Selective Intervention and Internal Hybrids: Interpreting ... · Selective Intervention and Internal Hybrids: Interpreting and Learning from the Rise and Decline of the Oticon Spaghetti

Selective Intervention and Internal Hybrids:Interpreting and Learning from the Rise andDecline of the Oticon Spaghetti Organization

Nicolai J. FossLINK, Department of Industrial Economics and Strategy, Copenhagen Business School,

Solbjergvej 3; 2000 Frederiksberg, [email protected]

AbstractInfusing hierarchies with elements of market control hasbecome a much-used way of simultaneously increasingentrepreneurialism and motivation in firms. However, thispaper argues that such "internal hybrids," particularly in theirradical forms, are inherently hard to successfully design andimplement because of a fundamental incentive problem ofestablishing credible managerial commitments to not inter-vene in delegated decision making. This theme is developedand illustrated, using the case of the world-leading hearingaids producer, Oticon. In the beginning of the 1990s, Oticonbecame famous for its radical internal hybrid, the "spaghettiorganization." Recent work has interpreted the spaghetti orga-nization as a radical attempt to foster dynamic capabilitiesby organizational means, neglecting, however, that about adecade later the spaghetti organization has given way to amore traditional matrix organization. In contrast, an orga-nizational economics interpretation of Oticon organizationalchanges is developed. This lens suggests that a strong liabilityof the spaghetti organization was the above incentive prob-lem: Frequent managerial meddling with delegated rights ledto a severe loss of motivation, and arguably caused the changeto a more structured organization. Refutable implications aredeveloped, and the discussion is broadened to more generalissues of economic organization.(Internal Hybrids, Organizational Change; Delegation; Managerial Com-mitment Problems; New Organizationat Forms)

IntroductionIn academic research as well as in managerial prac-tice, the search for the sources of competitive advan-tage has increasingly centered on organization-relatedfactors (e.g., Barney 1986, Kogut and Zander 1992,Mosakowski 1998, Nahapiet and Ghoshal 1999). Thus,it is argued that many firms radically change the way inwhich they structure their boundaries (e.g.. Helper et al.2000) as well as their internal organization (e.g.. Miles

et al. 1997). They arguably do this in an attempt to fosterthe dynamic capabilities that are necessary for compet-ing in the emerging knowledge economy. Fundamentaladvances in IT and measurement technologies have facil-itated these changes (Zenger and Hesterly 1997), whileequally fundamental developments in the organizationand motives of capital markets as well as increasinginternationalization are claimed to have made them nec-essary (Halal and Taylor 1998).

From an organizational economics perspective, theseexperiments with economic organization fall into thecategories of either external hybrids (Williamson 1996)(that is, market exchanges infused with elements of hier-archical control), or internal hybrids (Zenger 2002) (thatis, hierarchical forms infused with elements of mar-ket control). The aims of the experimental efforts areto reduce coordination costs, improve incentives, andhelp to clarify the nature of the businesses the firm isin, thereby improving entrepreneurial capabilities andthe ability to produce, share, and reproduce knowledge(Grant 1996, Day and Wendler 1998, Miles et al. 1997,Mosakowski 1998). Although both internal and externalhybrids are means to reach these aims, they would seemto be highly imperfect substitutes. For example, adopt-ing an internal hybrid form has the benefit of involv-ing fewer layoffs relative to adopting external hybrids.Also, spin-offs, carve-outs and the like are often legallycomplex operations, whereas adopting an internal hybridmay simply be a matter of managerial fiat. Further, man-agement may fear that leaving too many activities inthe hands of other firms will hollow out the corporation(Teece et al. 1994), or make it difficult to protect valu-able knowledge (Liebeskind 1996). Given this, one maywonder why firms should ever make governance choicesin favor of external hybrids. However, a main point ofthis paper is that internal hybrids are beset by distinctincentive costs that external hybrids tend to avoid, and

1047-7039/03/1403/0331$05.001526-5455 electronic ISSN

ORGANIZATION SCIENCE © 2003 INFORMSVol. 14, No. 3, May-June 2003, pp. 331-349

NICOLAI J. FOSS Selective Intervention and Internal Hybrids

that this may explain why external hybrids are chosenover internal hybrids.

Research on new organizational forms is an emergingfield (Daft and Lewin 1993, Zenger and Hesterly 1997,Foss 2002), and rather little is known about the costsand benefits of these organizational forms.' This papermixes empirical observation with theoretical reasoning,mostly drawn from organizational economics, in order togain a better understanding of the organizational designproblems of internal hybrids. The theoretical empha-sis is on the (neglected) costs of internal hybrids, andin particular on motivational and commitment problemsthat derive from the delegation of decision rights. Theroot of such problems is that in firms, (delegated) deci-sion rights are not owned; they are always loaned fromthe holder(s) of ultimate decision-making rights, namelythe top management and/or the shareholders. Given this,a fundamental problem for top management/owners isto commit to real delegation and refrain from selectiveintervention (Williamson 1996) that harms motivation,and may reduce effort and investments in firm-specifichuman capital.

These ideas are developed and discussed empiricallywith reference to organizational changes that took placein the Danish electronics (primarily hearing aids) pro-ducer, Oticon A/S, beginning in 1991. Oticon becameworld famous for its radical delegation experiment. The"spaghetti organization," as it came to be called, wasexplicitly conceived of by its designers as an attempt toinfuse the Oticon organization with strong elements ofmarket control (Kolind 1990, Lyregaard 1993) and wasseen as a hard-to-replicate source of knowledge-basedcompetitive advantage (e.g., Gould 1994). In fact, arecent cottage industry has treated the Oticon experienceas an outstanding example of the sustained benefits thatradical project-based organizations may provide (e.g.,Lovas and Ghoshal 2000, Ravasi and Verona 2000,Verona and Ravasi 1999). However, his literature failsto note that the spaghetti organization in its initial radi-cal form does not exist anymore; it has been supersededby more structured administrative systems since around1996. In the following, these organizational changeswill be discussed from an organizational economicsstarting point. The approach followed with respect tounderstanding the nature of organizational changes inOticon is a historical one that relies heavily on the largenumber of thick descriptions of Oticon that have beenproduced by a number of mainly Danish academics,journalists, and Oticon insiders throughout the 1990s(in particular, Lyregaard 1993; Poulsen 1993; Mors-ing 1995; Morsing and Eiberg 1998; Eskerod 1997,

1998; Jensen 1998). However, these sources were sup-plemented with semistructured interviews with the primemover behind the spaghetti experiment, then-CEO,Lars Kolind, as well as the present Oticon HRM officer(both June 2000).

The paper begins by developing an organizational eco-nomics interpretation of the spaghetti organization {TheSpaghetti Organization: A Radical Internal Hybrid). Thespaghetti organization appears to have been a particu-larly well-crafted internal hybrid. Still, it gave way to amore traditional matrix structure. It is not plausible toascribe this organizational change to outside contingen-cies or to dramatic changes in strategic intent. This sug-gests that the spaghetti organization may have been besetby organizational costs that came to dominate the benefitaspects, necessitating a change of administrative systems{Spaghetti and Beyond). The Oticon spaghetti experi-ment carries lessons for the design of internal hybrids. Inparticular, it directs attention to the incentive problemsof delegating rights within a firm when top managementkeeps ultimate decision rights. Refutable propositionsfor the design of internal hybrids are derived {Discus-sion: Implications for Internal Hybrids).

It should be clear already at this stage that the follow-ing is an attempt to pursue a specific interpretation ofthe Oticon spaghetti episode. Organizational economicsper se is hardly in an early stage of theory developmentanymore, given that early work goes back more than sixdecades (Coase 1937), and the last three decades havewitnessed a fiurry of work in this field. There is there-fore little need for following a logic of grounded the-ory (Glaser and Strauss 1967). Moreover, organizationaleconomics is a particularly appropriate tool of interpre-tation in the present context, because only this bodyof theory simultaneously frames internal hybrids theo-retically, casts the analysis in the relevant comparative-institutional terms (e.g., allows comparison of externaland internal hybrids), and frames the kind of incentiveproblems that will be central in the following analysis.For example, neither information processing or motiva-tion theory can accomplish all this.^

In sum, the contributions of this paper are to (1) pres-ent a novel, and in key respects more encompassing,account and interpretation of a well-known organiza-tional change case, exemplifying the interpretive use-fulness of organizational economics in the process;(2) analyze the (neglected) costs of internal hybrids interms of the problem of selective intervention, thus con-tributing to understanding the efficient design of suchhybrids; and (3) argue that the analysis under (2) is alsohelpful for understanding broader issues of economic

332 ORGANIZATION SCIENCEA'OI. 14, No. 3, May-June 2003


organization, such as the governance choice betweeninternal and external hybrids.

The Spaghetti Organization:A Radical Internal HybridRecent work has used the Oticon Spaghetti experimentfor the purpose of developing notions of strategy mak-ing as "guided evolution" (Lovas and Ghoshal 2000),and to discuss how the deliberate introduction of "struc-tural ambiguity" through the choice of loosely coupledadministrative systems (Ravasi and Verona 2000) mayhelp to build "organizational capabilities for continuousinnovation" (Verona and Ravasi 1999). This literatureplaces all of the emphasis on the benefit side (mostlyinnovation performance) of the spaghetti experiment andfails to note that (and explain why) the spaghetti orga-nization has been largely abandoned. In contrast, thispaper accounts for the costs of this particular internalhybrid in terms of organizational economics, and usesthis account to explain the change from the spaghettiorganization.

Oticon: BackgroundFounded in 1904 and based mainly in Denmark, Oticon(now William Demant Holding A/S) is a world leader inthe hearing aids industry.^ In the first part of the 1990s,Oticon became a famous and admired instance of rad-ical organizational change. CEO Lars Kolind and hisnew organizational design became favorites of the press,consultants, and academics alike.'* The new organizationwas cleverly marketed as the very embodiment ofempowering project-and team-based organization. More-over, it quickly demonstrated its innovative potentialby re-vitalizing important, but "forgotten" develop-ment projects that, when implemented in the produc-tion of new hearing aids, produced significant financialresults, essentially saving the firm from a threateningbankruptcy, as well as by turning out a number of newstrong spin-off products. The background to the intro-duction of the spaghetti organization was the loss ofcompetitive advantage that Oticon increasingly sufferedduring the 1980s as a result of increasingly strong com-petition (mainly from the United States). Also, a changein the technological paradigm (Dosi 1982) in the hear-ing aids industry was gradually taking place throughthe 1980s from "behind-the-ear" hearing aids to "in-the-ear" hearing aids (Lotz 1998). Oticon's success inthe 1970s was founded on miniaturization capabilities.While these had been critical for competitive advan-tage in the "behind-the-ear" hearing aid paradigm, newtechnological capabilities in electronics which were not

under in-house control by Oticon were becoming cru-cially important in the emerging in-the-ear paradigm.

