Co-Evolution of Capabilities and Industry the Evolution of Mutual Fund Processing

8/3/2019 Co-Evolution of Capabilities and Industry the Evolution of Mutual Fund Processing

1/19

Co-Evolution of Capabilities and Industry: The Evolution of Mutual Fund ProcessingAuthor(s): Daniel Levinthal and Jennifer MyattSource: Strategic Management Journal, Vol. 15, Special Issue: Competitive OrganizationalBehavior (Winter, 1994), pp. 45-62Published by: John Wiley & SonsStable URL: http://www.jstor.org/stable/2486810 .

Accessed: 18/07/2011 07:34

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless

you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you

may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at .http://www.jstor.org/action/showPublisher?publisherCode=jwiley. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed

page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

John Wiley & Sons is collaborating with JSTOR to digitize, preserve and extend access to Strategic

Management Journal.

http://www.jstor.org
http://www.jstor.org/action/showPublisher?publisherCode=jwileyhttp://www.jstor.org/stable/2486810?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/action/showPublisher?publisherCode=jwileyhttp://www.jstor.org/action/showPublisher?publisherCode=jwileyhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/2486810?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=jwiley


2/19

Strategic Management Journal, Vol. 15, 45-62 (1994)

CO-EVOLUTIONOF CAPABILITIESND INDUSTRY:THE EVOLUTIONOF MUTUALFUND PROCESSINGDANIELLEVINTHALnd JENNIFERMYATTThe WhartonSchool, University of Pennsylvania, Philadelphia,Pennsylvania, U.S.A.

The resource view of the firm has made substantial progress in identifying what attributesof a firm may provide a source of competitive advantage; however, the literature has farless to say about the emergence of these distinctive capabilities. We develop a simpleframework based on the role of positive feedback effects of market activity, organizationalfactors that cause a firm to focus on a particular capability trajectory, and lastly the roleof managerial choice with respect to anticipated feedback effects, which we termfeedforwardeffects. We apply this framework to the emergence of competitive positions in the mutualfund-processing business from its inception to 1984.

The field of Strategic Management, or BusinessPolicy as it was termed in its early incarnations,has had as a primary mission the analysis ofdiversity. In early work, the focus was ondistinctive competencies among organizations(Selznick, 1957; Andrews, 1971). As the fieldbegan to relabel itself Strategic Management,the primary interest became diversity acrossindustries (Porter, 1980). The movement to theindustry level of analysis resulted from thesuccessful application of the techniques of indus-trial organization economics to questions ofcompetitive strategy (Porter, 1980). The recentmovement back to the firm level of analysisseems to have been driven more by empiricalanomalies and challenges to management practicethan the 'technological opportunity' representedby a set of powerful analytical techniques.In many industries, global competition hasincreased the number of players in a given productmarket and thereby reduced the importance ofstrategic (in the game theoretic sense of theword) behavior. For instance, in the 1950s

Key words: Organizational apabilities,mutual undsCCC0143-2095/94/100045-18(? 1994by John Wiley & Sons, Ltd.

the automobile industry was fertile ground forexamining price collusion (Bresnahan, 1987), whilecurrentanalysisof the industry focuses on variationin manufacturing systems (Womack, Jones, andRoos, 1990) and product development (Clark andFujimoto, 1991). Empirical research regarding themodest returns to diversification efforts (Rumelt,1974;Porter, 1987)provides some indirectevidencethat the choice of industry may be less critical tofirm performance than the presence of distinctivecapabilities to operate within a given industry.Rumelt's (1991) efforts at disentangling businessunit effects from corporate and industry effectsprovide more direct evidence of this.This shift to a firm level of analysis on thepart of researchers has engendered tremendousexcitement in the research literature (Barney,1991) and has had a profound impact on thediscourse of practitioners whose conversationsare now replete with words such as corecompetence and capabilities. In assessing thisemerging research stream, which generally goesunder the label 'the resource view of the firm,'there appears to have been substantial progressin identifying what attributes of a firm mayprovide a source of competitive advantage


3/19

46 D. Levinthal and J. Myatt(Barney, 1991) as well as significant insightas to the sustainability of these advantages(Ghemawat, 1986; Rumelt, 1984). However, asFoss, Knudsen, and Montgomery (1993) note,the characterizations of what constitute a valuableresource or capability tend to be ex-post. Theliterature has far less to say about the emergenceof these distinctive capabilities. There is also asense that the ebb and flow of strategy researchmay have swung excessively to firm-centeredanalyses and has tended to ignore industrydynamics (Levinthal, 1995).The present paper is an effort to make somemodest progress toward redressing these twoweaknesses in the literature on firm capabilities.We consider these issues from the perspective ofevolutionary systems, in particular emphasizingtherole of positive feedback effects (Arthur, 1990).Positive feedback effects reinforce a process in itscurrenttrajectory. Arthur (1989) has applied thesenotions to understand the role of installed base ofusers on competition among alternativetechnologi-cal standards. We explore the implications ofpositive feedback of market activity on capabilityformation. Some of these are relation specific, i.e.within a particular customer-supplier pair, whileothers are more diffuse and result from overallmarketactivity.We also consider the organizationalfactors that influence the degree to which a firmmight pursue a given set of opportunities. Inaddition to recognizing these organizationalresources and constraints, managers make choicesconcerning resource allocation and market initiat-ives based on anticipated feedback effects. Weterm such considerations feedforward effects. Thisapproach allows us to address both the issue ofthe emergence of diversity among firms, as wellas allows us to examine the coupling betweenindustry-levelforces and these firm-leveldynamics.We apply this conceptual framework to thedevelopment of the mutual fund-processing indus-try.

CO-EVOLUTION OF CAPABILITIESAND INDUSTRY'How a firm's capabilities evolve is intimatelylinked with how the markets that a firm serves

1 The material in this subsection draws upon Levinthal (1995).

evolve. This coupling is not merely the result offinancial flows derived from profitable marketsproviding the basis for investment resources.There are clearly instances in which such acoupling is important, but the presence of well-developed capital markets in Western economiestends to mitigate the importance of purelyfinancial couplings. More importantly, manyorganizational capabilities emerge, are refined,or decay as a result of, or an absence of,product market activity. Therefore, the particularsubmarkets a firm serves will engender a distinc-tive, though not necessarily unique, set ofcapabilities. Certainly many capabilities, parti-cularly research activity, do not follow directlyfrom current operations. However, even withregard to such investments, the incentives that afirm has to make such investments and thepolitical forces internal to the firm that mayinfluence such decisions are not independent ofits current product market activities.The fate of a firm's capabilities, however, isnot purely a deterministic outcome of suchcouplings. Indeed, the couplings themselves area critical management decision. What marketsshould the firm serve? What activities should beperformed within the firm and what sorts ofexternal linkages should the firm make? Thesechoices provide managerial discretion over theevolutionary path that the firm's capability settakes. The framework suggested here is somewhatakin to the distinction Argyis and Schon (1978)make between first-order and second-order learn-ing processes. The impact on capabilities ofserving particular markets is analogous to a first-order learning process. While not automatic,first-orderlearning processes are a direct outcomeof the existing structure. By establishing a newset of linkages, whether by choice of a newsubmarket to serve, a new set of customerrelations, or a new internal organizational struc-ture, management sets in motion a new directionfor the development of the firm's capabilities.FeedbackFeedback effects can greatly amplify the existingheterogeneity in a population of organizations.Perhaps the most prominent examples of suchfeedback effects in the strategy literature revolvearound the reenforcing advantages of marketposition (Rumelt, 1984). In particular, the notion


4/19

Co-evolution of Capabilities and Industry 47of a learning or experience curve implies thatgreater market share will lead to lower costs,which in turn will tend to enhance the firm'smarket share (Boston Consulting Group, 1972).The notion of absorptive capacity (Cohen andLevinthal, 1989, 1990) has a similar implication.Organizations with more expertise in a particulartechnical domain will more readily acquiresubsequent knowledge in that domain. Marketposition advantages associated with a firm's scaleof operations may also be self-reinforcing. AsCohen and Klepper (1992) argue in the contextof firms' investment in research and developmentactivity, the ability of firms to appropriate thereturns to such investments is often a functionof their existing sales base.

