Balancing vertical integration and strategic outsourcing ... · BALANCING VERTICAL INTEGRATION AND ... forward integrated but rely on outsiders for a por- ... Balancing Vertical Integration

Strategic Management JournalStrat. Mgmt. J., 27: 1033–1056 (2006)

Published online 13 September 2006 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/smj.559

Received 2 November 2001; Final revision received 4 April 2006

BALANCING VERTICAL INTEGRATION ANDSTRATEGIC OUTSOURCING: EFFECTS ON PRODUCTPORTFOLIO, PRODUCT SUCCESS, AND FIRMPERFORMANCE

FRANK T. ROTHAERMEL,1* MICHAEL A. HITT2 and LLOYD A. JOBE3

1 College of Management, Georgia Institute of Technology, Atlanta, Georgia, U.S.A.2 Mays Business School, Texas A&M University, College Station, Texas, U.S.A.3 VizX Labs, LLC, San Diego, California, U.S.A.

Most prior research has focused on vertical integration or strategic outsourcing in isolationto examine their effects on important performance outcomes. In contrast, we focus on thesimultaneous pursuit of vertical integration and strategic outsourcing. Our baseline propositionis that balancing vertical integration and strategic outsourcing in the pursuit of taper integrationenriches a firm’s product portfolio and product success, and in turn contributes to competitiveadvantage and thus to overall firm performance. We derive a set of detailed hypotheses, and testthem on a unique and fine-grained panel of longitudinal data documenting over 3,500 productintroductions in the global microcomputer industry. The results provide strong support for thenotion that carefully balancing vertical integration and strategic outsourcing when organizingfor innovation helps firms to achieve superior performance. Copyright 2006 John Wiley &Sons, Ltd.

A long and venerable tradition of cross-disciplinaryscholarship is concerned with the boundaries ofthe firm (e.g., Coase, 1937; Thompson, 1967;Williamson, 1975; Galbraith, 1977). Determiningthe boundaries of the firm appears to be critical forfirm performance, especially in high-technologyindustries (Teece, 1986, 1992; Bettis and Hitt,1995; Hill and Rothaermel, 2003). Most of theextant research, however, examines the trade-offsbetween internalizing activities vs. sourcing themexternally through market transactions (Walkerand Weber, 1984; Jones and Hill, 1988; Leiblein,Reuer, and Dalsace, 2002) or through strategicalliances (Pisano, 1990; Folta, 1998; Steensma and

Keywords: boundaries of the firm; vertical integration;strategic outsourcing; innovation; dynamic capabilities∗ Correspondence to: Frank T. Rothaermel, College of Manage-ment, Georgia Institute of Technology, 800 West PeachtreeStreet, NW, Atlanta, GA 30308-1149, U.S.A.E-mail: [email protected]

Corley, 2001). While firms often trade off econo-mizing on transaction costs vs. access to dispersedknowledge stocks and enhanced flexibility in mak-ing these important governance decisions, manyfirms are partially integrated and simultaneouslyoutsource some activities (Harrigan, 1984; Afuah,2001). We argue that these firms seek to identifythe most effective balance in both organizing alter-natives to leverage their benefits and mitigate theircosts.

Following Harrigan’s theoretical contribution,we label this organizing approach taper integra-tion, which occurs ‘when firms are backward orforward integrated but rely on outsiders for a por-tion of their supplies or distribution’ (Harrigan1984: 643). Thus, taper integration arises whena firm sources inputs externally from independentsuppliers as well as internally within the bound-aries of the firm, or disposes of its outputs throughindependent outlets in addition to company-owneddistribution channels. Taper integration implies

Copyright 2006 John Wiley & Sons, Ltd.

1034 F. T. Rothaermel, M. A. Hitt and L. A. Jobe

that some activities are pursued in a parallel man-ner, both in-house and through outsourcing. Weargue that taper integration enhances a firm’s prod-uct portfolio, new product success, and firm per-formance, in particular, if a firm effectively bal-ances the two strategic components of taper inte-gration, i.e., vertical integration and strategic out-sourcing.