There is evidence (e.g., Poulsen 1993, Gould 1994,Morsing 1995) that at the end of the 1980s Oticon waslocked into a competence trap that was reinforced bystrong groupthink characterizing both the managementteam and the employees. A symptom of this was thatthe dominant opinion among managers and developmentpersonnel at Oticon was that the in-the-ear hearing aidwould turn out to be a commercial fiasco. Moreover,in-the-ear hearing aids were not perceived to be Oticonturf, in terms of both technological and marketing capa-bilities (Poulsen 1993). The self-image of the companyclearly was one of being a traditional industrial companywith its strongest technological capabilities in miniatur-ization and specializing in mass-producing behind-the-ear hearing aids, developing the underlying technologyincrementally. Administrative systems were organizedtraditionally with functional departments, the managersof which together constituted the senior executive group.When problems began to accumulate, various attemptswere made to change the situation, which, however, wereeither too insignificant or did not survive political jock-eying inside Oticon. In 1988, Lars Kolind assumed theposition of new CEO, concentrated all decision-makingpower in his own hands, and implemented drastic cost-cutting measures. However, he also quickly realized thatsomething else had to be done to cope with the decisivechanges that were underway with respect to products andprocesses in the industry. More radical measures wereneeded with respect to the strategic orientation of thefirm, the administrative systems that could back this up,and the technology that the firm sourced, leveraged, anddeveloped.

Trying SpaghettiThe new, radical measures were first sketched in a six-page memo (Kolind 1990), which described a funda-mental change of corporate vision and mission: Thecompany should be defined broadly as a first-classservice firm with products developed and fitted indi-vidually for customers, rather than narrowly, as a man-ufacturing company producing standard behind-the-earhearing aids. A new organizational form, namely the"spaghetti organization" (called so in order to empha-size the point that it should be able to change rapidly,yet still possess coherence), would support this strate-gic reorientation. The new form should be explicitly"knowledge-based," that is, consisting of " . . . knowledgecentres... connected by a multitude of links in a non-hierarchical structure" (Kolind 1994, pp. 28-29). Mak-ing the organization "anthropocentric," that is, designing

ORGANIZATION SciENCEA'bl. 14, No. 3, May-June 2003 333


jobs SO that these would " . . . fit the individual person'scapabilities and needs" (ibid., p. 31), was argued toprovide the motivational support for this knowledge net-work. Furthermore, basing the network on "free mar-ket forces" (Lyregaard 1993) would make it capable ofactually combining and recombining skills in a fiexiblemanner, where skills and other resources would move tothose (new) uses where they were most highly valued.Clearly, the aim was to construct a spontaneously work-ing internal network that would work with only minimalintervention on the part of Kolind and other managers,that is, "essentially, a free market at work" (LaBarre1996).

The new organizational form was primarily imple-mented in the Oticon headquarters (i.e., administration,research and development, and marketing). To sym-bolically underscore the fundamental transformation ofOticon, headquarters moved at 8 am on 8 August 1991to a completely new location north of Copenhagen. Inthe new building, all desks were placed in huge, openoffice spaces, and employees did not have permanentdesks, but would move depending on which projectsthey were working on. The number of formal titles wasdrastically reduced, resulting in a two-layered structurewith Kolind and 10 managers representing the manage-rial team and the remaining part of the organizationbeing organized into projects (Kolind 1994). Thus, thenew organization represented a breakdown of the oldfunctional department-based organization into an almostcompletely flat, project-based organization. Departmentsgave way to "Competence Centers" (e.g., in mechanicalengineering, audiology, etc.) that broke with the bound-aries imposed by the old departments. The "multijob"concept represented a notable break with the traditionaldivision of labor in organizations. It was based on twokey features: First, there were no restrictions on thenumber of projects that employees could voluntarilyjoin, and, second, employees were actively encouraged(and in the beginning actually required) to develop andinclude skills outside of their existing skill portfolio.^The underlying notion was that this would increase thelikelihood that project teams would consist of the rightmix of complementary skills and knowledge, because ofthe increase in the scope of the knowledge controlledby each team member. Moreover, the multijob conceptwould ease knowledge transfer because of the increasein the overlap of knowledge domains that it would pro-duce, as employees familiarized themselves with otheremployees' specialized fields.

These changes were accompanied by an extensivedelegation of the rights to make decisions on resourceallocation. Notably, employees would basically decide

themselves which projects they would join rather thanbeing assigned to tasks and projects from "above."Project managers were free to manage projects in theirpreferred ways. Wage negotiations were decentralized,with project managers receiving the right to negoti-ate salaries.^ Finally, although project teams were self-organizing and basically left to mind their own businessonce their projects were ratified, they were still requiredto meet with a "Projects and Products Committee" onceevery three months for ongoing project evaluation.

To meet the two potentially confiicting aims ofmaking it possible for project teams to rapidly andfiexibly combine the right skills, and achieving overallcoherence between rather independently taken decisions,the new organization was founded on four fundamen-tal ideas (Kolind 1994). First, as noted, the tradi-tional functional department structure was eliminatedin favor of a project organization that went consid-erably beyond the traditional matrix structure. Whilethis served to increase fiexibility, other measures weredirected towards achieving organizational coherence.Thus, secondly, new information technology systemswere designed and implemented to make it possible tocoordinate plans and actions in this decentralized orga-nization. Everybody was supposed to have full accessto the same information. Third, the traditional conceptof the office was abandoned, as already mentioned.Finally, Kolind worked hard to increase intrinsic motiva-tion by developing a corporate value base that stronglystressed responsibility, personal development, and free-dom. These fundamental organizing principles werebacked up by other measures. For example, to increasemotivation Kolind introduced an employee stock pro-gram, in which shop floor employees were invited toinvest up to 6.000 Dkr (roughly 800 USD) and man-agers could invest up to 50.000 Dkr (roughly 7.500USD). Although these investments may seem relativelysmall, in Kolind's view they were sufficiently large tosignificantly matter for the financial affairs of individualemployees; therefore, they would have beneficial incen-tive effects. More than half of the employees made theseinvestments.

The implementation of the spaghetti organization hadquick and strong performance effects (Peters 1992,Poulsen 1993). Improved performance in terms of theuse and production of knowledge was almost immediate,resulting in a string of remarkable innovations duringthe 1990s (Verona and Ravasi 1999, Ravasi and Verona2000). Improved growth and financial performance fol-lowed somewhat later (see Table 1).^

With respect to improvements in the use of knowl-edge, the spaghetti organization allowed significant



Table 1 Oticon Financial and Technoiogicai Performance

1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

Net rev. (mio. Dkr)Profit mg. (%)RoE (%)Product innovation

423.81.6

-8.5

449.68

11.6

455.43.79.4

476.51.8

-1.5Muitifocus

538.85.87.2

Personic

661.313,137

Oticon4 kids

750.317.937.9

Noaii

940.212.425.9

Microfocus

1,087.312.824.3

Digifocus

1,413.413.830.6

Spin-offinnovationsof digifocus

1,613.115.435.7

Spin-offinnovationsof digifocus

1,884.317.953.8

Ergo swiftdigifocus

II

Note. Sources: Ravasi and Verona (2000), Annual Reports of Oticon A/S, and Wiiliam Demant Hoiding A/S.

shelved projects to be revitalized. For example, it wasrealized that Oticon already had embarked upon devel-opment projects for in-the-ear hearing aids as far back as1979. These projects provided essential inputs into manyof the product innovations that Oticon launched duringthe 1990s. Another effect of the spaghetti organizationwas that product development time was reduced by 50%.In 1993, half of Oticon's sales stemmed from productsintroduced in 1993, 1992, and 1991. A total of 15 newproducts had been introduced since the implementationof the new organization, whereas none had been intro-duced in the last years of the earlier organization.

A recurring theme in academic treatments of theOticon spaghetti organization (Morsing 1995, Veronaand Ravasi 1999, Ravasi and Verona 2000) is that animportant cause of the observed increase in Oticon'sinnovativeness was the introduction of "structuralambiguity"—that is, the deliberate engineering of free-dom and ambiguity in the role system and in the author-ity structure by means of the introduction of a radicalproject organization. This condition facilitated the effi-cient and speedy integration and production of knowl-edge, resulting in the observed improvement of Oticoninnovativeness in the 1990s. This interpretation fails,however, to explain why the spaghetti organization wasgradually abandoned from about 1996 in favor of a moretraditional matrix organization. It also fails to accountfor the possible costs of the spaghetti organization. Thefollowing section presents a complementary interpreta-tion, based mainly on organizational economics.

The Spaghetti Organization as an Internal HybridA striking aspect of the spaghetti organization is theprevalence of the market metaphor in the commentarieson the new form by both insiders and outsiders (Peters1992, Lyregaard 1993, LaBarre 1996).^ The spaghettiorganization may indeed be interpreted as a radicalinternal hybrid, because the organization was stronglyinfused with elements characteristic of market exchange(see Table 2). Although there was no attempt to priceinternal services in the spaghetti organization and Oticon

employees did not become legally independent suppli-ers of labor services, in many other relevant dimensionsOticon was more like a market than a traditional hierar-chical firm. Thus, employees (particularly project lead-ers) were given many and quite far-reaching decision-making rights. Development projects could be initiatedby, in principle, any employee, just like entrepreneurs ina market setting, although these projects had to pass notthe market test, but the test of receiving approval fromthe Projects and Products Committee. Project groupswere self-organizing in much the same way that, forexample, partnerships are self-organizing. The setting ofsalaries was decentralized to project leaders, acting likeindependent entrepreneurs {Business Intelligence 1993).Incentives became more high-powered (i.e., efforts andrewards were more closely tied together) as performance

Table 2 iVIarket Organization and the Spaghetti Simulation

Market Organization The Spaghetti Organization

Allocation by means of pricingLegal independence between

parties (contract law)Freedom of contract

High-powered incentives

Dispersed residual claimancyDispersed decision rights

Dispersed ultimate decisionrights (dispersed formalauthority)

Resource allocation decen-tralized and strongly influ-enced by local entrepre-neurship

Strong autonomous adaptationproperties

Transfer prices not usedEmployment contracts

(employment law)Approximated by delegating

rights to suggest and joinprojects

Variable pay; initially based onobjective input and outputmeasures

Employee stock schemesVery widespread delegation of

rightsConcentrated ultimate deci-

sion rights (concentratedformal authority)

Local entrepreneurship verystrongly encouragedProjects approval easilyobtained

Secured through extensivedelegation of decisionrights



pay was increasingly used and as the employee stockownership program was introduced, thus mimicking thesuperior incentive properties of the market. Most hier-archical levels were eliminated and formal titles doneaway with, etc., mimicking the nonhierarchical natureof the market. In sum, market organization was indeedemulated in a number of dimensions.