The notion of brand equity (Aaker, 1991)offers another illustration of such market positionadvantages. Not only does a strong brandnameenhance the margins a firm receives on aparticular product, but the ability of a firm toplace that brandname on other related productsenhances the ability of the firm to extend itsproduct line. In addition, the introduction ofthese product extensions which share the estab-lished name result in a higher level of brandawareness, further enhancing the brand equity.Of course the fungibility of brandname is notboundless and, indeed, extensions of brandnamesto inappropriate domains may negate their value.An important caveat to these market positionaladvantages is that they are self-reinforcing incompetitive environments in which the bases ofcompetitive advantage are stable. Conversely,in changing environments, these same self-reinforcing mechanisms may lead to a decline ina firm's competitive position (Levinthal, 1992).The needs of existing customers may change orthe subfields and customers that the firm servesmay decline. More basic changes such as thetechnology underlying an industry or the regula-tory structure may disrupt the drivers of success(Abernathy and Clark, 1985).Indeed, feedback effects may reduce thelikelihood that a firm will successfully adapt tosuch changes. For instance, an established firmmay have more incentive to invest in incrementalchanges in a current technology than in exploringmore radical innovations. More generally, Levin-thal and March (1993) argue that myopia tends tobe an important property of many organizationallearning processes. The returns to exploiting

existing knowledge and capabilities tend to bemore certain and immediate than the returnsto the exploitation of novel capabilities andopportunities. Furthermore, the past exploitationactivities in a given domain tend to make furtherexploitation in that domain even more attractivedue to various sorts of competency learning. Thispositive reinforcement of activities in the currentdomain in which returns are relatively certainand favorable tends to drive out search foralternative bases of action.2Initial conditionsFeedback effects influence a firm's capabilitieswithin the context of an ongoing relationship.However, when a firm enters a product area ornew exchange relationship, its capabilities areheavily influenced by its prior activities. Suchfeedback effects that precede the current marketactivity will be referred to as initial conditions.While a number of initial conditions are specificto a particular firm, such as its bundle ofcospecialized assets (Teece, 1986), others, suchas geographic effects, influence a broad set offirms.Geographic effects on firms' competitivenesshave been a recent source of inquiry in theeconomics and strategy literatures (Arthur, 1986;Krugman, 1991; Porter, 1990). Krugman drawson Marshall's (1920) arguments regarding non-pecuniary externalities. Contiguous location offirms within the same industry or that draw onsimilar skills allow firms to take advantage of apool of workers with specialized skills. Moregenerally, an infrastructure of physical systems,specialized supporting organizations, as wellas skilled workers, develops. In addition, thecontiguous location of the firms enhances theirability to learn from each other (Enright, 1993).Many initial conditions are specific to a firm.Fichman and Levinthal (1991) characterize awide variety of 'assets' that an organizationmay have even at its founding. For instance,organizations may have pre-existing ties tocustomers. Alternatively, the founders themselves2 Christensen and Bower (1994) provide a compelling exampleof this phenomenon in the context of the disk drive industry.Strong existing customer relations provide an incentive toadvance the existing technological trajectory; however, whennew technological opportunities do not correspond to currentcustomer needs they tend not to be pursued.


5/19

48 D. Levinthal and J. Myattmay bring with them important skills and industryexpertise. Few firms enter a new industry subfieldwith a clean slate (Mitchell, 1989). The vastmajority of entrants possess some existing base ofcapabilities. These capabilities may have beendeveloped through firm operations in other mar-kets, but even new firms inherit capabilities fromthe previous experiences of the firm's founders.FeedforwardThe previous section characterizes some ways inwhich heterogeneity among firms may becomeamplified over time. Prescient managers appearto recognize such feedback effects in makingdecisions about what industries or emergingsubfields to enter and which clients may helpfurther the firm's development. Thus, in makinga choice about what markets to serve, a firm ismaking a bet on a coevolutionary process. Thefirm is, or should, not only be concerned aboutits current capability to compete within thatdomain, but also with how participating in thatparticular industry or subfield will affect thefirm's future capabilities.Perhaps the most basic attribute of the marketsand customers served that will impact thedevelopment of the firm's capabilities is theirgrowth rate. Is the firm serving customers andmarket segments that are growing rapidly andthereby can provide a rich experience base forthe firm? In addition, leading-edge customersmay expose the firm to advances in technology(Von Hippel, 1988) and product offerings.For instance, consider a firm that wishes toparticipate in the mutual fund industry. One ofthe critical choices that such a firm must makeis its mode of distribution. Will the firm relyon direct sales or selling through a financialintermediary such as a bank or independentfinancial agent? The firm's choice of distributionsystem may set it on a virtuous or destructivecoevolutionary process. The firm must bothassess the overall market penetration of thatparticular mode of distribution and the feedbackeffects of serving that submarket.Similarly, consider the coevolution of a firm'scapabilities and industry when the industry is indecline. As markets shrink, so does reinvestmentin equipment. This yields a vintage effect on thefirm's production capabilities. Indirectly, thishas further implications for a firm's technical

capabilities. For instance, top-flight engineersvalue being able to work on the latest equipmentand the associated technical challenges; therefore,as the equipment begins to become dated thefirm begins to lose some of its most valuableengineering talent.Certain subfields impose greater or fewertechnical demands on a firm than others. Forinstance, in printed circuit boards, if a firm onlyparticipates in the defense sector it will notdevelop the technical competence to deal withthe high-density requirement of commercialmarkets. Thus, serving leading commercial mar-kets is a critical means of developing newcapabilities within printed circuit board manufac-turing.

Along the same lines, certain customers maydrive the firm to enhance their capabilities. Forinstance, State Street Bank, within the mutualfund industry, has such a large proportion of themarket for mutual fund processing that it islikely to experience first any new demandassociated with new instruments, such as financialderivatives or new services. In the process ofserving this demand for a new service, the firmenhances its capabilities relative to competitors.Similarly, R. R. Donnelley, a commercial printingfirm which traditionally focused exclusively onthe U.S. market, has developed significantforeign operations as it followed its clienttelecommunication companies overseas (Carey,1993). In addition, serving the computer industrywas the catalyst for Donnelley learning aboutemerging technologies (Carey, 1993).The role suggested here of leading-edge cus-tomers is analogous to Porter's (1990) discussionof demand factors associated with industryperformance across nations. Porter (1990) pointsto two critical attributes of home country demand.One is timing. Does the home country tend tobe early or late in its demand for a particularclass of new products or services? The second isthe level of sophistication and the degree towhich customers are demanding in their qualityrequirements. These factors influence the speedand direction with which organizations proceedalong their evolutionary trajectory.