The benefits of pursuing either vertical integra-tion or strategic alliances have been highlighted innumerous prior studies (e.g., Hill and Hoskisson,1987; Jones and Hill, 1988; Kogut, 1988; Hage-doorn, 1993; Dyer, 1996; Dyer and Singh, 1998;Gulati, 1998; Stuart, 2000; Rothaermel, 2001; Ire-land, Hitt, and Vaidyanath, 2002). In contrast,while the use of taper integration has been grow-ing in prominence in a number of industries, littleempirical research has examined the effects of thispractice on firm outcomes. The dearth of empir-ical research is likely due to the methodologicaldifficulty of capturing taper integration in a theo-retically proximal fashion.

We undertake herein a first step towards closingthe gap in our knowledge of the effects of taperintegration. Our overarching hypothesis is that bal-ancing vertical integration and strategic outsourc-ing in a prudent manner helps to optimize a firm’sproduct portfolio and achieve product success. Inso doing, it contributes to a firm’s competitiveadvantage and thereby increases firm performance.To test this hypothesis, we examine the joint effectsof simultaneously pursuing vertical integration andstrategic outsourcing on a firm’s product portfolio,new product success, and firm performance.

First, we suggest that vertical integration andstrategic outsourcing interact to synergisticallyincrease a firm’s product portfolio. Building a port-folio of related products can contribute to a com-petitive advantage, particularly in highly dynamicmarkets (Brown and Eisenhardt, 1997). Next, weassess the individual effects of vertical integrationand of strategic outsourcing on a firm’s productportfolio, product success, and performance. Here,we suggest that these relationships resemble aninverted U-shape, reflecting the theoretical notionthat vertical integration and strategic outsourcingshould be balanced to achieve the desired firm-level outcomes.

To the extent that the products in the portfolioare at the cutting edge of technology and servecustomer needs, they are likely to enjoy successin the marketplace. It is unlikely, however, that

all products in a firm’s portfolio will be success-ful, especially when there are a large number ofthem (Sanderson and Uzumeri, 1995). Firms musthave the capability to manage a large number ofproducts to target them for the appropriate mar-ket segments and to ensure their differentiationas perceived by the consumers. The difficulty ofsuccessfully accomplishing these tasks increasesas the number of related products in the firm’sportfolio grows, thereby enhancing the manage-rial challenges. This effect is likely to be particu-larly pronounced when firms face a high level ofuncertainty, common in high-technology industries(Bettis and Hitt, 1995). We propose a synergisticinteraction effect of vertical integration and strate-gic outsourcing on the size of a firm’s productportfolio, while also suggesting that a moderateproduct portfolio size will have the strongest effecton subsequent new product success and firm per-formance.

The theoretical model presented herein makesa contribution by highlighting the performance-enhancing consequences of balancing two differ-ent organizational forms, vertical integration andstrategic outsourcing, constituting taper integra-tion. We draw on a longitudinal sample of firmsin the global microcomputer industry to test thehypothesized relationships on an original paneldataset. In this research, we also make a method-ological contribution by developing a unique andtheoretically proximate measure of individual newproduct success, derived from a detailed analysisof over 3,500 product introductions. We concludethat successfully organizing for innovation can beaccomplished through taper integration that bal-ances the costs and benefits inherent in this hybridorganizational form to create product portfoliosthat enhance new product success and overall firmperformance.

THEORY AND HYPOTHESESDEVELOPMENT

Taper integration and product portfolios

Firms must continuously develop and introducenew products to the marketplace to achieve andmaintain a competitive advantage. New productintroductions have the potential to create transi-tory advantages for firms, because they enable thefirms to capture returns to innovation (Schumpeter,

Copyright 2006 John Wiley & Sons, Ltd. Strat. Mgmt. J., 27: 1033–1056 (2006)DOI: 10.1002/smj