As a general matter, the attraction of infusing hier-archical forms with elements of market control is thatsome of the basic advantages of the hierarchy, suchas the superior ability to perform coordinated adap-tation to disturbances (Williamson 1996), build spe-cialized social capital (Nahapiet and Ghoshal 1999),and share knowledge (Osterloh and Frey 2000), canbe combined with the superior incentive properties ofthe market (Williamson 1996) and its superior flexibil-ity with respect to autonomous adaptation (Hayek 1945,Williamson 1996). Along similar lines, Kolind explicitlysaw the spaghetti organization as combining the superiorabilities of a hierarchy to build knowledge-sharing envi-ronments and foster a cooperative spirit with the flexi-bility and creativity of a market-like project organization(Kolind 1994).

The Structure of Rights in theSpaghetti OrganizationOrganizational economics suggests that understandingthe costs and benefits of any organizational form requiresexamining the structure of decision and income rights inthe relevant form (Fama and Jensen 1983; Jensen andMeckling 1992; Hart 1995; Williamson 1996; Barzel1997; Baker et al. 1999, 2002; Holmstrom 1999). Thebenefits and the costs of the spaghetti organization canbe comprehended through this lens. The remaining partof this section concentrates on the benefit side.

Centralized decision-making systems, particularlylarge ones, have well-known difficulties with respect tomobilizing and efficiently utilizing important "sticky"knowledge (von Hippel 1994) such as the precise char-acteristics of specific processes, employees, machines,or customer preferences (Jensen and Wruck 1994). Theytherefore often also have difficulties combining suchknowledge into new products and processes (Laursenand Foss 2002). As Hayek (1945) explained, the mainproblem is that much of this knowledge is transitory,fleeting, and/or tacit, and therefore costly to articulateand transfer to a (corporate) center. Markets have advan-tages relative to pure hierarchies with respect to utilizingsuch knowledge, particularly when it is not required toutilize the relevant knowledge in conjunction with otherknowledge sets (where a hierarchy may have compara-tive advantages).'" Thus, markets economize on the costs

of transferring knowledge by allocating decision rightsto those who possess the relevant knowledge, rather thanthe other way around (Hayek 1945, Jensen and Meckling1992). Rights will move towards the agents who placethe highest valuation on those rights. Since these agentsbecome residual claimants, effective use will be made ofthe rights they acquire. From this perspective, internalhybrids are fundamentally attempts to mimic, inside thehierarchy, the decentralization of decision and incomerights that characterizes the market in an attempt toimprove the efficiency of processes of discovering, cre-ating, and using knowledge.'^

The spaghetti organization may be understood throughthis lens, that is, as a hybrid organizational design thataimed at improving the colocation of knowledge andrights through extensive delegation, and backed up thisdelegation of decision rights by giving employees moreincome rights. By giving project teams extensive deci-sion rights, requiring that ideas for projects be madepublic, and ensuring that project teams possessed thenecessary complementary skills for a particular market-ing, research, or development task, the spaghetti orga-nization stimulated a colocation of decision rights withknowledge. High-powered incentives were provided inan attempt to make sure that efficient use was made ofthose rights. This improved the use of existing knowl-edge (cf. the revitalization of "forgotten" projects) andeased the combination of knowledge in the productionof new knowledge.

However, Oticon's use of "free market forces"(Lyregaard 1993) was fundamentally a simulation,because the allocation of decision rights in thatorganization (as in any firm) remained in importantrespects different from the allocation that characterizesmarket organization. In contrast to markets, firms cannotconcentrate income rights (i.e., residual claimancy) anddecision rights to the same extent, in the same hands. Anagency problem results from this separation. Many ofthe elements of the spaghetti organization may be seenas responses to this fundamental agency problem, mostobviously the increased use of high-powered incentives.Consider also the rights to allocate resources to a par-ticular project. These may be broken down into groupsof decision-making rights, namely rights to (1) initiateprojects, (2) ratify projects, (3) implement projects, and(4) monitor and evaluate projects (cf. Fama and Jensen1983). The efficiency of decision-making processes inproject-based firms rests on the allocation and exerciseof such rights. For reasons of efficiency, firms usuallydo not concentrate these rights in the same hands; rather,initiation and implementation rights may be controlled



by one person (or team), while ratification and mon-itoring rights are controlled by other persons, usuallyhierarchical superiors."

This allocation of decision rights was characteristicof the spaghetti organization. Whereas anybody couldinitiate a project, projects had to be evaluated by theProducts and Projects Committee that was staffed byKolind, the development manager, the marketing man-ager, and the support manager. The Committee eitherrejected or approved of the project. The only formal cri-teria for getting a project accepted were that the rele-vant project relate to the business areas of Oticon andyield a positive return over a three-year period and witha discount rate of 30%. Apparently, the Products andProjects Committee did not control the use of corpo-rate resources by means of controlling the budgets ofindividual projects at the project ratification stage. Inparticular, the use of human resources—the main inputcategory—across projects was not monitored. The rightsto implement a project following approval included theright to hire employees in open competition with otherprojects (Eskerod 1998). Operating projects would meetevery third month with the Products and Projects Com-mittee, or a representative thereof, for project evaluation(i.e., monitoring).

The fact that the Projects and Products Committeecould veto a project ex ante suggests that it was thereal holder of power in Oticon. Frequent interventionon the part of the Committee ex post project approvalconfirms this (Eskerod 1998). Thus, it became increas-ingly clear that the Committee could at any time halt,change, or even close projects. This kind of interventiontook place frequently. The Projects and Products Com-mittee's exercise of their ultimate decision rights may beseen as simply reflecting the separation discussed abovebetween decision management (i.e., initiation, imple-mentation, and daily project management) and decisioncontrol (i.e., project evaluation and monitoring).'^ How-ever, this separation does not logically imply the kind ofex post intervention that the Committee engaged in. Forexample, one may imagine that the relevant rights mightbe allocated so precisely and with so much foresight thatthere are no incentives to intervene ex post, as in the caseof a very detailed contract between two legally inde-pendent firms. However, the way in which the Projectsand Products Committee exercised their ultimate deci-sion rights is more akin to reneging on a contract, per-haps even to performing a "hold-up" (Williamson 1996).Thus, the Committee effectively reneged in implicit con-tracts with the projects as the efforts of project became,in the eyes of the Committee, superfluous (e.g., because

of new technological developments), moved in unfore-seen directions, or were revealed to have been foundedon ill-conceived ideas. In turn, this exercise of ultimatecontrol rights caused unforeseen incentive problems, aswill be discussed later.

Organizational ComplementaritiesAn interesting aspect of the spaghetti organization is thatan explicit logic of complementarity was present in thereasoning of its main designer. Observed Kolind: "It wasnot strictly necessary to do all these things at the sametime, but we opined that with a simultaneous implemen-tation of the changes [in organizational elements]... theywould reinforce each other" (in Mandag Morgen 1993,p. 17; my translation). Complementarities between ele-ments of an organizational form exist when increas-ing the level of one element increases the marginalreturn from increasing the level of all remaining ele-ments (Milgrom and Roberts 1990, Hemmer 1995,Zenger 2002). Loosely, when complementarity obtains,the dynamics of organizational elements imply that theymove together. Changing one element in an isolated wayis likely to set in motion (possibly unforeseen) processesof change in other elements because the system willgrope towards an equilibrium where all elements havechanged (Zenger 2002). The process of groping may beassociated with serious inefficiencies. Therefore, organi-zational change initiatives should "get the complemen-tarities right."

Apparently, the spaghetti organization did exactly this.Thus, the change in the rights structure of Oticon wassuch that decision rights changed in a way that was com-plementary to the change in income rights; specifically,widespread delegation of decision rights was accompa-nied by making incentives more high powered throughperformance pay and employee ownership. In turn, thechange in incentives was backed up by complementarychanges in measurement systems. Thus, a performanceevaluation system was implemented in which employeeperformance was measured in three to eight differentdimensions (depending on the type of employee) andpay was made dependent on these measures (Poulsen1993).

Other initiatives may also seen to be complementaryto the increase in the delegation of rights in the spaghettiorganization. For example, the open office landscapeand the strategically placed coffee bars and staircaseswere complementary to rights delegation in terms ofutilizing and building knowledge because they helpedfoster the knowledge exchange that gave rise to newideas for project teams. With respect to the moral haz-ard problem introduced by delegating rights, the new



much-more information-rich environment was also com-plementary to this delegation, because it helped to buildreputational effects (cf., Eskerod 1997, 1998) and easedmutual monitoring among employees, keeping agencyproblems at bay. Kolind's (1990) strong emphasis onbuilding culture in the new organization may be seenin a similar light: Influencing preferences through thebuilding of shared values became an important activityin the spaghetti organization, because its strong dele-gation of rights introduced both problems of coordi-nating independently made decisions (Miller 1992) andagency problems, problems that are reduced as prefer-ences become more homogeneous. The complementarynature of these organizational elements also explains thespeed and toughness with which Kolind managed thetransition from the old organization.'^ This is becauseit is usually inefficient to change systems of comple-mentary elements in an incremental manner; transitionbetween such systems should normally be accomplishedin a "big bang" manner (cf., Dewatripont and Roland1995).

Spaghetti and BeyondRetreating from SpaghettiIn his account of the spaghetti organization, Gould(1994, p. 470) noted that " . . .Lars Kolind's vision wasthe right one for Oticon. In any case, one thing was cer-tain: there could be no turning back." However, begin-ning in 1996, a considerable "turning back" actuallybegan: Oticon embarked upon a partial abandonment ofthe spaghetti organization and gradually adopted a moretraditional matrix structure. In 1996, Oticon headquar-ters was divided into three "business teams" that areessentially new administrative layers. In addition to thebusiness teams, a "Competence Center" was set up. Thisunit is in charge of all projects and their financing andof an operational group that controls administration, IT,logistics, sales, and exports. It is one of the successorsto the now abandoned Projects and Products Committee.However, its style of managing projects is very different.In particular, care is taken to avoid the kind of interven-tion in already-approved-of projects that characterizedthe Products and Projects Committee. The team lead-ers and the head of the Competence Center comprise,together with the CEO, the "development group," whichmay be seen as a second successor to the Products andProjects Committee of the original spaghetti organiza-tion. The development group, which essentially is thesenior executive group, is in charge of overall strategymaking. It is also the unit from which most of the initia-tive with respect to starting new projects comes. Many of

the decision-making rights earlier held by project leadershave now been concentrated in the hands of the Com-petence Center, or the managers of the business teams.For example, project leaders' rights to negotiate salarieshave been constrained. Project leaders are appointed bythe Competence Center, so that the right to be a projectleader is not something that one grabs, as under thespaghetti organization. Although the multijob concept isstill present, the extreme forms that characterized thespaghetti organization are not.

To sum up, recent changes of administrative systemsat Oticon have amounted to a break with the radi-cal bottom-up approach that characterized the originalspaghetti structure. Thus, although Oticon is still charac-terized by considerable decentralization and delegationof rights, many of the crucial elements of the spaghettiorganization have been left.

