establish the basis for an appropriate balance.If, as we expect, increased support for open innova-tion is found to be desirable from the perspective ofsocial welfare, then expansion of fair use rights andother forms of safe harbors for those seeking to freelyuse and reveal innovation-related information will beamong the types new policies that should be considered(Strandburg 2008).

With respect to corporate practice of innovation,research should be pursued on new business models thatcreate private profits for producers without proprietarycontrol of innovative product and service designs. Crea-tion of new product and service designs is a cost for

producers. Early evidence shows there are ways for pro-ducers to thrive by adopting open innovation designsand shifting to profit seeking based on other forms ofprivate advantage, such as lead time, high-quality ser-vice, and superior distribution channels (Raymond 1999,von Hippel 2005). Of course, producers will predictablyresist open innovation until they have made successfultransitions to these new business models. Energetic pur-suit of practitioner-oriented research can aid and speedup these needed transitions. In addition, when and whereeach organizational form is most efficacious is an impor-tant question for both theory and practice in innovationmanagement.

The paradigm shift we describe from producer innova-tion to user and open collaborative innovation offers bothinescapable challenges and opportunities to researchers,to policy makers, to firmsindeed, to all of us whohave a stake in innovation. We think that both personalfreedoms and social welfare will increase as a result ofthis shift. We suggest that further explorations will bevaluable.

AcknowledgmentsThe authors thank Jeroen de Jong, Joachim Henkel, KarimLakhani, Anita McGahan, Eric Raymond, Steven Roper,Suzanne Scotchmer, Scott Stern, and Michael Tushman, aswell as three anonymous reviewers and the associate editor,Lori Rosenkopf, for comments and suggestions that led to sig-nificant improvements of this paper.

ReferencesAghion, P., P. Howitt. 1998. Endogenous Growth Theory. MIT Press,

Cambridge, MA.

Alchian, A. A., H. Demsetz. 1972. Production, information costs, andeconomic organization. Amer. Econom. Rev. 62(5) 777795.

Allen, R. C. 1983. Collective invention. J. Econom. Behav. Organ.4(1) 124.

Aoki, M. 1988. Information, Incentives, and Bargaining inthe Japanese Economy. Cambridge University Press, Cam-bridge, UK.

Aoki, M. 2001. Toward a Comparative Institutional Analysis. MITPress, Cambridge, MA.

Arundel, A., V. Sonntag. 1999. Patterns of advanced manufactur-ing technology (AMT) use in Canadian manufacturing: 1998AMT survey results. Report 88F0017MIE, 12, Statistics Canada,Ottawa, ON, Canada.

Baldwin, C. Y. 2008. Where do transactions come from? Modular-ity, transactions and the boundaries of firms. Indust. CorporateChange 17(1) 155195.

Baldwin, C. Y., K. B. Clark. 2000. Design Rules: The Power of Mod-ularity, Vol. 1. MIT Press, Cambridge, MA.

Baldwin, C. Y., K. B. Clark. 2006a. Between knowledge and theeconomy: Notes on the scientific study of designs. B. Kahin,D. Foray, eds. Advancing Knowledge and the Knowledge Econ-omy. MIT Press, Cambridge, MA, 299328.

Baldwin, C. Y., K. B. Clark. 2006b. The architecture of participation:Does code architecture mitigate free riding in the open sourcedevelopment model? Management Sci. 52(7) 11161127.


Baldwin, C. Y., C. J. Woodard. 2009. The architecture of platforms:A unified view. A. Gawer, ed. Platforms, Markets and Innova-tion. Edward Elgar, London, 1944.

Baldwin, C. Y., C. Hienerth, E. von Hippel. 2006. How user innova-tions become commercial products: A theoretical investigationand case study. Res. Policy 35(9) 12911313.

Baumol, W. J. 2002. The Free-Market Innovation Machine: Analyzingthe Growth Miracle of Capitalism. Princeton University Press,Princeton, NJ.

Benkler, Y. 2002. Coases penguin, or, Linux and The Nature of theFirm. Yale Law J. 112(3) 369447.

Benkler, Y. 2006. The Wealth of Networks: How Social Produc-tion Transforms Markets and Freedom. Yale University Press,New Haven, CT.