Focusing forcesThe notion of feedforward suggests a significantrole for managerial discretion. By its choice of


6/19

Co-evolution of Capabilities and Industry 49customers and submarkets, top management hasan important role in influencing the evolution ofthe firm's capability set. These choices aboutexternal linkages, however, may be constrainedby the internal attributes of the firm. Obviously,the existing capability set determines what exter-nal linkages are competitively viable. As Teece,Pisano, and Shuen (1990) suggest in theirdiscussion of dynamic capabilities, there are avariety of focusing forces that tend to constraina firm to a particular capability trajectory.The term 'learning' often carries with it theconnotation of an activity or idea being novel.However, learning processes often serve toconstrain the range of organizational activities(Levinthal and March, 1993). As noted earlier,rapid learning may result in a competency trapwhereby increasing skill at the current proceduresmake experimentation with alternatives progres-sively less attractive. Along similar lines, Cohenand Levinthal (1989, 1990) argue that theability of firms to evaluate and utilize outsideknowledge-what they term absorptivecapacity-is a function of their prior relatedknowledge. As a result, firms will tend to confinethemselves to a limited set of technologicaldomains and have difficulty responding to devel-opments outside these areas.Teece et al. (1990) point out that the presenceof cospecialized assets (Teece, 1986) will also actas a constraining force of the development of firmcapabilities. They suggest that complementary orcospecialized assets whose value is enhanced bya particular class of innovative activities will steerthe firm's resources in that direction. One canthink of this argument in the context of Sutton's(1991) work on the effect of sunk costs onfirm and industry development. A preexistingcospecialized asset is a sunk cost, which thencan influence the firm's subsequent investmentdecisions.The political structure of a firm representsanother important class of constraining forces.There is a large literature in the managementfield (Hambrick and Finkelstein, 1987; Tushmanand Romanelli, 1985) concerned with the role oftop management in both inhibiting and facilitatingorganizational change. A particular form of thisargument is that incumbent executives tend tobe committed to the company's current courseof action. Some variants of this argument arepsychological, either associated with psychologi-

cal commitment (Staw, 1981) or cognitive blindersthat restrict the management's awareness ofalternative courses of action (Daft and Weick,1984; Dutton and Dukerich, 1991; Zajac andBazerman, 1991). Other variants revolve aroundissues of organizational politics either internal tothe firm (Boeker, 1989) or external (Zajac andKraatz, 1993). A switch in strategy may changethe importance of particular units of the organiza-tion and, thereby, threaten existing power bases.Alternatively, changes in strategy may disruptexternal constituencies.Figure 1 offers a simple schematic represen-tation of the interplay between organizationalprocesses and industry dynamics in determiningthe firm's 'capability trajectory.' Such an opensystems perspective is clearly not new. However,as suggested earlier, the strategy field has notbeen terribly effective at bridging such levels ofanalysis and perspectives. Subsequent sections ofthe paper apply this general perspective to thecompetition among providers of custody andtransfer agency services in the mutual fundbusiness. In this context, the ability to sustainmarket relations and to acquire new ones is quitedependent on a variety of feedback mechanismsfrom general product market activity, but alsoimportantly a history of relationships betweenintermediaries, inluding relationships betweenprocessors and funds, and also relationshipsbetween processors and both investment man-agers and distributors, and even geographiclocation.

EnvironmentFeedforward

Feedback

Focusing Forcesrganization

Figure1. Coevolutionof firmand industry.


7/19

50 D. Levinthal and J. MyattMUTUAL FUND INDUSTRYBirth of an organizationalormWe explore these ideas of coevolution of capabili-ties and industry in the context of the mutualfund industry. The Massachusetts InvestmentTrust, established in 1924, is usually identifiedas the first mutual fund. However, investmenttrusts, organized in Britain to invest in theU.S.A., had existed for more than 100 years bythe time that the Massachusetts Investment Trustwas started. Massachusetts Investment Trustdifferentiated itself from these prior trusts throughits novel offering structure. Under the traditionaltrust structure, a group of individuals entrustedtheir financial assets to a 'trustee' who investedthe assets as a pool. This type of investmentoccurred through a private arrangement betweenthe investors and the trustee. Unlike these earlytrusts, Massachusetts Investment Trust offeredparticipation in a group of investments to thepublic through an offering of shares. The numberof shares available and the time limit of theoffering were not fixed. This type of fund cameto be known as an open-end fund.3One of the main attractions of mutual fundsto investors is the degree of liquidity they offerin comparison to other investment vehicles. Whatdistinguishes open-end investment companies istheir constant readiness to repurchase or redeemtheir outstanding shares at the current value.Most of these companies continuously offer newshares as well. Another feature of mutual fundsis regular reports concerning the operations andportfolio holdings of their companies and, inparticular, the daily reporting of their net assetvalue (NAV). Due to regulatory requirements,funds are required to publicly disclose their asset,income and expense structure annually. In orderto qualify for investment company status, mutualfunds are also required to distribute substantiallyall net income from dividends, interest andcapital gains directly to investors.

3 In contrast, what become known as closed-end funds arefunds that have a fixed offering of shares. Purchases andredemptions of closed-end funds require a transaction on astock exchange analogous to the purchase or sale of corporatestock.

Mutual fund functionsThe provision of a mutual fund involves four basicfunctions: investment management, distribution,custody and fund accounting, and transfer agencyor shareholder services. While all these activitiesunderlie the provision of a mutual fund, themutual fund itself need not provide any ofthese services. The fund sponsor, however, isresponsible for contracting for these services orproviding them itself. Mutual fund companiesdiffer according to how each function isaccomplished and whether the activity is carriedout by the fund complex itself or some inde-pendent party. Despite the differences in con-figuration, each function must be completed inorder to offer mutual funds to potential investors.Investment management-the portfoliodecisions regarding the fund's investments-isoften carried out by an independent investmentadvisor but more typically is performed by themutual fund sponsor itself. This internal provisionof investment decisions may reflect the fact thatthe rate of return on invested funds is themost basic means by which funds differentiatethemselves. A rather different motivation stemsfrom the fact that the provision of investmentmanagement services is viewed as the mostprofitable activity in the mutual fund business.Therefore, keeping this activity in-house tendsto be in the interest of the management of themutual fund.Funds are offered to investors through twotypes of distribution channels. Investors may buyshares directly from a fund via phone or directmail marketing. They may also buy sharesthrough a sales force. In the latter case, theunderwriters of the fund act as wholesalersand establish sales agreements with variousdistributors. Several different types of distributorsmay be contracted by the underwriters, includingstockbrokers, financial planners, and insuranceagents. Some underwriters employ their owncaptive sales force. The term 'captive' refers tothe fact that the sales force only sells the mutualfunds of that particular underwriter, as opposedto brokers who may sell funds offered by severaldifferent underwriters. As of 1992, 59 percent offunds were sold through a sales force and 41%through direct marketing.The custodianship, or trust and custody func-tion, is responsible for the physical holding of


8/19

Co-evolution of Capabilities and Industry 51securities in which fund assets are invested.Securities regulations require that this functionbe handled by a bank. Related to the custodyactivity is fund accounting and administration.Fund accounting is responsible for the recordkeeping of the fund from an investment, ratherthan shareholder, perspective. This functioninvolves tracking the investment of a fund'sassets. Mutual fund regulations require that fundscalculate a net asset value on a daily basis.Essentially a simple record-keeping function,fund accounting is made more difficult by complexsecurities, including derivatives and junk bondsamong others, whose values are not always easilydetermined. In general, the discipline of same-day reconciling of all security trades, as opposedto the five working days allowed for stocktransactions, poses significant demands on theprocessor. While fund accounting is responsiblefor maintaining the accounting records of thefund, fund administration is responsible forcompliance records. As both areas use the samedata for their records, they are generally treatedas the same function. Fund administration isresponsible for reporting to the SEC and otherregulatory agencies and the fund board of trusteesregarding fund activities.