Searching for Possible Causes of the PartialFailure of the Spaghetti ExperimentAlthough the spaghetti organization at first inspectionseems to have been a particularly well-crafted internalhybrid, closer inspection may reveal design mistakes thatcaused its abandonment. An organizational economicsperspective suggests that a number of candidates fordesign mistakes be discussed. They may be grouped intoproblems of allocating competence, eliminating tourna-ments, sacrificing specialization advantages, coordina-tion, knowledge sharing, and infiuence activities.'"* Theyare discussed in the following.

Allocating Competence. Demsetz (1988) and Casson(1994) argue that firms are hierarchical because this isan efficient way of utilizing different yet complementaryknowledge; direction may be less costly than instructionor joint decision making. When this is the case, thosewith more decisive knowledge should direct those withless decisive knowledge. Thus, the hierarchy is an effi-cient method of allocating competence. The spaghettiorganization eliminated most hierarchical levels. Thus,the extent to which hierarchy could be used as a sort-ing mechanism for allocating skills was much smallerin the spaghetti organization. For example, the delega-tion of project initiation rights implied that competentand less competent employees had the same rights toinitiate projects and get a hearing before the Projectsand Products Committee. Knowledge-based inefficien-cies may have resulted that may have been avoided in atraditional hierarchy.

However, this explanation implicitly asserts that man-agers are, on average, more knowledgeable with respectto what actions subordinate employees should optimally

338 ORGANIZATION SCIENCE/VOI. 14, No. 3, May-June 2003


take than these employees are themselves. If this is notthe case, bottom-up selection processes may sort betterthan hierarchical processes. In fact, the spaghetti orga-nization was (at least in the official rhetoric) very muchfounded on the notion that bottom-up processes wouldselect more efficiently than hierarchical processes. Hier-archical superiors may be more knowledgeable aboutwhich actions should be optimally taken by subordi-nates when there are strong complementarities betweenthe actions of subordinates, and hierarchical superiorspossess superior information about these complemen-tarities and/or they possess private information aboutwhich states of the world have been realized (Fossand Foss 2002). To be sure, complementarities betweensubordinates' actions and knowledge sets obtained inthe spaghetti organization. However, the purpose ofthe spontaneous, marketlike, bottom-up processes wasexactly to discover and utilize such complementarities—something that the earlier hierarchical organization hadnot been capable of. Thus, it seems unlikely that abolish-ing the hierarchy in Oticon led to serious inefficienciesrelated to the allocation of competence.

Eliminating Tournaments. From an incentive perspec-tive, the extremely flat spaghetti organization impliedthat one particular incentive instrument was no longeravailable to the organization: Hierarchical job lad-ders could no longer function as incentive mechanismsin their own rights because the spaghetti organiza-tion essentially abolished what agency theorists call"tournaments" between managers (Lazear 1995). Pro-motion was no longer a "prize" that could be obtainedthrough expending effort. However, while the spaghettiorganization may have eliminated this particular incen-tive instrument, it introduced a number of new incentiveinstruments, such as performance pay. From the pointof view of individual employees, these new instrumentsmay have had stronger motivational effects than tourna-ments because they were less open to political manipu-lation. Thus, the sacrifice of tournaments as an incentiveinstrument may not have been a major problem.

Sacrificing Specialization Advantages. A key com-ponent of the spaghetti organization was the multi-job concept which implied that each employee was(1) encouraged to develop skills outside her present skillportfolio and (2) free to join projects as she saw fit.Much work on Oticon has treated the multijob conceptas a strong stimulus to knowledge exchange and integra-tion (e.g., Verona and Ravasi 1999, Ravasi and Verona2000), presumably quite rightly so. However, the con-cept may also have introduced distinct costs, most obvi-ously the sacrifice of specialization advantages that it

would seem to imply. However, there are indications thatthis was actually not the case. For example, an Oticonengineer may have been encouraged to develop Englishwriting skills, which would place him in a better positionto undertake technical translation relevant to his project,and do so in a more informed way than a professionaltranslator would be capable of. Thus, this aspect of themultijob concept may have led to beneficial exploitationof complementarities between different skills.

Problems of Coordination. However, there is strongevidence that the second part of the multijob concept,the freedom to join projects, had significant costs.'^Nobody kept track of the total time that employees spenton projects.'* Moreover, project leaders were free totry to attract those who worked on competing projects,and in many cases they succeeded in doing so. Thiswas a consequence of the explicit aim to emulate themarket, but the effect was that it was hard to commitemployees to projects and to ensure an efficient allo-cation of attention to projects (Gifford 1992). This ledto severe coordination problems because project lead-ers had no guarantee that they could actually carry aproject to its end. Moreover, many employees joinedmore projects than their time resources possibly allowedfor, creating problems of coordinating schedules andwork hours. The Products and Projects Committee hadno routines for dealing with these problems. Appar-ently, reputation mechanisms were not sufficient for cop-ing with them either. It would perhaps seem that theseproblems could have been reduced by simply prohibit-ing employees from working on more than, say, twoprojects that could not add up to more than 100% of theemployee's total work hours.'' Establishing such con-trols in the original spaghetti organization would, how-ever, have run against the official rhetoric of autonomy,empowerment, and delegation. Alternatively, monitoringsystems might have been refined to control dimensionsof employee behavior that related to their attention andwork allocation across the projects in which they par-ticipated, so as to reduce coordination problems. How-ever, the very elaborate monitoring system that wasimplemented together with the spaghetti organizationand that involved the construction of objective measureson half a dozen aspects of employee behavior (Poulsen1993) appears to have been quickly and tacitly shelvedand substituted with a simpler system that relied muchmore on subjective performance assessment (BusinessIntelligence 1993). This suggests that the problem withmonitoring systems under the original spaghetti organi-zation rather was that they were already too complexand costly to administer in practice.

ORGANIZATION SCIENCEA/OI. 14, No. 3, May-June 2003 339


Problems of Knowledge Sharing. The multijob con-cept promoted knowledge sharing and, in turn, knowl-edge creation. However, there is evidence (Eskerod1997, 1998) that knowledge sharing was not alwaysspontaneous and uninhibited. In fact, in some cases,knowledge tended to be held back within projects,because of the widespread and correct perception thatprojects were essentially in competition over corporateresources. Thus, by stressing so strongly a marketlikecompetitive ethos and by making incentive systems more"high-powered" (Williamson 1996) than they had beenunder the old organization, the spaghetti organizationto some extent worked against its stated purposes. Theorganization's measurement and reward systems appar-ently could not fully cope with these problems. It maybe questioned how significant this problem was. Theimpressive innovation record of Oticon in the 1990sindicates that the firm's creation of knowledge may nothave been significantly harmed by the competitive rela-tions existing within the spaghetti organization. Still,the relevant question is whether the knowledge-sharingenvironment could have been better designed. Knowl-edge sharing is not necessarily best stimulated by a kindof project organization that simulates competitive mar-kets. To the extent that knowledge sharing is a hard-to-measure performance variable, employees are likely toput less of an emphasis on this (Holmstrom and Milgrom1991). Upon realizing this, resorting to lower-poweredincentives is likely (Holmstrom 1999). This correspondsto what has happened in Otcion. Although the perfor-mance measurement systems in Oticon now includesattempts to measure employees' contribution to knowl-edge sharing, it is also tfie case that the strong competi-tive ethos which characterized the spaghetti organizationhas been significantly dampened in the successor form.

Influence Activities. Influence activities are activitiesthat subordinates engage in when they infiuence hier-archical superiors to make decisions that are in theirown interests, rather than in the organization's (Milgrom1988, Argyres and Mui 2000). Resources expended oninfluence activities are, from the point of view of theorganization, waste. It is arguable that it is relativelymore difficult under an organization such as the spaghettiorganization to protect against infiuence activities. Thisis because everybody has, in principle, direct access tothe management team. A comparative advantage of thetraditional, hierarchical, and rule-governed organizationis exactly that it may be better at protecting itself againstinfiuence activities because access to those who holdultimate decision rights is more difficult. In fact, thespaghetti organization which actively stimulated compe-tition between project groups for the approval of the only

relevant "hierarchical superior" left, namely the Projectsand Products Committee, did produce such influenceactivities (Eskerod 1998). In contrast, under the hierar-chical form existing prior to the spaghetti organization,such activities had been much less prevalent becauseof the aloof management style of the old management(Poulsen 1993). Personal relations to those who staffedthe Committee became paramount for having a projectratified by the Committee. As Eskerod (1998, p. 80)observed:

Part of being a project group may be lobbying in the PPC

trying to obtain a high priority status by influencing the PPC

members. The reason for doing this is that a high priority

project is regarded as a very attractive place for the employees,

because the management sees this project as important.

It is, however, not clear from the existing empiricalstudies of the spaghetti organization that this was per-ceived of as a serious problem in the organization—forexample, whether it resulted in obviously unimportantprojects being approved of by the Committee. Rather, itwas taken as an unavoidable, and relatively small, costof the spaghetti organization.'^

To sum up, the search for the causes of the par-tial abandonment of the spaghetti organization so farseems to lead only to inefficiencies stemming from thelack of well-functioning project management routineson the part of Products and Projects Committee beinga serious problem. However, handling this problem didnot necessarily require a major organizational change.Still, the many possible small liabilities of the spaghettiorganization (problems of knowledge being held backin projects, influence activities, etc.) may together haveadded up to significant costs that could be reduced byadopting a more structured organizational form (B0rsensNyhedsmagasin 1999, interview with Henrik Hoick onJune 2000). Moreover, there is one fundamental problemleft that was clearly present in the spaghetti organiza-tion, and which is a strong candidate for explaining theabandonment of that organizational form.

The Problem of Selective InterventionAlthough infusing hierarchical forms with elements ofmarket control seems attractive, crafting and implement-ing such internal hybrids is a highly complicated prob-lem. One reason is a fundamental incentive problemthat plagues all hierarchies, but is arguably particu-larly prevalent in the kind of very fiat organizations ofwhich the Oticon spaghetti organization is an exam-ple. An early statement of the nature of this problemcan be found in the comparative systems literature in



economics, that is, the literature taken up with the eco-nomic differences between capitalist and socialist sys-tems. Thus, Mises (1949, p. 709) argued that thereare fundamental problems involved in "playing market"inside hierarchies.'^ Specifically, schemes for a socialistmarket economy would not work because the concentra-tion of ultimate decision-making rights and responsibil-ities (i.e., ownership) in the hands of a central planningboard would dilute the incentives of managers. Thus,while planning authorities could delegate rights to makeproduction and investment decisions to managers, theserights would be inefficiently used. First, because man-agers could always be overruled by the planning author-ities, they were not likely to take a long view, notablyin their investment decisions. Second, because managerswere not the ultimate owners, they were not the fullresidual claimants of their decisions and, hence, wouldnot make efficient decisions.