Blaxill, M., R. Eckhardt. 2009. The Invisible Edge: Taking Your Strat-egy to the Next Level Using Intellectual Property. Portfolio,New York.

Boudreau, K. J., N. Lacetera, K. R. Lakhani. 2011. Incentives andproblem uncertainty in innovation contests: An empirical analy-sis. Management Sci. 57(5) 843863.

Brusoni, S., A. Prencipe. 2006. Making design rules: A multidomainperspective. Organ. Sci. 17(2) 179189.

Casadesus-Masanell, R., P. Ghemawat. 2006. Dynamic mixedduopoly: A model motivated by Linux vs. Windows. Manage-ment Sci. 52(7) 10721084.

Chandler, A. D., Jr. 1977. The Visible Hand: The Managerial Rev-olution in American Business. Harvard University Press, Cam-bridge, MA.

Chesbrough, H. 2003. Open Innovation: The New Imperative for Cre-ating and Profiting from Technology. Harvard Business SchoolPress, Cambridge, MA.

Colfer, L. J., C. Y. Baldwin. 2010. The mirroring hypothesis: The-ory, evidence and exceptions. Working Paper 10-058, HarvardBusiness School, Boston.

Dasgupta, P., P. A. David. 1994. Toward a new economics of science.Policy Res. 23(5) 487521.

de Jong, J. P. J., E. von Hippel. 2009. Transfers of user processinnovations to process equipment producers: A study of Dutchhigh-tech firms. Res. Policy 38(7) 11811191.

Demsetz, H. 1988. The theory of the firm revisited. J. Law, Econom.Organ. 4(1) 141161.

Dosi, G., R. R. Nelson. 2010. Technical change and industrialdynamics as evolutionary processes. B. H. Hall, N. Rosenberg,eds. Handbook of the Economics of Innovation, Vol 1. North-Holland, Amsterdam, 51127.

Dreyfuss, R. C. 2010. Does IP need IP? Accommodating intellectualproduction outside the intellectual property paradigm. CardozoLaw Rev. 31(5) 14371473.

Enos, J. L. 1962. Petroleum Progress and Profits: A History of ProcessInnovation. MIT Press, Cambridge, MA.

Fama, E. F., M. C. Jensen. 1983a. Agency problems and residualclaims. J. Law Econom. 26(2) 327349.

Fama, E. F., M. C. Jensen. 1983b. Separation of ownership and con-trol. J. Law Econom. 26(2) 301325.

Ferguson, C. H., C. R. Morris. 1993. Computer Wars: How the WestCan Win in a Post-IBM World. Times Books, New York.

Foray, D. 2004. The Economics of Knowledge. MIT Press, Cam-bridge, MA.

Ford, H. 1922. My Life and Work. Garden City Publishing Company,Garden City, NY.

Franke, N., S. Shah. 2003. How communities support innovative activ-ities: An exploration of assistance and sharing among end-users.Res. Policy 32(1) 157178.

Franke, N., E. von Hippel. 2003. Satisfying heterogeneous user needsvia innovation toolkits: The case of Apache security software.Res. Policy 32(7) 11991215.

Gallini, N., S. Scotchmer. 2002. Intellectual property: What is the bestincentive system? A. Jaffe, J. Lerner, S. Stern, eds. InnovationPolicy and the Economy, Vol. 2. MIT Press, Cambridge, MA,5178.

Gault, F., E. von Hippel. 2009. The prevalence of user innovationand free innovation transfers: Implications for statistical indica-tors and innovation policy. Working Paper 4722-09, MIT SloanSchool of Management, Cambridge, MA.

Gawer, A. 2009. Platform dynamics and strategies: From productsto services. A. Gawer, ed. Platforms, Markets and Innovation.Edward Elgar, London, 4576.

Gawer, A., M. A. Cusumano. 2002. Platform Leadership: How Intel,Microsoft and Cisco Drive Industry Innovation. Harvard Busi-ness School Press, Boston.

Gawer, A., R. Henderson. 2007. Platform owner entry and innova-tion in complementary markets: Evidence from Intel. J. Econ

Documents

Von Hippel 2011