Transfer agency is the function responsible formaintaining records of shareholder transactions.Accounting books are kept on the shares heldby individuals. The information is then usedto generate shareholder reports. This functionbecomes more complex in fund sponsors thatoffer the same fund under different fee structures,where the differences in fee structures typicallyreflect different modes of distribution. Thetransfer agency function is also responsible forkeeping track of the amount of money investedin the fund, which is important from an investmentmanagement perspective, to ensure that thefund's assets are fully invested. This function isusually highly mechanized, and as a result issubject to significant economies of scale. Inrecent years, this function has become labeled'shareholder services'. This change in labelingreflects an increased emphasis on customerrelations, particularly for fund complexes thatengage in direct sales to end customers. Thesefunds have developed more extensive customersupport in order to compete more effectivelywith funds that offer customers a broker/dealeras an intermediary. Shareholder service

encompasses all activities concerning contactbetween the shareholder and the fund. Activitiesinclude generating and sending shareholderaccount statements and responding to shareholderinquiries. Larger mutual fund complexes oftenhandle this function internally in order to controlthe quality of service provided to its customerand to protect client relationships.Our analysis focuses on those activities thathave historically not been provided by the mutualfunds themselves. The reason for this is thatover the 70-year time span of the mutual fundindustry we can not hope to have measures ofefficiency or productivity with respect to thevarious underlying activities. However, for thoseactivities that are subject to market forces,competitive viability should reflect the firm'scompetence in the given domain. The twoactivities that fall under this category are custodyand transfer agency. Custody is the 'cleanest'activity in this regard since, by law, it must beprovided by an independent firm. In contrast,transfer agency activity has seen a considerabledegree of change over the years in the extent towhich funds provided this service themselves orcontracted externally for the service. Early on inthe mutual fund industry, almost all funds reliedon an external transfer agent. For instance, in1940 89 percent of the assets in the industrywere managed by an external transfer agent. Inthe 1970s, there was a movement towards in-sourcing of transfer agency among the largermutual fund complexes. This movement wasdriven by both a desire to have more controlover their customer relations and a belief on thepart of many funds that they had achievedsufficient scale of operations to be efficient inthe provision of these services. Many funds,however, later reversed themselves and againengaged an outside firm to provide these services.As of 1984, 66 percent of the assets and 57percent of the funds were handled by an outsidefirm.

EVOLUTION OF FUND PROCESSORSInitial conditionsWithin the mutual fund business, there are avariety of initial conditions that influence a firm'sability to compete. Perhaps most important hasbeen the existing base of investment assets. Early


9/19

52 D. Levinthal and J. Myatton in the industry,a large pool of investmenttrust assets provided a basis for the initialcapitalization f a fund. More recent years havewitnessed a similarphenomenonwith respecttoretail banksand certificatesof deposit. As thesecertificatesmature, banks have tried to retaincontrol over the funds by encouraging ustomersto reinvest in a bank-operatedmutual fund.Similarly, brokerage houses have successfullyentered the mutual fund businessby leveragingtheir existingcustomerpool. Mutualfund com-plexes themselvesexhibit this pattern when theydraw investorsto a new fund from their existingaccounts.With regardto processing activities, the fundmanagers lso mayhaveexistingbusinessrelation-ships with the processor.The most direct formof these ongoing relationships s a situation inwhich the fund processor is currently servicingother mutual funds under the control of thesamemanagement.Directors of the mutualfundmay have other business or even nonbusinessties to potential providers of fund-processingservices. Indeed, the very first mutual fund,MassachusettsInvestment Trust, chose StateStreet Bank as its custodian and transferagentdue to existingrelationships etweenmanagementof the two firms(Wilson, 1949).We predictthat funds which are part of fundcomplexes that have existing relationshipswitha given processorwill be more likely to choosethat processor when they are initiatinga newrelationship.Furthermore,as a crude proxyforthe possible importanceof socially embeddedrelationships(Granovetter, 1985), we examinethe importance of geographic location. Co-location of the fund and the processor s weaklysuggestiveof social embeddedness.FeedbackTransferagency and custody operations, parti-cularly in the last two decades, are viewed asscale-intensivebusinesses. In our analysis, weexamine the effect of scale, as measured bymarketshare,on both the probabilityof existingrelationshipscontinuingas well as on the likeli-hood of a fund that is switching processorschoosinga given processor.

Independentof scale effects, one mayobservepositive feedback in the form of experientiallearning.Againwithregard o experienceeffects,

the relevant experience base may be across allfunds or a more narrow category. Experiencemay also be specific to a particular client(Levinthal and Fichman, 1988). Such relationship-specific assets may consist of well-groundedcommunication patterns and the developmentof trust among those individuals involved inboundary-spanning jobs (Adams, 1976), andknowledge of the peculiarities of a firm's account-ing system and products. They may be inthe form of skills embodied in people andorganizational routines, as well as skills embodiedin assets like computer software or special-purpose written procedures (Nelson and Winter,1982). Whatever their form, as a result ofinvestments by both parties to make the relation-ship function effectively and of learning overtime, the two parties to the exchange developrelation-specific assets. The common feature ofthese assets, which is central to our analysis, isthat they can only emerge over time.Therefore, we examine the effect of durationof fund-processor relationships on the probabilityof those relationships continuing. In addition, afund that is engaging in the choice of a newprocessor may be part of a fund complex thathas ongoing relations with a variety of processors.Therefore, we also examine the effect of theduration of these complex level-processorrelationships on the fund's choice of a newprocessor.FitThese forces of positive feedback may at timesbe attenuated, and perhaps even prove dysfunc-tional, as the needs and requirements of clientschange. For instance, if a fund changes its modeof distribution then the experiential base of itsprocessor with regard to the prior mode ofdistribution is no longer relevant. With regardto the custody business, an important source ofrelationship-specific learning may occur betweenthe investment manager and the custodian.Therefore, a switch in investment manager mayeliminate this relationship-specific capital. Weexamine the impact of a change in investmentadvisors on the probability of a fund continuingto engage its current custodian. Similarly, welook at the impact of a change in distributor onthe persistence of the relationship between thetransfer agent and the fund. Moving from


10/19

Co-evolution of Capabilities and Industry 53one distribution mode to another (e.g. captivebroker-dealer to noncaptive or broker-dealer todirect) changes the demands faced by the transferagent; furthermore, even moving from onedistributor to another within the same mode maydisrupt the systems and routines developed bythe transfer agent.Focusing forcesIn the early days of the mutual fund business,the processing of fund accounts was not viewed asa particularly high-growth or profitable business.Therefore, banks that were strong in traditionallines of business may not have aggressivelypursued fund processing. Furthermore, financialinstitutions for which fund processing representsa more substantial portion of their overall businessactivitymay be more likely to aggressively reinvestresources and pursue new opportunities in thisarea.We do not have reliable profitability data oncustodian banks over the life history of theindustry; however, we do have data on the levelof assets of these banks. This information onoverall assets for the banks allows us to constructa measure of the ratio of mutual fund assets forwhich the bank serves as custodian to the bank'sown assets. This measure provides an indicationof the intensity of the bank's involvement in thefund's business relative to their overall activity.Transfer agents may be banks but may,particularly in more recent years, be nonbanksthat specialize in transaction-processing activities.Therefore, the measure of intensity of involve-ment used in the case of custodians can not beapplied. Instead we indicate with a dummyvariable whether the transfer agent is a nonbanktransaction processor or a banking institution. Ina dichotomous way, this measure captures thedegree to which the processor is focused on themutual fund business.Our preceding argument suggests that themeasures of focus should be positively related tothe persistence of ongoing relationships as wellas with the likelihood of a fund choosing toinitiate a relationship with that processor.Table 1 summarizes the categories of variablesto be considered and indicates the particularmeasures to be used as well. In addition to thesemeasures, a few control variables were includedin the analysis. First, some observers have

suggested that relationships between funds andprocessors have become less stable in recentyears. As a result, we include a time perioddummy variable to account for the possibility ofa shift in the longevity of these relationships.The other control variables are related to thesize of the processor and the fund; they are themarket share of the particular mutual fund andthe market share of the mutual fund complex ofwhich the fund is a part. One might believe thata large fund complex would leverage its bargainingpower by playing off service providers on oneanother over time. Alternatively, larger fundcomplexes may pose more complex demands ontheir processors and engender greater opportuni-ties for relation-specific learning (Levinthal andFichman, 1988). The point made here is merelythat it may be worth controlling for such effects.In addition, most measures are expressed inrelational terms (i.e., market share indicators,relative reliance on a given processor) so thatthe measures are not distorted by the dramaticgrowth in the industry over time.