Later research has clarified that (1) handling the prob-lem requires that the planning authorities can crediblycommit to a noninterference policy, and (2) the prob-lem goes beyond the comparative systems context. Itis latent (or manifest) in all relations between "rulers"and "ruled" (North 1990, Miller 1992, Williamson 1996,Foss and Foss 2002). The problem arises from the factthat it is hard for the ruler to commit to a noninterfer-ence policy because reneging on a promise to delegatewill in many cases be extremely tempting, and thoseto whom rights are delegated will anticipate this. Lossof motivation results. The problem is not unknown inorganizational studies, (e.g., Vancil and Buddrus 1979,p. 65). In particular, Williamson's (1996) concept of the"impossibility of selective intervention" is highly rele-vant. He describes it as

. . . a variant on the theme, "Why aren't more degrees of free-dom always better than less?" In the context of firm and mar-ket organization, the puzzle is, "Why can't a large firm doeverything that a collection of small firms can and more." Bymerely replicating the market the firm can do no worse thanthe market. And if the firm can intervene selectively (namely,intervene always but only when expected net gains can be pro-jected), then the firm will sometimes do better. Taken together,the firm will do at least as well as, and will sometimes dobetter than, the market (1996, p. 150).

Williamson directly argues that (efficient) selective inter-vention of this kind is "impossible." Incentives arediluted, because the option to intervene " . . . can be exer-cised both for good cause (to support expected netgains) and for bad (to support the subgoals of the inter-venor)" (Williamson 1996, pp. 150-151). Promises toonly intervene for good cause can never be credible,Williamson argues, because they are not enforceable

in a court of law. The wider implication of this rea-soning is that because decision rights cannot be dele-gated in a court-enforceable manner inside firms (i.e.,are not contractible), authority can only reside at the top.Authority cannot be delegated, even informally, becauseany attempt to do this will run into the problem of theimpossibility of selective intervention. One would there-fore expect to see little use of delegation. Given thatdelegation is clearly a viable and widespread organi-zational practice, this suggests that this implication isgoing too far.

In fact, it is conceivable that the "intervenor" maycredibly commit to not intervene in such a way that the"subgoals of the intervenor" are promoted. The logicmay be stated in the following way (cf. Baker et al.1999). Assume that a subordinate initiates a project.^"Assume further that the manager has information thatis necessary to perform an assessment of the project,but that he decides upfront to ratify any project that thesubordinate proposes. Effectively, this amounts to fullinformal delegation of the rights to initiate and ratifyprojects—"informal," because the formal right to ratifyis still in the hands of the manager and because that rightcannot be allocated to the subordinate through a court-enforceable contract (cf. Williamson 1996). Because thesubordinate values being given freedom—she is partly aresidual claimant on the outcomes of his activities—thiswill induce more effort in searching for new projects(Aghion and Tirole 1997, Foss and Foss 2002). To theorganization, the expected benefits of these increasedefforts may be larger than the expected costs from thebad projects that the manager has to ratify. However, aproblem arises when the manager has information aboutthe state of a project ("bad" or "good"), because he maythen be tempted to renege on a promise to delegate deci-sion authority, that is, intervene in a "selective" manner.If he overrules the subordinate, the latter will lose trustin him, holding back on effort. Clearly, in such a gamea number of equilibria, each one characterized by dif-ferent combinations of employee trust and managerialintervention, are feasible. What determines the particu-lar equilibrium that will emerge is the discount rate ofthe manager, the specific trigger strategy followed bythe subordinate (e.g., will he lose trust in the managerfor all future periods if he is overruled, or will he bemore forebearing?), and how much the manager valueshis reputation for not reneging relative to the benefits ofreneging on a bad project (Baker et al. 1999).

All of the above build on standard economics assump-tions on motivation and cognition: Employees are moti-vated solely by being able to share in the outcomesof their activities, and managerial intervention decreases

ORGANIZATION SCIENCEA'OI. 14, No. 3, May-June 2003 341


motivation because it means that the expected gainof putting effort into a project diminishes. Includingricher motivational and cognitive concerns aggravatesthe problem of selective intervention. As argued in anextensive literature in psychology (summarized in Frey1997), people are also likely to be intrinsically moti-vated. Such motivation may be sustained by psycholog-ical contracts that involve loyalties and emotional ties(Brockner et al. 1992, Robinson and Morrison 1995,Osterloh and Frey 2000, p. 541). Managerial interven-tion, particularly when it is perceived to be essentiallyarbitrary, may break such contracts and harms intrinsicmotivation (Robinson and Rousseau 1994). Other partsof psychological research (summarized in Bazerman1994) suggest other ways in which the problem of selec-tive intervention may be aggravated in practice. Thus,robust findings in experimental psychology show thepresence of a systematic overconfidence bias in judg-ment; that is, people tend to trust their own judgmentsmore than is "objectively" warranted. Managers are notexceptions to this bias, perhaps quite the contrary. Thepresence of the overconfidence bias in the judgments thatunderlie managerial decision making is likely to aggra-vate the problem of selective intervention, because itproduces additional meddling in subordinates' decisions(Nickerson and Zenger 2001, p. 15).

Selective Intervention in OticonIt is arguable that the main reason that the spaghettiorganization was changed into a more hierarchicalorganization has to do with the kind of incentive andmotivational problems described above. The officialOticon rhetoric, stressing bottom-up processes in a fiex-ible, marketlike, and essentially self-organizing systemwith substantial autonomy and a management team (i.e.,the Projects and Products Committee) that acted as lit-tle more than facilitator (Kolind 1990, Lyregaard 1993),became increasingly at odds with the frequent selectiveintervention that was undertaken by the Projects andProducts Committee.^' The need for selective interven-tion was rationalized by an external observer in the fol-lowing terms:

. . . PPC [the Products and Projects Committee] does not make

general written plans, which are accessible to the rest of

the organization... if this were done, plans would have to

be adjusted or remade in an ever-continuing process, because

the old plans had become outdated (Eskerod 1998, p. 80).

This entirely ad hoc approach was taken by the Prod-ucts and Projects Committee to be an unavoidable fea-ture of a fiexible, project-oriented organization (Eskerod

1998, p. 89). However, it was also a direct signal toemployees that the "contract" between any project andthe Products and Projects Committee was very incom-plete (Williamson 1996), and that the Committee mightat any time exercise its ultimate control rights for thepurpose of intervening in projects. This produced dilutedincentives and badly harmed motivation (as documentedat length by Eskerod 1997, 1998). Accumulating frustra-tion finally resulted in a major meeting in 1995, whichmarked the beginning of the retreat from the spaghettiorganization. At the meeting employees dramaticallyexpressed their concerns about the contrast between onthe one hand, the Oticon value base, including the strongrhetoric of delegation, and on the other hand, the wayin which the company was actually managed. Frustra-tion that projects were interrupted in seemingly arbitraryways and that the organization was far better at gen-erating projects than at completing them was explicitlyvoiced.

The preceding discussion suggests that a fundamentalproblem in the spaghetti organization was that Kolindand the Products and Projects Committee never com-mitted to a policy of not intervening selectively; nei-ther, apparently, did they intend to do so or even seeany rationale in it. Kolind's view appears to have beenthat in important respects and in many situations, heand the Products and Projects Committee would pos-sess accurate knowledge about the true commercial andtechnical possibilities of a given project, and that effi-cient utilization of corporate resources dictated interven-ing in, and sometimes closing down, projects. However,that view clashed on a basic level with the rhetoric ofwidespread delegation of decision rights, leading to thedemise of the spaghetti organization and the adoption ofthe present more-structured matrix organization.

In principle, Kolind and the Products and ProjectsCommittee could have committed to a policy of nonin-terference from the beginning, rather than acting on thebelief that organizational fiexibility required that theyselectively intervene in projects. Conceivably, this mighthave made this radical internal hybrid viable. However,even if Kolind and the Products and Projects Commit-tee had announced initially that they would refrain fromselective intervention, there are reasons why this com-mitment may not have been sustainable in the longerrun. Thus, it was increasingly clear that the elabo-rate system of measures that was initially installed wasinadequate. It did not capture important dimensions ofbehavior (e.g., employees' contribution to knowledgesharing) and it may have contributed to some projectsholding back knowledge. Rather than trying to refinethe system further, it was abandoned.^^ However, the

342 ORGANIZATION SCIENCEAÔI. 14, No. 3, May-June 2003


implication was that management could no longer takeplace solely through incentives (following initial ratifi-cation of projects). The employee stock ownership pro-gram was arguably not sufficiently high powered to trulymotivate, and did not confer sufficient decision rightsto halt the practice of selective intervention to employ-ees. The implication was that Kolind and the Productsand Projects Committee had to engage in much moremonitoring of the projects. Doing this without com-promising team autonomy and harming motivation wasunlikely.

The New OrganizationA notable feature of the present Oticon organization liesin its much more consistent approach towards projects.Organizational expectations are that priorities do notchange in the rapid and erratic manner that characterizedthe original spaghetti organization, and that employeescan be much more sure that the projects they are workingon are taken all the way to the end. In the new orga-nization, projects are rarely stopped or abandoned, andthere is an explicitly stated policy of sticking to ratifiedprojects. Two reasons are given for this. First, projectsare more carefully examined with respect to technicalfeasibility and commercial implications. An aspect ofthis is that the Competence Center now much moreactively puts forward project ideas and contacts poten-tial project leaders, rather than relying on the bottom-up approach that characterized the original spaghettiorganization. Thus, hierarchical selection has to someextent substituted for selection performed by bottom-up processes. Second, the wish to avoid harming moti-vation (i.e., diluting incentives) by overruling ongoingprojects is strongly stressed. The management team hasopenly announced this policy, and has made it credi-ble by (1) consistently sticking to it and (2) researchingproject ideas carefully ex ante so that employees' per-ceived probability that intervention will occur is low.Some reasons why a more traditional hierarchy may bebetter at making such commitment credible is discussedin the following section.

Discussion: Implications forInternal HybridsProponents of internal hybrids argue that their advantagehes in the ability to integrate the virtues of more conven-tional organizational forms (Miles et al. 1997). Specif-ically, internal hybrids combine the ability to achieveefficiencies through specialization that characterizes thefunctional form with the relative independence that canbe granted in a divisional form, and the ability to transfer

resources and capabilities across division and businessunit boundaries that characterize the matrix organiza-tion (e.g.. Miles and Snow 1992). The designers of theOticon spaghetti organization invoked strikingly similararguments (Kolind 1990, 1994; Lyregaard 1993). Thissuggests that broader lessons with respect to the efficientdesign of internal hybrids may emerge from the Oticonexperience.

Getting Complementarities RightA basic proposition in much of organization theory isthat for reasons of efficiency, organizational forms arealigned with environmental conditions, strategies, andexchange conditions in a systematic and discriminatingmanner (Thompson 1967, Meyer et al. 1993, Williamson1996, Nickerson and Zenger 2000). Thus, Zenger (2002,p. 80) argues that many attempts to infuse hierarchieswith elements of market control break with this basicproposition and often " . . . violate pattems of comple-mentarity that support traditional hierarchy as an organi-zational form." For example, managers implement newstructures without new performance measures and newpay systems, or they implement new pay systems with-out developing new performance measures. This resultsin unstable, possibly inefficient, hybrid forms. In con-trast, viable internal hybrids are characterized by orga-nizational elements clustering in certain characteristic,complementary combinations, just as in the case of mar-kets and hierarchies (Williamson 1996).