Balancing Vertical Integration and Strategic Outsourcing 1035

1942). These advantages, however, are frequentlyshort lived in dynamic industries (Bettis and Hitt,1995; Bayus, Erickson, and Jacobson, 2003). As aresult, firms often build portfolios of related prod-ucts to deter entry through product proliferation,increasing their customer base, better serving cus-tomer needs, and maintaining strategic flexibility(Brander and Eaton, 1985; Hill, 1997). This inturn allows firms to respond rapidly to changesin the marketplace initiated by, for example, acompetitor’s introduction of new products (Brownand Eisenhardt, 1997). To build competitive prod-uct portfolios, firms increasingly attempt to com-bine the benefits from economizing on transac-tion costs through vertical integration (Williamson,1975) with those derived through strategic out-sourcing such as enhanced flexibility and access toa broader stock of knowledge external to the focalfirm (Powell, Koput, and Smith-Doerr, 1996).

Firms vertically integrate to build entry barriers,facilitate investments in specialized assets, pro-tect product quality, and improve scheduling andcoordination (Williamson, 1975; Chandler, 1977;Harrigan, 1984). Vertical integration has the poten-tial to enrich a firm’s new product developmentbecause it provides the opportunity to integratetacit knowledge with complementary assets acrossdifferent value chain activities (Teece, 1986). Intechnologically advanced industries, where suppli-ers often control vitally important new technol-ogy, internalizing these technological capabilitiesaffords control and assures access to the knowl-edge necessary to build a portfolio of productsbased on cutting-edge technology (Afuah, 2001).

While the determination of firm boundariesappears to be mediated by industry structure andfirm capabilities (Argyres, 1996), a fundamentalassumption of transaction cost economics is thatfirms can alter their boundaries based on man-agerial discretion (Williamson, 1975, 1985). Onecould argue, however, that this assumption may nothold in some industries. For example, in the micro-computer industry, the setting for this study, it maybe considered an unlikely option for microcom-puter manufacturers to integrate backwards intochips or forward into operating systems, becausethese value chain activities have been basicallymonopolized by Intel and Microsoft, respectively.While this example provides some evidence thatthe determination of firm boundaries is mediatedby the existing industry structure, recent devel-opments in microcomputing suggest that some

firms have successfully integrated backwards intodesigning their own chips and are beginning todevelop proprietary operating systems (Engardioand Einhorn, 2005). Nonetheless, backward andforward integration along the value chain clearlypose significantly differential strategic challenges.While integration into some value chain activitiesseems to be fairly effortless, integration into otheractivities can be quite difficult, the latter dependingon the industry structure and the capabilities heldby the integrating firm. Nonetheless, Leiblein andMiller (2003) found that internalization is likelyunder conditions of uncertainty regardless of thelevel of asset specificity. Therefore, vertical inte-gration should facilitate new product developmentand thereby contribute to expanding the firm’sproduct portfolio.

Because cutting-edge knowledge necessary forinnovation tends to be widely dispersed acrossdifferent firms, continual innovation in highlydynamic industries appears only possible if a firmreaches beyond its boundaries. This observationhas prompted some to suggest that the locusof innovation might be found in a network ofalliances rather than within individual firms, espe-cially in high-technology industries (Powell et al.,1996; Rothaermel and Deeds, 2004). Access toknowledge external to the firm, in addition tointernal knowledge, enriches a firm’s absorptivecapacity (Cohen and Levinthal, 1990), and enablesfirms to avoid path dependence in the developmentof internal technological knowledge stocks (Collis,1991; Lei, Hitt, and Bettis, 1996). Nicholls-Nixonand Woo (2003), for example, show that the abilityto produce product innovations requires both inter-nal and external R&D investments, leading firms toengage in a variety of strategic alliances. Integrat-ing internal and external technological knowledgestocks allows a firm to build a larger and broaderportfolio of related products in order to gain andmaintain a competitive advantage or to achieve atleast competitive parity.

These arguments suggest that under conditionsof uncertainty prevalent in high-technology indus-tries (Bettis and Hitt, 1995), firms tend to sourcesome of the knowledge necessary for new prod-uct development through strategic alliances. Intheir inductive study of innovation in the computerindustry, Brown and Eisenhardt (1997), foundthat firms using strategic alliances to probe andaccess cutting-edge knowledge external to the



focal firm were more successful in their new prod-uct introductions. Similarly, others established thatthe sharing of knowledge across firm boundariesimproves firm-level innovation when studying flatpanel displays (Spencer, 2003), an important com-ponent in the microcomputer industry. Thus, whenconsidering vertical integration and strategic out-sourcing in isolation, prior research provides con-vincing evidence that each organizational form hasthe potential to expand a firm’s product portfolio.