DATAData were collected for bond and equity fundsfrom an annual reference put out by Wiesenbergerstarting in 1934. Wiesenberger provides a varietyof information on each fund, including assetsunder management, information as to the invest-ment manager, transfer agent, and distributor.Since 1965, Wiesenberger explicitly indicates thatit profiles all mutual funds with at least $10million in assets; the earlier reporting of mutualfund activity appears consistent with this criterion.In cases where the data from Wiesenberger wereincomplete, the appropriate years of Moody'sBank and Finance Manual were used. Data onmoney market funds were collected from theIDD Mutual Fund Directory. The IDD FundDirectory attempts to include all money marketfunds.From these data on the individual funds, weconstructed measures of assets under manage-ment for the various transfer agents and custo-dians in operation during the sample period.Duration data as well were constructed byexamining breaks in relationships with a givenprocessor. All breaks in relationships wereexamined to make sure that there was indeed a


11/19

54 D. Levinthal and J. MyattTable 1. Measurement of feedback and focus effectsFeedbackGeneral expertise*Overall market share in the processing businessRelationship-specific expertise*Years in the relationship*Proportion of mutual fund complex assets served by processor*Proportion of mutual fund complex funds served by processor*Common geographic settingDisruptive forcesPrior relationship expertise is negated by changes in the structure of the relationship*Change in investment advisor (custody relationships)*Change in distributor (transfer agent relationships)Focus effectsIndicators of the importance of the fund-processing business to the focal firm*Assets under management relative to total assets (custody business)*Dummy variable indicating a specialist processor organization vs. a bank (transfer agent relationships)

change in the relationship and not a change inownership or name in one of the parties. Datawere collected to form a complete census ofthe population as characterized by the abovepublications at 2-year intervals from 1934 to1984. While the industry came into being in 1924with the first open-ended mutual fund, reliableinformation on service providers was not availableuntil 1934.

EMPIRICAL ANALYSISThe empirical analysis consists of two basiccomponents. One is a multinomial logit analysisof the initiation of relationships with a processorand the other is a logit analysis of the probabilityof an ongoing relationship continuing from oneperiod to the next. The significance of a variableor set of variables is assessed with the likelihoodratio test.4 The test compares the goodness offit of a pair of nested models, with the variableor variables of interest being the constraints thatdistinguish the models (Fienberg, 1980). Themeasure of goodness of fit is the G2 statistic,minus two times the value of the log likelihood.4 The likelihood ratio test is preferable to the t-test for assessingthe effect of a single variable since tests based on t-values ofindividual coefficients can be misleading in logit analysis(Anderson et al., 1980).

Under the null hypothesis, this difference isasymptotically distributed as a chi-squared(Fienberg, 1980). The degrees of freedom forthe likelihood ratio test are equal to the numberof parameters that distinguish the two models.The effect of individual variables within each setwas assessed by comparing a model excludingthe variable of interest with a model containingthe full set of variables.

LOGIT ANALYSIS OF RELATIONSHIPPERSISTENCEThe logit analysis examines the probability that,from one period to the next, a given relationshipwill end. Therefore, positive values for anestimated coefficient indicate that for that parti-cular variable an increase in the value of thevariable will tend to increase the risk of therelationship ending. Similarly, a negative esti-mated coefficient indicates that larger values ofthe variable will tend to be associated with alower risk of the relationship ending. Table 2gives the results of the logit analysis of custodianrelationships, while Table 3 provides the resultsfor transfer agent relationships.

The control variables as a set, in contrast to amodel consisting of just a constant term, werehighly significant (X2 = 11.982, d.f. = 3,


12/19

Co-evolution of Capabilities and Industry 55p < 0.01) in the analysis of switching of transferagents but did not contribute significantly to theestimation of persistence of custodian relation-ships (X2 = 3.197, d.f. = 3, p > 0.10). Withinthe set of control variables for the analysis oftransfer agents, the variables for the effect ofmutual fund market share (X2= 6.220, d.f. = 1,p < 0.05) and for the effect of the market shareof the mutual fund complex (X2 = 5.995, d.f. = 1,p < 0.05) were individually significant.This lack of significance for the controlvariables in the analysis of custodian relationshipsholds in spite of the fact that there were moreobservations in the analysis of custodians, 3257cases vs. 1906 cases for the analysis of transferagents, and a relatively similar total frequencyof switching events in the two settings. Out ofthe 3257 observations of custodian-mutual fundrelations, there were 591 switches; similarly, forthe 1906 observations of transfer agent-mutualfund relations there were 317 switches. Thereason for the greater magnitude of data oncustodian-mutual fund relationships is that manymutual fund complexes carry out the transferagent responsibilities. As our analysis is restrictedto market relationships, we are left with asomewhat smaller sample of transfer agent-fundpairs. Along these same lines, it is important tonote that we are examining the likelihood of anexternal transfer agent-mutual fund relationshipending in a switch to a new external transferagent. Other terminating events, such as theclosure of the fund or, more commonly, thedecision by the fund complex to internalizetransfer agent activities, are considering a cen-soring event and not a switch.Product market activity is strongly self-reinforc-ing. Processors, whether they are custodians ortransfer agents, have a substantially higherprobability of continuing their existing customerrelationships the larger is their overall marketshare of assets under management. For thecase of custodians, the estimate is x2 = 8.276,d.f. = 1, p < 0.05 and for transfer agentsX2 = 12.843, d.f. = 1, andp < 0.001.Not only does greater current product marketactivity tend to lead to greater continuity ofexisting relationships, but the subsequent multi-nomial analysis of the choices made by fundsinitiating a new relationship indicates that greatercurrent market activity has a significant impacton the likelihood of gaining new customers.

Obviously, the archival record of business activityin the industry does not give a direction insightinto the source of those scale advantages. Whilethere is clearly a sense in the industry today thatscale economies are important, much of thatdiscussion is focused on the distribution of mutualfunds and not their processing (Alexander et al.,1994; Makadok, 1992). The major contributorto scale economies in the processing sector ofthe industry would seem to lie in the large,primarily fixed, costs of software and systemsdevelopment.Much of the self-reinforcing feedback stemsfrom particular market relationships. Greaterduration of past relations and more extensiveties, either as measured by assets or number offunds within the same mutual fund complex,lead to greater likelihood of a given relationshipcontinuing. In addition, geographic proximity, asmeasured by the fund and the processor beinglocated in a common state, may facilitate thedevelopment of a broader set of ties betweenthe two organizations. In contrast, ongoingrelationships are threatened if those in importantboundary-spanning roles depart (Seabright, Lev-inthal, and Fichman, 1992). Overall, theserelational variables are highly significant(X2 = 108.639, d.f. = 5, p < 0.001 for custodyrelationships and x2 = 85.69, d.f. = 5, p < 0.001for transfer agent relationships).With the exception of the measures of theintensity of the current relationship in the caseof transfer agents, the various measures ofrelational feedback are individually significant aswell. In the case of transfer agents, both theratio of the assets of the mutual fund complexthat are managed by the processor (X2 = 2.703,d.f. = 1, p > 0.10) and the ratio of the numberof funds in the overall mutual fund complexmanaged by the processor (X2 = 1.422, d.f. = 1,p > 0.10) fall a bit short of statistical significance.The measures of the intensity of ongoing relation-ships are significant in the case of custodyrelationships for both the proportion of funds ofthe overall mutual fund complex managed bythe processor (X2 = 16.591, d.f. = 1, p < 0.001)and for the proportion of the assets of themutual fund complex managed by the processor(X2 = 3.115, d.f. = 1, p < 0.10).The duration of the relationship between thefund and processor was individually significantin the case of custody relationships (X2 = 9.006,