Did the spaghetti organization get the complemen-tarities between organizational elements right? On firstinspection, it did, as has been argued. However, closerinspection reveals a somewhat different picture. Thus,it may be argued that Oticon did not get the organiza-tional complementarities exactly right, because the kindof radical internal hybrid that was adopted requires thatprojects be managed almost exclusively through the pro-vision of incentives and ownership (Miles et al. 1997,Zenger 2002). The performance measurement systemsin the spaghetti organization were not adequate to sup-port precise performance evaluation. Some relevant per-formance dimensions (e.g., contribution to knowledgesharing) were not measured at all. Also, the incentiveeffects of the employee stock ownership program appearto have been limited. Thus, remuneration schemes maynot have rested on sufficiently precise and encompass-ing measures and were not sufficiently high powered tocomplement the widespread delegation of decision rightsin the organization. This fostered a need for selectiveintervention on the part of Kolind and the Products andProjects Committee that went beyond what would have

ORGANIZATION SCIENCEA'OI. 14, No. 3, May-June 2003 343


been necessary with better measures of employee perfor-mance, and which had the unintended effect that motiva-tion was seriously harmed. This reasoning suggests thefollowing proposition:

PROPOSITION 1. Internal hybrids that violate patternsof complementarity characteristic of this organizationalform will be subject to more problems of selective inter-vention than hybrid forms that get the complementaritiesright.

A corollary to this proposition is that advances in mea-surement methods will result in less selective inter-vention because the measurement of performance isimproved so that the moral hazard stemming from thedelegation of rights is reduced.

Problems of Intervention and Organizational FormThe motivational and incentive problems that mayemerge from managerial selective intervention are notindependent of organizational structure, notably, thenumber of hierarchical layers in the organization, andtherefore the distribution of information and authorityin a firm. Arguably, organizations that adopt internalhybrids that amount to drastically reducing the num-ber of hierarchical layers, such as Oticon's spaghettiexperiment, are more prone to the problem than moretraditional hierarchical firms. There are (at least) threereasons for this.

First, decision rights are more solidly established ina traditional hierarchy, which is associated with well-defined, distinct positions than in a flat, project-basedorganization where decision rights are more fleeting.Organizational expectations that certain positions comewith certain decision rights are very well established,and potentially costly for a top manager to break withthrough selective intervention. The same kind of orga-nizational expectations are not likely to be establishedin a flat, project-based organization. Second, a top man-ager who selectively intervenes in a hierarchical organi-zation risks overruling the whole managerial hierarchy(all those below him), whereas this may be a smallerconcern in a flat organization where the CEO may onlyharm motivation in a specific project team if he over-rules that team. Third, information-processing perspec-tives (Thompson 1967, Galbraith 1974) suggest that thehierarchy is not just a structure of authority, but alsoone of information. The informational distance betweenprojects and top manager may be increased by havinga multilayered hierarchy. This implies that the top man-ager knows that he is in key dimensions ignorant aboutthe project (Aghion and Tirole 1997). In this case, hisincentives to selectively intervene will be small. The pre-ceding arguments suggest the following proposition:

PROPOSITION 2. An internal hybrid form that is orga-nized within a firm with few hierarchical layers will beassociated with larger efficiency losses caused by prob-lems of selective intervention than an internal hybridform that is organized within a firm with more hierar-chical layers.

Problems of Intervention and theExternal EnvironmentA key reason why the Products and Projects Com-mittee considered that frequent selective interventionwas necessary had to do with the impossibility ofmaking detailed plans for future business developmentin an industry where unforeseen contingencies (e.g.,new technologies) often occurred. This suggests a thirdproposition:

PROPOSITION 3. There will be more selective interven-tion in internal hybrid forms that operate in turbulentindustries than in internal hybrid forms that operate intranquil industries.

This proposition may be taken to be the otherside of the coin of the transaction cost argument thatexternal hybrids are unstable in "dynamic" industries(Williamson 1996) because in such industries unex-pected contingencies that may give rise to holdups aremore likely. Along similar lines, the argument under-lying Proposition 3 is that in dynamic industries, theimplicit contract between teams/projects and manage-ment in internal hybrids is likely to be relatively moreincomplete than in more tranquil industries. Therefore,management is likely to engage in more selective inter-vention in an attempt to infiuence how projects react tounexpected contingencies.

Internal and External Hybrids andInternal and External MarketsThe problem of selective intervention casts a novel lightover governance choices between internal and exter-nal hybrids and internal and external markets (Poppo1995). These organizational forms may be seen as ratherclose substitutes. For example, they may be adoptedin order to better exploit local knowledge (Cowen andParker 1997), or to strengthen incentives, because theymake agents residual claimants to a higher degree thanis the case in traditional hierarchies. However, whereasinternal hybrids/internal markets may suffer from prob-lems stemming from selective intervention, externalhybrids/external markets do not suffer from these. Ofcourse, external hybrids and markets may suffer frominefficiencies caused by holdup problems when spe-cific assets are deployed. However, creating competi-tion between suppliers, investing in hostages, having

344 ORGANIZATION SCIENCEAÔI. 14, No. 3, May-June 2003


some tapered integration, etc., may strongly reduce prob-lems related to holdup. The legal system also con-strains the holdup possibility, however imperfectly. Incontrast, a solution to the problem of avoiding harmfulselective intervention cannot rely on market forces orcourt-enforceable contracts. The implication is that, onaverage, external markets and external hybrids are likelyto have incentive properties that are superior to those ofinternal markets and internal hybrids, so that there willbe (transaction and production) cost penalties associatedwith the use of the latter. This results in the followingproposition:

PROPOSITION 4. Firms that choose external hybrids{markets) over internal hybrids {markets) will havea cost performance that is superior to those thatchoose internal hybrids {markets) over external hybrids{markets).

This reasoning may be seen as a variation of a familiartheme of transaction cost economics (WilUamson 1996),namely that vertical integration be considered the optionof last resort.

Managing Commitment to Not Selectively InterveneWhile theory suggests that the problem of committingto not selectively intervene is a tough one, we do seemto observe a substantial amount of delegation in real-world firms. This indicates that it is possible to crediblycommit to nonintervention. There are two fundamentalmethods that managers may use for this purpose. Bothessentially tie the hands of a would-be intervenor.

The first one is to commit oneself to being (ratio-nally) ignorant. Thus, a manager may choose not tobe informed about a number of critical dimensions inprojects. In very hierarchical organizations this may beeasy to accomplish because of the large informationaldistance between topmanagement and projects. A secondapproach proceeds by managers making it harmful tothemselves to selectively intervene. Open announcementof a nonintervention policy, making such policy recordedin company documents, working to install it in corporateculture, etc., all contribute to meeting this aim becauseit makes the possible clash between the communicatedvalues and managerial interventionist practice extremelysharp, and makes very obvious the break of the explicitlystated psychological contract (Brockner et al. 1992).

The Spaghetti Organization as a ModulationBetween Stable Organizational FormsAlthough organizational forms that break with a logic ofcomplementarity may incur penalties in terms of staticefficiency (i.e., economizing with transaction costs and

costs of production), they may still conceivably yieldbenefits in terms of dynamic efficiency (i.e., innova-tiveness). Calls for "chaotic" organization (Peters 1992)often implicitly make such arguments. Organizationdesign needs to consider both types of efficiencies(Ghemawat and Ricart i Costa 1993). An implication isthat in an intertemporal perspective, choosing "consis-tent" configurations of organizational elements may notnecessarily maximize the value of the firm. An inge-nious argument of this kind has been developed byNickerson and Zenger (2000). They suggest that consid-erations of efficiency may require modulating betweendiscrete organizational forms—such as the old hierar-chical Oticon organization and the postspaghetti matrixstructure—even in response to a stable set of environ-mental conditions. This is because the steady-state func-tionality delivered by a discrete organizational form mayitself be discrete, and the desired functionality may liein-between those delivered by the discrete organizationalforms. Efficiency gains may then be obtained by modu-lating between the forms.

If indeed the Oticon spaghetti organization may haveincurred inefficiencies with respect to the organization ofits administrative systems, it is hard to dispute that it wasalso a quite innovative organization (cf. Table 1). Thesebenefits may likely have overwhelmed the organizationalcosts. Although the spaghetti organization was not stablein the presence of the problem of selective intervention,it would still have made sense to choose this form, evenif the designers had known it to be inherently unstable.In fact, much of the early discussion of the spaghettiorganization made reference to the need to try some-thing entirely new and admittedly chaotic, for the pur-pose of drastically shaking up the original, bureaucraticorganization (Kolind 1990, Peters 1992, Poulsen 1993).This is consistent with Nickerson and Zenger's theory:The spaghetti organization may indeed be an exampleof modulating between the stable organizational formof the traditional, prespaghetti hierarchy, and the stablematrix organization post the spaghetti. What lends cre-dence to this interpretation is that although the hearingaids industry was technologically quite dynamic in therelevant period (Lotz 1998), it is not possible to iden-tify environmental changes that might have caused theorganizational change away from the spaghetti.

Firms and MarketsThe present discussion casts light over the classical issueof what the fundamental differences are between firmsand markets, and supports the original Coasian posi-tion that the key difference is that markets do not relyon resource allocation by means of authority, whereas

ORGANIZATION SciENCEAtoL 14, No. 3, May-June 2003 345


firms do (Coase 1937). "Authority" is a problematicword because it is often invested with a too-narrowmeaning, for example, detailed direction and supervi-sion (Foss and Foss 2002). Ultimately, the meaning ofhaving authority is that one can restrict the decisions ofone's subordinate, overrule him, and perhaps fire him.This means that although decision rights may be dele-gated, we can still trace the chain of authority in a firm,and we will always realize that ultimate decision-makingpower resides at the top. As this paper has illustrated,all subordinates' decision rights "are loaned, not owned"(Baker et al. 1999, p. 56). Fundamentally, it can neverbe otherwise. This is because ultimate decision-makingrights can only be transferred from bosses to subor-dinates in one way, namely by transferring ownership(Hart 1995). However, transferring ownership amountsto spinning-off the person to whom ownership is given.It means creating a new firm. It is this fundamentaldifference in how ownership is allocated that underliesthe problem of selective intervention. The analysis inthis paper thus makes direct contact with importantmodem theories of economic organization (Hart 1995,Williamson 1996, Baker et al. 2002) that stress theimportance of ownership for the understanding of thenature of firms and firm boundaries.

ConclusionsTo many firms, the adoption of new, hybrid organiza-tional forms is increasingly seen as imperative. However,rather little theoretical and empirical research has treatedparticularly internal hybrids. This paper has examineda specific experiment with adopting and later stronglymodifying a radical intemal hybrid in an attempt to iden-tify some possible liabilities of the adoption of suchorganizational forms. In particular, the focus has beenon motivational problems that may be caused by prob-lems of committing to refraining from harmful selec-tive intervention. A main argument was that problems ofselective intervention are particularly prevalent in orga-nizations that adopt radical intemal hybrids. In contrast,firms with more traditional hierarchical structures bettershield themselves from these problems. Managers maycommit to nonintervention by means of rationally choos-ing to be ignorant or by making it harmful to themselvesto selectively intervene. Finally, the problem of selectiveintervention is a prime candidate for understanding theincentive liabilities of hierarchies and intemal hybridsvis-a-vis markets or extemal hybrids.

Although this paper has thus exemplified the interpre-tive power of organizational economics, admittedly orga-nizational economics only tells a part of the story. From

an organizational economics perspective, the spaghettiorganization represented a matrix of rights and incen-tives that are helpful for understanding its liabilities andhow these liabilities gave rise to certain organizationaldynamics (i.e., the partial abandonment of the spaghettiorganization). However, it may indeed also be under-stood in terms of an attempt to, for example, fosterdynamic capabilities (Ravasi and Verona 2000), a per-spective that lies outside of organizational economics.Thus, the full story of the Oticon spaghetti experimentrequires that more than one perspective be considered.Relatedly, the paper has suggested that organizationaleconomics should consider to a fuller extent psycholog-ical insights in motivation and in cognition. While it ispossible to tell stories of managerial commitment, selec-tive intervention, and stified incentives based only onorganizational economics, there is little reason to be sonarrow. A vast literature on procedural justice in orga-nization, psychological contracts, and (biased) cognitionexists, the insights of which may be combined withorganizational economics insights in order to further theunderstanding of problems of managerial commitment,including problems of selective intervention (cf. alsoMiller 1992, Lindenberg 2000).

AcknowledgmentsThe author acknowledges discussion with, and helpful comments from,Kirsten Foss, Robert Gibbons, Anna Grandori, Henrik Hoick, Lars Kolind,Kristian Kreiner, Peter Lotz, Volker Mahnke, Peter Maskell, Mette Mors-ing, two anonymous reviewers, and Senior Editor Axel von Werder, aswell as seminar audiences at Copenhagen Business School, the UniversitatFreiburg, Norges Handelsh0yskole, and the University of Pisa on earlierdrafts of this paper. All mistakes, errors of interpretation, etc. are solelythe responsibility of the author.

Endnotes'Zenger (2002) argues that much more work exists on extemal hybridsthan on intemal hybrids, investigation of the latter being largelyconfined to work on the multidivisional form. And Poppo (1995,p. 1,845) points out that "[ejmpirical work that examines the differ-ences between intemal and extemal markets are r a r e . . . . Theory inthis area is also limited."

^However, a main purpose of conducting analysis of single cases oftenis to be able to pose competing explanations for the same set of events(and perhaps to indicate how these explanations may be applied toother situations) (Yin 1989). Moreover, basic considerations of inter-nal validity dictate that altemative explanations be considered. How-ever, while I shall indeed make reference to and discuss other possibleexplanations of some of the relevant events (e.g., ideas from motiva-tion theory and information processing theory), the main emphasis ison developing one specific interpretation. While an eclectic, multiple-perspective approach may be superior in the abstract, more insightmay arguably be provided in the concrete by pursuing, in a relativelynarrow fashion, one specific interpretation and explore the limits ofthis interpretation.



'See Lotz (1998) for a careful analysis of the hearing aids indus-try, with particular emphasis on pattems of innovation. The historyof Oticon prior to the introduction of the spaghetti organization isextensively covered in Poulsen (1993) and Morsing (1995), and, morebriefly, in Gould (1994) and Lovas and Ghoshal (2000, pp. 877-878).•"The Oticon case is reportedly the best-selling IMD case (Gould1994) ever (B0rsens Nyhedsmagasin 8 November, 1999). Kolind'sdramatic and symbol-laden way of implementing the spaghetti orga-nization, as well as the form itself, are still being given extensivetreatment in management textbooks (e.g., Boddy and Paton 1998).'As Kolind explained to Gould (1994, p. 465): "We quickly agreedthat all employees would have a portfolio of jobs, and we were tough;we said at least three jobs, with the main one in their profession orusing their greatest competence, and the other two in outside areas.This concept really expands an organization's resources: engineersare doing marketing, marketing people manage development projects,and financial people help with product development."'Although the variance on the distribution of salaries was increasedas a result of the new reward schemes, average salaries do not appearto have changed. In fact, average Oticon salaries have been, and stillare, comparatively low, particularly for software developers. Intrinsicmotivation is a key aspect of Oticon motivation policies, and is seenas complementary to (rather than substituting for) extrinsic motiva-tion. On intrinsic and extrinsic motivation, and its implications fororganizational theory, see Frey (1997) and Osterloh and Frey (2000).'Oticon's growth in the 1990s largely represented growth of marketshare, because the size of the market for hearing aids stagnated inthat decade.

*Much recent management literature has suggested that firms involatile elements need to emulate markets to the largest possibleextent (e.g., Halal et al. 1993, Cowen and Parker 1997).'For a full comparative analysis, see Nickerson and Zenger (2001).'"The possibility that extemal hybrids or market contracting may bealtematives to intemal hybrids never seems to have been consideredin Oticon. Thus, that incentives may be strengthened by relying on thereal market (rather than the simulated intemal one) through spinning-off functions and departments (Aron 1991) does not appear to havebeen seen as a serious altemative to intemal disaggregation."Exceptions may occur when giving subordinates more extensiverights (e.g., a package of initiation, ratification, and implementationrights) strengthens employee incentives (see Aghion and Tirole 1997,Baker et al. 1999, and Foss and Foss 2002 for analyses of this).'^For example, it could reflect attempts to curb moral hazard in projectteams. However, the increased use of high-powered incentives andmore widespread employee ownership were designed to remedy prob-lems of moral hazard.

' 'The change was assisted by the symbolic acts undertaken by Kolind,which helped to signal his commitment to the change (Hermalin1998). For example, Kolind invested 25 million Dkr (approximately4 million USD) of his own funds in the firm,'''in addition, a motivation theory perspective would suggest thatwhile employees' lower-level needs were not sufficiently satisfied(low income, uncertainty due to the reorganization and layoffs),management already tried to address their higher-level needs (morecomprehensive tasks, more responsibility). Thanks to an anonymousreviewer for this point.

''Eskerod (1997, 1998) in particular documents this. My later inter-view with the chief HRM officer strongly confirmed Eskerod's findingthat the multijob concept had severe costs in terms of problems ofcoordination and fmstrating employees.

"And neither would this have been possible, as nobody in Oticon,not even the Projects and Products Committee, kept track of the totalnumber of development projects. Records were only kept ofthe 10-20major projects. An estimate is that under the spaghetti organization,an average of 70 projects were continously running (Eskerod 1998,p. 80).

" in fact, the more structured project organization gradually imple-mented from 1996 has established controls that secure that the coor-dination and time-allocation problems that beset the original spaghettiorganization are kept at bay.'Înterview with HRM manager Henrik Hoick."Somewhat later, the literature on intemal transfer prices revealedthe existence of various incentive problems that may beset this orga-nizational practice (e.g., Holmstrom and Tirole 1991).^"This should be understood in a broad sense: A "project" may referto many different types of decisions or clusters of decisions.^'See Simons (2002) for a highly pertinent discussion of employees'perception of the fit between managers' words and actions and themotivational consequences of this perception.^^Since behavior was apparently difficult to measure, a more output-based system could have been tried (Prendergast 1999), for example,contracts that specified rewards for specific accomplishments (e.g.,a system that rewarded according to milestones in a developmentproject). However, it is doubtful whether such a contract could actu-ally be made court enforceable. A managerial commitment problemwould again result.

ReferencesAghion, Philippe, Jean Tirole. 1997. Formal and real authority in

organization. J. Political Econom. 105 1-29.

Argyres, Nicholas, Vai-Lam Mui. 2000. Rules of engagement, infor-mal leaders, and the political economy of organizational dissent.Unpublished paper.

Aron, Debra J. 1991. Using the capital market as a monitor: Cor-porate spin-offs in an agency framework. RAND J. Econom. 22505-518.

Baker, George, Robert Gibbons, Kevin J. Murphy. 1999. Informalauthority in organizations. J. Law, Econom. Organ. 15 56-73.

, , . 2002. Relational contracts and the theory of thefirm. Quart. J. Econom. 117 39-83.

Bamey, Jay B. 1986. Organizational culture: Can it be a source ofsustained competitive advantage? Acad. Management Rev. 11656-665,

Baron, James N., David M. Kreps. 1999. Consistent human resourcepractices. California Management Rev. 41 29-53.

Barzel, Yoram. 1997. Economic Analysis of Property Rights.

Cambridge University Press, Cambridge, MA.

Bazerman, Max H. 1994. Judgment in Managerial Decision Making.Wiley, New York.

Boddy, David, Robert Paton. 1998. Management: An Introduction.Prentice Hall, London, U.K.

O R G A N I Z A T I O N SciENCEA'bl. 14, No. 3, May-Jutie 2003 347

NICOLAt J. FOSS Selective Intervention and Internal Hybrids

B0rsens Nyhedsmagasin. 1991. For Gud, Schr0der og Oticon.(1 Marts).

. 1999. Opg0r med Kolinds kaos. (8 November) 14-22.

Brockner, J., T. Tyler, R. Cooper-Schneider. 1992. The influence of

prior commitment to an institution on reactions to perceived

unfaimess: The higher they are, the harder they fall. Admin. Sci.

Quart. 37 241-261.

Business Intelligence. 1993. A non-traditional performance and pro-cess measurement system. Mette Morsing, Kristian Eiberg, eds.Managing the Unmanageable for a Decade. Oticon, Hellemp,Denmark.

Casson, Mark. 1994. Why are firms hierarchical? Internat. J. Econom.

Bus. 1 47-76.

Coase, Ronald H. 1937. The nature of the firm. Economica (N.S.) 4

386-405.

Cowen, Tyler, David Parker 1997. Markets in the Firm: A Market

Process Approach to Management. The Institute of Economic

Affairs, London, U.K.

Daft, Richard L., Arie Lewin. 1993. Where are the theories for the

"new" organizational forms? An editorial essay. Organ. Sci. 4

i-iv.

Day, Jonathan, Jim Wendler. 1998. The new economics of organiza-

tion. McKinsey Quart. 23 4-18.

Demsetz, Harold. 1988. The theory of the firm revisited. J. Law,Econom. Organ. 4 141-161.

Dewatripont, Matthias, G. Roland. 1995. The design of reform pack-ages under uncertainty. Amer. Econom. Rev. 85 1207-1223.

Dosi, Giovanni. 1982. Technological paradigms and technological tra-

jectories. Res. Policy t t 147-162.