13/19

56 D. Levinthal and J. MyattTable 2. Logistic analysis of custodian relationships

FeedbackVariables Constant Control Market Relational FocusConstant -1.507 -1.544 -1.481 -0.023 -0.163(0.046) (0.082) (0.084) (0.218) (0.230)Market share 0.032 0.029 -0.036 0.010of fund (0.031) (0.034) (0.060) (0.061)Market share -0.015 -0.011 -0.018 0.003of complex (0.017) (0.017) (0.022) (0.021)1974-1984 period 0.116 0.168 -0.217 -0.218dummy (0.096) (0.098) (0.122) (0.133)Custodian market -0.012 -0.005 -0.003share (0.004) (0.005) (0.006)Proportion of assets 0.007 0.010with custodian (0.004) (0.004)Proportion of funds -0.018 -0.021with custodian (0.004) (0.005)Duration of fund/custodian -0.034 -0.026relationship (0.012) (0.013)Common state -0.684 -0.581(0.111) (0.120)Switch investment 0.858 0.736advisor (0.149) (0.163)Ratio of custody assets 0.003to bank assets (0.004)

Observations 3257 3257 3253 2369 2069Parameters 1 4 5 10 11G 3084.986 3081.789a 3063.343b 2181.161c 1877.521Difference in G2 3.197 8.276 108.639 0.455p>O. 10 p


14/19

Co-evolution of Capabilities and Industry 57Table 3. Logistic analysis of transfer agent relationships

FeedbackVariables Constant Control Market Relational FocusConstant -1.612 -1.618 -1.467 -0.736 -0.675(0.062) (0.107) (0.114) (0.327) (0.331)Market share 0.092 0.134 0.113 0.116of fund (0.036) (0.038) (0.049) (0.049)Market share -0.061 -0.061 -0.076 -0.079of complex (0.027) (0.027) (0.038) (0.038)1974-1984 period 0.201 0.264 -0.166 -0.125dummy (0.131) (0.133) (0.171) (0.174)Transfer agent overall -0.038 -0.009 -0.012market share (0.011) (0.013) (0.013)Proportion of assets -0.005 -0.005with custodian (0.003) (0.003)Proportion of funds 0.004 0.003with custodian (0.003) (0.003)Duration of fund transfer -0.017 -0.018agent relationship (0.017) (0.017)Common state -0.802 -0.837(0.165) (0.167)Switch distributor 0.412 0.410(0.206) (0.207)Nonbank -0.345(0.237)Observations 1906 1906 1906 1327 1327Parameters 1 4 5 10 11G2 1715.392 1703.409 1690.566a 1155.789 1153.579Difference in G2 11.982 12.843 85.69 2.21p


15/19

58 D. Levinthal and J. Myattwith that processor, the processor is not in thesame state, and the processor did not previouslyserve the client in the prior period. For custodyrelationships, the probability of the high relationalexchange ending is 0.18 while the probability ofthe low relational exchange ending is 0.47. Thatis, there is nearly a 50/50 chance that a lowrelational exchange will end over a 2-year period.Similarly, for transfer agent relationships, theprobability of a high relational exchange endingover a 2-year period is 0.12 while the probabilityof a low relational exchange ending is 0.30. Incontrast to the relatively modest impact of marketshare on the persistence of relationships, thevariation in the intensity of the degree ofrelational feedback yields impacts on the prob-ability of persistence by a factor of two or more.Both the switch in advisor in the case ofcustody relationships (X2 = 30.946, p < 0.001)and the switch in distributor (X2 = 3.802,p < 0.10) were individually significant. Thesechanges served to disrupt the funds relationshipswith each external processor, even though themutual fund, not the investment advisor ordistributor, is responsible for contracting forcustody and transfer agent activities. The factthat the change in the identity of these boundaryspanning agents reduces the likelihood of thefund-processor relationship continuing is sugges-tive of the relational nature of these exchanges.The measure of focus was not associated withthe likelihood of the processor relationshipcontinuing in either the case of custodians(X2 = 0.455, d.f. = 1, p > 0.10) or for transferagents (X2 = 2.21, d.f. = 1, p > 0.10). Theability to retain clients, while tied to feedbackeffects at the market level and, more importantly,to feedback effects within ongoing relationships,is not influenced by the prominence of themutual fund business to the processor.

MULTINOMIAL LOGIT ANALYSIS OFPROCESSOR CHOICEThese same constructs were examined in thecontext of a mutual fund's choice of processorwhen it is initiating a new processor relationship.The exact same set of variables is not, however,used. First, any variable that does not vary overthe alternative set can not be included in amultinomial ogit. This clearlyincludes both the

time period dummy variable and those variableswhich are attributes of the mutual funds them-selves (i.e., the market share of the fund andthe market share of the fund complex). Inaddition, the relational tie between the processorand mutual fund will vary by processor-funddyads. Since a given mutual fund complex isgoing to have ongoing relationships with only asmall subset of the processors that are active ina given year, this was treated as a dichotomousvariable indicating the presence or absence ofsuch a relationship and not as a measure ofintensity of the relationship (i.e., proportion ofassets and proportion of funds within a complex)as in the case of the logit analysis of thepersistence of ongoing relationships. Similarly,each fund-potential processor pairing is codedas sharing or not sharing the same geographiclocation. The remaining variable, the processor'smarket share, is obviously an attribute of theprocessor and is independent of the particularpotential client mutual fund. Table 4 providesthe multinomial analysis of the choice of custod-ian, while Table 5 gives the results of themultinomial analysis of the choice of transferagent.Relative to a model without covariates, theeffect of market share is highly significant in boththe case of custody relationships (X2 = 286.916,d.f. = 1, p < 0.001) and for transfer agentrelationships (X2 = 56.041, d.f. = 1, p < 0.001).Furthermore, examining a model in which amarket share square term is included indicatesthat higher market share results in a higherprobability of engaging a new client over therange of empirically observed market sharevalues.5The three measures of relationship ties collec-tively have a significant positive effect on thelikelihood of a mutual fund choosing a givenprocessor (X2 = 211.19, d.f. = 3, p < 0.001 forchoice of transfer agent and x2 = 880.394,d.f. = 3, p < 0.001 for custodians). In addition,the three measures are individually significant:(X2 = 143.472, p < 0.001 custodian;X2= 181.708, p < 0.001 transfer agent) for theeffect of a common state location, (X2= 22.432,

5 The exception to this finding is that there are a few instancesin which the maximum value of market share in a given yearexceeds the point at which the probability of engaging a newclient reaches its maximum value with respect to market share.


16/19

Co-evolution of Capabilities and Industry 59Table 4. Multinomial analysis of custodian choice

FeedbackMarket Relationship Focus

Market share of 0.078 0.061 0.062custodian (0.004) (0.005) (0.006)Existing custodian! 1.387 1.337fund complex tie (0.164) (0.184)Duration of custodian! 0.234 0.259fund complex relationship (0.051) (0.058)Common state 1.624 1.600(0.116) (0.128)Ratio of custody assets 0.004to bank assets (0.003)Observations 572 525 386Parameters 1 4 5G2 4089.870a 2836.320 2182.736Difference in G2 880.394 1.167pO. 10aThe value of G2 for N of 525 is 3716.714bThe value of G2for N of 386 is 2183.903

Table 5. Multinomial analysis of transfer agent choiceFeedback

Market Relationship FocusOverall market share 0.072 0.059 0.060of transfer agent (0.009) (0.010) (0.010)Existing transfer agent/ 1.032 1.048fund complex ties (0.219) (0.219)Duration of transfer agent/ 0.064 0.047fund complex relationship (0.043) (0.043)Common state 1.732 1.742(0.141) (0.142)Nonbank 0.246

(0.176)Observations 302 276 276Parameters 1 4 5G2 1970.267a 1583.828 1581.942Difference in G2 211.19 1.886pO. 10aThe marketmodel has a G2 value of 1795.015 or 276 cases


17/19

60 D. Levinthal and J. Myattp < 0.001 custody;x2= 17.821,p < 0.001 transferagent) for the effect of duration of relationshipbetween the overall mutual fund complex and theprocessor, and (X2 548.242, p < 0.001 custody;X2 = 19.071, p < 0.001 transfer agent) for thepresence of an existing relationship between themutual fund complex and the processor. Whilethe focus variables have the predicted sign forboth custody and transfer agent relationships,in neither instance is the estimated coefficientstatistically significant. For custody relationshipsthe x2 test yields (X2= 1.167, p > 0.10) and inthe case of transfer agent relationships it yields(X2= 1.886, p > 0.10).To get some idea of the magnitude of theseeffects, consider the probabilitythat a given mutualfund would choose a particular processor, whereprocessors vary according to their value withrespect to some subset of the independent variablesin the multinomial analysis and share a commonvalue for the remaining variables, where thesecommon values are set at the mean relationships,market share has an enormous impact on theprobability that a given fund will initiate a newrelationship with a particular processor. In thecontext of custodian relationships, a custodian witha 5 percent market share has a 0.03 probability ofbeing chosen by an arbitrary mutual fund, whilea custodian with a 25 percent market share has a0.08 probability of being chosen. Similarly, atransfer agent with a market share of 5 percenthas a 0.04 probability of being chosen by a mutualfund in search of a new service provider, while atransfer agent with a 25 percent market share hasa probabilityof 0.12 being chosen. Thus, the effectof variation in market share yields a difference oftwo or three orders of magnitude in the likelihoodof a relationship being initiated.

The impact of existing relationships betweenthe processor and potential client is even morepronounced. If the mutual fund is part of a fundcomplex that has an existing relationshipwith theprocessor, the likelihood of a relationship beinginitiated changes by several orders of magnitude,changingfrom a probabilityvalue of 0.03 to 0.10 inthe case of transfer agent relationshipsand from0.02 to 0.08 in the case of custody relationships.Surprisingly, he mere fact of being located in thesame state has an even larger impact on thelikelihood of choosing a processor. For the case ofcustodians,the probabilitychangesfrom 0.01 to 0.09and for transferagents it changes from 0.01 to 0.13

One might think that this enormous impact ofthe geographic variable is an artifact of the heavyconcentration of financial service activity in afew states. However, this effect is only triviallydiminished when a control variable is introducedinto the estimating equation that indicates whatshare of the total processing activity is carried outin the processor's state. The coefficient on thevariable indicating a common state changes from1.732 to 1.710 in the case of transfer agents andfrom 1.624 to 1.618 in the case of custodyrelationships.

CONCLUSIONHeterogeneity in competitive position among firmsin the transfer and custody business for mutualfunds is influenced by a variety of forces. Somefactors, such as market share, amplify the existingheterogeneity. Firms with a strong market positiontend to have a higher probability of sustainingongoing relationships and, much more pronounced,a greater tendency to initiate new relationships.Other factors tend to sustain over time theexisting position of firms. Existing relationshipsare fairly stable. The more extensive the tiesbetween the fund complex and the processorthe greater the stability of these relationships.Furthermore, the powerful effect of the geographicvariable suggests that these ties do not just consistof narrow task-specific knowledge, but may reflectto some extent a more socially embedded relation-ship (Granovetter, 1985).Some of these relational factors, such as theeffect of duration and intensityof ongoing relations,tend to sustain existing competitive positions.However, other relational effects will tend toamplifythe existing heterogeneity in the population.Processors that have an existing tie with a mutualfund complex are much more likely to acquireadditional funds within that complex as clients. Inaddition, the analysis is strongly supportive ofrecent discussion of geographic agglomerationeffects (Arthur, 1986; Krugman, 1991).The lack of significance of the focus variablessuggests that the evolution of competitive position isdrivenby marketposition and externalrelationshipsand that organizationalcommitments and resourceallocations do not exert an independent effect onthis trajectory. Clearly, however, the current dataset is primarily based on measures of market


18/19

Co-evolution of Capabilities and Industry 61activity and only in a crude way reflects somesense of internal organizational attributes.Taken as a whole, the empirical evidence offersrather strong support for the arguments regardingamplification of heterogeneity by market activityand the inertial effect on heterogeneity of therelational nature of the exchange relationships.The relative role of these two sorts of forces,however, is clearly context dependent. Financialinstitutions, by definition, serve an intermediaryrole and, as a result, client-firm exchanges arelikely to have a significant relational component.Contexts will also differ in the degree to whichthe positive feedback of market activity is enduringover time. The relationships studied here weresubject to disruptive forces of changes in theaffiliated intermediaries but were not subject tomore radical changes in the basis of competitiveadvantage. Nevertheless, the qualitative distinctionbetween these two sorts of feedback mechanisms,one amplifyingheterogeneity and the other sustain-ing the current level of heterogeneity, is likely tobe robust. Heterogeneity in competitive positionis sustained by existing market relations and tendsto be amplified by overall market position. Inthese two important respects, the evolution of agiven firm's competitive position can be seen asan outcome of linkages to particular clients andthe overall market.

ACKNOWLEDGEMENTSThis research has been supported by a SloanFoundation grant to the Financial InstitutionsCenter at the Wharton School and the Sol C.SniderEntrepreneurialCenter, also at the WhartonSchool. We gratefullyacknowledge the time severalparticipants in the industry took to enrich ourunderstanding of the industry. In particular, wewould like to acknowledge Barbara Weidlich ofState Street Bank and Dushyant Pandit of PNCFinancial Services. We thank Ed Zajac forcomments on a prior draft.

REFERENCESAaker D. A. (1991). Managing Brand Equity. FreePress, New York.Abernathy, W. J. and K. B. Clark (1985). 'Innovation:Mapping the winds of creative destruction', ResearchPolicy, 14, pp. 3-22.

Adams S. K. (1976). 'The structure and dynamics ofbehavior in organizational boundary roles'. In M.Dawned (ed.), Handbook of Organizational andIndustrial Psychology. Rand McNally, Chicago, IL,pp. 1175-1199.Alexander, J., M. Hurley, S. Meers, Y. Aliredha andE. Wang (1994). 'The changing economics of themutual fund industry', Industry Research Group-Goldman Sachs, New York.Anderson, S., A. Auquier, W. Hauck, D. Oakes, W.Vandaele and H. Weisberg (1980). StatisticalMethodsfor ComparativeStudies. Wiley, New York.Andrews, K. R. (1971). The Concept of CorporateStrategy. Richard D. Irwin, Homewood, IL.Argyris, C. and D. Schon (1978). OrganizationalLearning. Addison-Wesley, Reading, MA.Arthur, W. B. (1986). 'Industry location patterns andthe importance of history', Center for EconomicPolicy, Stanford University, Stanford, CA.