Eskerod, Pemille. 1997. Nye perspektiver pa fordeling af menneske-lige ressourcer i et projektorganiseret multiprojekt-milj0. Ph.D.thesis, Handelsh0jskole Syd, S0nderborg.

. 1998. Organising by projects: Experiences from Oticon's prod-uct development function. Mette Morsing, Kristian Eiberg, eds.Managing the Unmanageable for a Decade. Oticon, Hellemp,Denmark.

Fama, Eugene, Michael C. Jensen. 1983. Separation of ownership and

control. / Law and Econom. 26 301-325.

Foss, Nicolai J. 2002. "Coase vs Hayek" Economic organization andthe knowledge economy. Intern. J. Econom. Bus. 9 9-36.

, Kirsten Foss. 2002. Authority and discretion: Tensions, cred-ible delegation, and implications for new organizational forms.Downloadable from http://www.cbs.dk/link/papers

Frey, Bmno. 1997. Not Just for the Money. Edward Elgar, Aldershot,U.K.

Galbraith, Jay R. 1974. Organization design: An information process-

ing view. Interfaces 4 28-36.

Ghemawat, Pankaj, Joan E. Ricart i Costa. 1993. The organizationaltension between static and dynamic efficiency. Strategic Manage-ment J. U 59-73.

Gifford, Sharon. 1992. Allocation of entrepreneurial attention. / .Econom. Behavior and Organ. t9 265-284.

Glaser, B., A. Strauss. 1967. The Discovery of Grounded Theory.

Aldine, Chicago, IL.

Gould, R. Morgan. 1994. Revolution at Oticon A/S: The spaghettiorganization. Soumitra Dutta, Jean-Francois Manzoni, eds. 1999.Process Re-engineering, Organizational Change and Perfor-mance Improvement. McGraw-Hill, London, U.K.

Grant, Robert M. 1996. Prospering in dynamically-competitive envi-ronments: Organizational capability as knowledge integration.Organ. Sci. 7 375-387.

Halal, William E., K. B. Taylor. 1998. Twenty-First Century Eco-nomics: Perspectives of Socioeconomics for a Changing World.St. Martin's Press, New York.

, Ali Geranmayeh, John Pourdehnad. 1993. Internal Markets:Bringing the Power of Free Enterprise Inside Your Organization.

Wiley, New York.

Hart, Oliver. 1995. Firms, Contracts and Financial Structure.Clarendon Press, Oxford, U.K.

Hayek, Friedrich A. von. 1945. The use of knowledge in society, inidem. 1948. Individualism and Economic Order. University ofChicago Press, Chicago, IL.

Helper, Susan, John Paul MacDuffie, Charles Sabel. 2000. Pragmatic

collaborations: Advancing knowledge while controlling oppor-

tunism. Indust. and Corporate Change 9 443-487.

Hemmer, Thomas. 1995. On the interrelation between produc-

tion technology, job design, and incentives. / Accounting and

Econom. 19 209-245.

Hermalin, Benjamin. 1998. Toward an economic theory of leadership:Leading by example. Amer. Econom. Rev. 88 1188-1206.

Holmstrom, Bengt. 1999. The firm as a subeconomy. / Law, Econom.Organ. t 5 74-102.

, Paul Milgrom. 1991. Multitask principal-agent analyses: Incen-tive contracts, asset ownership, and job design. / Law, Econom.Organ. 7 24-52.

, Joan Ricart i Costa. 1986. Managerial incentives and capitalmanagement. Quart. J. Econom. 101 835-860.

, Jean Tirole. 1991. Transfer pricing and organizational form. J.

Law, Econom. Organ. 7 201-228.

Jensen, Frank Dybdal. 1998. Vcerdibaseret Ledelse—styring mellemregler og visioner. Jurist- og 0konomforbundets Forlag,Copenhagen, Denmark.

Jensen, Michael C , William H. Meckling. 1992. Specific andgeneral knowledge and organizational structure. Lars Werin,Hans Wijkander, eds. 1992. Contract Economics. Blackwell,Oxford, U.K.

, Karen Wmck. 1998. Science, specific knowledge and total

quality management. Michael C. Jensen, ed. Foundationsof Organizational Strategy. Harvard University Press, Cam-bridge, MA.

Kogut, Bruce, Udo Zander. 1992. Knowledge of the firm, combina-tive capabilities, and the replication of technology. Organ. Sci. 3387-397.

Kolind, Lars. 1990. Think the unthinkable. Mette Morsing, KristianEiberg, eds. 1998. Managing the Unmanageable For a Decade.Oticon, Hellemp, Denmark.

. 1994. The knowledge-based enterprise. Mette Morsing, Kris-tian Eiberg, eds. 1998. Managing the Unmanageable for aDecade. Oticon, Hellemp, Denmark.

348 ORGANIZATION SCIENCE/VOI. 14, No. 3, May-June 2003


LaBarre, Polly. 1996. This organization is dis-organization.Wysiwyg// 93/http://fastcompany.com/online/03/oticon.html

Laursen, Keld, Nicolai J. Foss. 2002. New HRM practices, comple-mentarities, and the impact on innovation performance. Cam-bridge J. Econom. Forthcoming.

Lazear, Edward. 1995. Personnel Economics. MIT Press, Cambridge,MA.

Liebeskind, Julia Porter. 1996. Knowledge, strategy and the theory ofthe firm. Strategic Management J. 17 441-452.

Lindenberg, Siegwart. 2000. It takes both trust and lack of mis-trust: The workings of cooperation and relational signaling incontractual relationships. J. Management and Governance 411-33.

Lotz, Peter. 1998. The paradox of high R&D and industry stability:Technology and structural dynamics in the global hearing instru-ments industry. Indust. Innovation 5 113-137.

Lovas, Bjom, Sumantra Ghoshal. 2000. Strategy as guided evolution.Strategic Management J. 21 875-896.

Lyregaard, Poul-Erik. 1993. Oticon: Erfaringer og faldgruber. SteenHildebrandt, Leif H. Aiken, eds. Pa vej mod helhedssyn i ledelse.Ankerhus.

Mandag Morgen. 1991. Oticon satser dristigt, men resultateme man-gier. (15 Feb.) 15-18.

Meyer, Alan D., Anne S. Tsui, C.R. Hinings. Configurationalapproaches to organizational analysis. Acad. Management J. 361175-1195.

Miles, Raymond E., Charles C. Snow. 1992. Causes of failure innetwork organizations. California Management Rev. 35 53-72.

, , John A. Mathews, Grant Miles, Henry J. Coleman, Jr.1997. Organizing in the knowledge age: Anticipating the cellularform. Acad. Management Executive 11 7-20.

Milgrom, Paul. 1988. Employment contracts, influence activities andefficient organization design. J. Political Econom. 96 42-60., John Roberts. 1990. The economics of modem manufacturing:

technology, strategy and organization. Amer. Econom. Rev. 80511-528.

Miller, Gary. 1992. Managerial Dilemmas. Cambridge UniversityPress, Cambridge, MA.

Morsing, Mette. 1995. Omstigning til Paradis? Oticon i processenfra hierarki til spaghetti. Copenhagen Business School Press,Copenhagen, Denmark.

, Kristian Eiberg, eds. 1998. Managing the Unmanageable for aDecade. Oticon, Hellerup, Denmark.

Mosakowski, Elaine. 1998. Entrepreneurial resources, organizationalchoices, and competitive outcomes. Organ. Sci. 9 625-643.

Nahapiet, Janine, Sumantra Ghoshal. 1999. Social capital, intellec-tual capital, and the organizational advantage. Acad. ManagementRev. 23 242-266.

Nickerson, Jackson, Todd Zenger. 2000. Being efficiently fickle: Adynamic theory of organizational choice. Unpublished paper,David Eccles School of Business, University of Utah.

, . 2001. A knowledge-based theory of govemance choice:A problem-solving approach. Unpublished paper, David EcclesSchool of Business, University of Utah.

North, Douglass C. 1990. Institutions, Institutional Change,and Economic Performance. Cambridge University Press,Cambridge, MA.

Osterloh, Margit, Bruno Frey. 2000. Motivation, knowledge transferand organizational form. Organ. Sci. 11 538-550.

Peters, Tom. 1992. Liberation Management. Alfred A. Knopf,New York.

Poppo, Laura. 1995. Influence activities and strategic coordination:Two distinctions of intemal and extemal markets. ManagementSci. 41 1845-1859.

Poulsen, Per Thygesen. 1993. Tcenk det utxnkelige—revolutionen iOticon. Schultz, Copenhagen, Denmark.

Prendergast, Canice. 1999. The provision of incentives in firms.J. Econom. Literature 37 7-63.

Ravasi, Davide, Gianmario Verona. 2000. Organizing the processof knowledge integration: The benefits of structural ambiguity.Scandinavian J. Management. Forthcoming.

Robinson, S. L., E. W. Morrison. 1995. Organizational citizen-ship behaviour: A psychological contract perspective. / Organ.Behaviour 16 289-298.

, Denise M. Rousseau. 1994. Violating the psychological con-tract: Not the exception but the norm. J. Organ. Behaviour 15245-259.

Simons, Tony. 2002. Behavioral integrity: The perceived alignmentbetween managers' words and deeds as a research focus. Organ.Sci. 13 18-35.

Teece, David J., Giovanni Dosi, Richard P. Rumelt, Sidney G. Winter.1994. Understanding corporate coherence: Theory and evidence.J. Econom. Behavior and Organ. 23 1-30.

Thompson, James D. 1967. Organizations in Action. McGraw-Hill,New York.

Vancil, Robert, Lee E. Buddrus. 1979. Decentralization: ManagerialAmbiguity by Design. Irwin, Homewood, IL.

Verona, Gianmario, Davide Ravasi. 1999. Organizational capabilitiesfor continuous innovation. Unpublished paper, Bocconi Univer-sity.

von Hippel, Eric. 1994. "Sticky information" and the locus of prob-lem solving: Implications for innovation. Management Sci. 40429^39.

von Mises, Ludwig. 1949. Human Action. Fox and Wilkes, San Fran-cisco, CA.

Williamson, Oliver E. 1996. The Mechanisms of Governance. OxfordUniversity Press, Oxford, U.K.

Yin, Robert K. 1989. Case Study Research: Design and Methods.Sage, London, U.K.

Zenger, Todd. 2002. Crafting intemal hybrids: Complementarities,common change initiatives, and the team-based organization.Intern. J. Econom. Bus. Forthcoming.

, William S. Hesterly. 1997. The disaggregation of corporations:Selective intervention, high-powered incentives, and molecularunits. Organ. Sci. 8 209-222.

Zucker, Lynne. 1991. Markets for bureaucratic authority and control:Information quality in professions and services. Res. Sociologyof Organ. 8 157-190.

ORGANIZATION SCIENCE/VOL 14, No. 3, May-June 2003 349

Documents

Selective Intervention and Internal Hybrids: Interpreting ... · Selective Intervention and Internal Hybrids: Interpreting and Learning from the Rise and Decline of the Oticon Spaghetti