Arthur, W. B. (1989). 'Competingtechnology, increasingreturns, and lock-in by historical events', EconomicJournal, 97, pp. 116-131.Arthur, W. B. (1990). 'Positive feedbacks in theeconomy', ScientificAmerican, pp. 92-99.Barney, J. B. (1991). 'Firm resources and sustainedcompetitive advantage', Journal of Management, 17,pp. 99-120.Boeker, W. (1989). 'The development andinstitutionaliz-ation of subunit power in organizations',Administrat-ive Science Quarterly,34, pp. 388-410.Boston ConsultingGroup (1972). Perspectiveson Experi-ence. Boston Consulting Group, Inc., Boston, MA.Bresnahan, T. F. (1987). 'Competition and collusion inthe American automobile industry: The 1955 pricewar', Journal of Industrial Economics, 35,pp. 457-482.Carey, S. (July 1, 1993). 'Donnelley follows its customersaround the world', Wall StreetJournal, p. Bi.Christensen, C. M. and J. L. Bower (1994). 'Catchingthe new wave: Why good customers make ithard'. Unpublished manuscript, Graduate School ofBusiness, Harvard University.Clark, K. B. and T. Fujimoto (1991). Product Develop-ment Performance:Strategy,Organization, and Man-agement n the WorldAuto Industry.HarvardBusinessSchool Press, Boston, MA.Cohen, W. M. and S. Klepper (1992). 'The anatomy ofindustry R & D intensity distributions', AmericanEconomic Review, 82, pp. 773-799.Cohen, W. and D. Levinthal (1989). 'Innovation andlearning: The two faces of R & D', EconomicJournal, 99, pp. 569-596.Cohen, W. and D. Levinthal (1990). 'Absorptivecapacity: A new perspective on learning and inno-vation', Administrative Science Quarterly, 35, pp.128-152.Daft, R. and K. Weick (1984). 'Toward a model oforganizations and interpretation systems', Academyof ManagementReview, 9, pp. 284-296.Dutton, J. E. and J. M. Dukerich (1991). 'Keeping aneye on the mirror:Image and identityin organizationaladaptation', Academy of Management Journal, 34,pp. 517-554.


19/19

62 D. Levinthaland J. MyattEnright, M. (1993). 'Emerging of technologies: Knowl-edge-intensive technologies and regional develop-ment', Business History Review, 67, pp. 347-349.Fichman, M. and D. Levinthal (1991). 'Honeymoonsand the liability of adolescence: A new perspectiveon duration dependence in social and organizational

relationships', Academy of Management Review, 16,pp. 442-460.Fienberg, S. E. (1980). The Analysis of Cross-ClassifiedCategoricalData. MIT Press, Cambridge, MA.Foss, N. J., C. Knudsen and C. A. Montgomery (1993).'Towards a synthesis of evolutionary and resource-based approaches to strategy: Some topics fordiscussion'.Unpublished working paper, CopenhagenBusiness School, University of Copenhagen.Ghemawat, P. (September-October 1986). 'Sustainableadvantage', Harvard Business Review, pp. 53-58.Granovetter, M. (1985). 'Economic action and socialstructure: The problem of embeddedness', AmericanJournal of Sociology, 78, pp. 1360-1380.Hambrick, D. and S. Finkelstein (1987). 'Managerialdiscretion: A bridge between polar views of organiza-tional outcomes'. In L. L. Cummings and B.M. Staw (eds.), New Directions in OrganizationalBehavior. JAI Press, Greenwich, CT.Krugman, P. (1991). Geography and Trade. MIT Press,Cambridge, MA.Levinthal, D. A. (1992). 'Surviving Schumpeterianenvironments; An evolutionary perspective', Indus-trial and Corporate Change, 1, pp. 427-443.Levinthal, D. A. (1995, forthcoming). 'Strategicmanage-ment and the exploration of diversity. In N. J.Foss, C. Knudsen and C. A. Montgomery (eds.),Evolutionary and Resource-Based Approaches toStrategy.Kluwer Academic Publishers, Orwell, MA.Levinthal, D. A. and M. Fichman (1988). 'Dynamicsof interorganizational attachments: Auditor-clientrelations', Administrative Science Quarterly, 33,pp. 345-369.Levinthal, D. A. and J. G. March (1993). 'The myopiaof learning', Strategic Management Journal, WinterSpecial Issue, 14, pp. 95-112.Makadok, R. (1992). 'The "winner'scompetitive cycle"in the money market mutual fund industry: Thedynamics of the competitive engine'. Unpublishedmanuscript, Wharton School, University of Pennsyl-vania.Marshall,A. (1920). Principlesof Economics. Macmillan,London.Mitchell, W. (1989). 'Whether and when? Probabilityand timing of incumbents' entry into emergingsubfields', Administrative Science Quarterly, 34,pp. 208-230.Moody's Bank and Finance Manual. Moody's Publishing,New York.Mutual Fund Directory. Investment Dealers' Digest,New York.Nelson, R. and S. Winter (1982). An EvolutionaryTheory of Economic Change. Harvard UniversityPress, Cambridge, MA.

Porter, M. E. (1980). Competitive Strategy: Techniquesfor Analyzing Industriesand Competitors. Free Press,New York.Porter, M. E. (May-June 1987). 'From competitiveadvantage to corporate strategy', Harvard BusinessReview, pp. 43-59.Porter, M. E. (1990). The Competitive Advantage ofNations. Free Press, New York.Rumelt, R. P. (1974). Strategy, Structure,and EconomicPerformance. Harvard Business School Press, Cam-bridge, MA.Rumelt, R. P. (1984). 'Towards a strategic theory ofthe firm'. In R. B. Lamb (ed.), Competitive StrategicManagement. Prentice-Hall, Englewood Cliffs, NJ,pp. 566-570.Rumelt, R. P. (1991). 'How much does industry matter',Strategic Management Journal, 12(3), pp. 167-185.Seabright, M., D. Levinthal and M. Fichman (1992).'Role of individual attachments in the dissolution

of interorganizational relationships', Academy ofManagement Journal, 35, pp. 122-160.Selznick, P. (1957). Leadership and Administration.Harper & Row, New York.Staw, B. M. (1981). 'The escalation of commitment: Areview and analysis', Academy of ManagementReview, 6, pp. 577-587.Sutton, J. (1991). Sunk Costs and Market Structure.MIT Press, Cambridge, MA.Teece, D. (1986). 'Profiting from technological inno-vation: Implications for integration, collaboration,licensing, and public policy', Research Policy, 15,pp. 285-305.Teece, D., G. Pisano and A. Shuen (1990). 'Firmcapabilities, resources, and the concept of strategy'.Consortium on Competitiveness & CooperationWorking Paper No. 90-8, University of California atBerkeley, CA.Tushman, M. and E. Romanelli (1985). 'Organizationalevolution: A metamorphosis model of convergenceand reorientation'. In L. Cummings and B. Staw(eds.), Research in Organizational Behavior, 7,pp. 171-222.Von Hippel, E. (1988). The Sources of Innovation.Oxford University Press, New York.Wiesenberger, A. InvestmentCompanies. Wiesenberger,New York.Wilson, J. A. (1949). A Story of Progress. MassachusettsInvestment Trust, Boston, MA.Womack, J. P., D. T. Jones and D. Roos (1990). TheMachine that Changed the World. Harper Collins,New York.Zajac, E. and M. Bazerman (1991). 'Blind spots inindustry and competitor analysis: Impact of infor-mation (mis)perception for strategic decisions', Acad-emy of Management Review, 16, pp. 37-56.Zajac, E. and M. Kraatz (1993). 'A diametric forcesmodel of strategic change: Assessing the antecedentsand consequences of restructuring in the highereducation industry. Strategic Management Journal,Summer Special Issue, 14, pp. 83-102.

Co-Evolution of Capabilities and Industry the Evolution of Mutual Fund Processing

Documents

Transcript of Co-Evolution of Capabilities and Industry the Evolution of Mutual Fund Processing