Herein, we argue that the simultaneous pursuitof vertical integration and strategic outsourcingallows for the integration of internal and exter-nal knowledge stocks to increase the number ofproducts in the firm’s product portfolio. Underconditions of uncertainty, firms can enrich theirproduct portfolios by internalizing current valu-able technological knowledge. At the same time,they must develop and maintain external sourc-ing relationships to gain access to new technicalknowledge developed beyond the firm’s bound-aries (Cohen and Levinthal, 1990; Kotabe, Martin,and Domoto, 2003). Furthermore, the integrationof currently valuable internal technical knowledgeand new external technical knowledge can pro-duce spillover effects that enable firms to profitablyimprove current products as well as to introduceadditional related products in order to gain andmaintain a competitive advantage.

Taper integration is thus a unique organizationalform in which the simultaneous pursuit of verticalintegration and strategic outsourcing has the poten-tial to create synergy that facilitates developmentof new products to increase a firm’s product portfo-lio. Taper integration enables a firm to economizeon transaction costs, to obtain access to diversesources of knowledge, to integrate tacit knowledgeand complementary assets, and thereby to enhanceits strategic flexibility. While this organizationalform clearly increases the complexity of manage-rial tasks such as coordinating and scheduling, andalso has non-trivial bureaucratic costs (Jones andHill, 1988), a larger product portfolio is neverthe-less a significant benefit. Thus, all else being equal,we suggest that firms with greater taper integrationtend to have more products in their portfolio.

Hypothesis 1: The interaction between a firm’sdegree of vertical integration and level of strate-gic outsourcing has a positive effect on the num-ber of related products in the firm’s portfolio.

Balancing vertical integration and strategicoutsourcing

The overarching hypothesis in this study is thatbalancing vertical integration and strategic out-sourcing helps to optimize a firm’s product portfo-lio and to improve product success, thereby con-tributing to a firm’s competitive advantage andfirm performance. Achieving balance implies thattaper integration should be based on a some-what equal emphasis on vertical integration andon strategic outsourcing. A balance in taper inte-gration is achieved when a firm neither focuses toomuch on vertical integration nor on strategic out-sourcing. This in turn implies that each organizingform in isolation should exhibit a curvilinear rela-tionship on valuable firm-level outcomes. Thus, wehypothesize that the direct effects of each verticalintegration and strategic outsourcing on a firm’sproduct portfolio, product success, and firm per-formance are characterized by diminishing returnswith the relationships resembling an inverted U-shaped function.

Extensive vertical integration can producediminishing effects on firm-level outcomes for anumber of reasons. Harrigan (1984), for example,highlights the disadvantages of extensive verticalintegration such as increased managerial costs incoordinating integration over multiple stages ofthe value chain, the potential for either excesscapacity or underutilization of resources becauseof unevenly balanced productivity across differentvalue chain activities, technological obsolescence,strategic inflexibility, increased mobility andexit barriers, tight coupling to poor performingbusiness units, lack of information and feedbackfrom suppliers and distributors, among otherproblems. Thus, the greater the extent of verticalintegration, the lower the degrees of strategicfreedom and the greater the bureaucratic costsassociated with it. When the loss in strategicflexibility and the increase in bureaucratic costsoutweigh the benefits gained through verticalintegration (Jones and Hill, 1988), diminishingreturns result. These arguments suggest thatthe relationship between the degree of verticalintegration and the firm-level outcomes of productportfolio, product success, and performance isinverted U-shaped.

Hypothesis 2: The effects of a firm’s degreeof vertical integration on the size of its prod-uct portfolio (2a), new product success (2b),



and firm performance (2c) are characterized bydiminishing returns such that the relationshipsresemble an inverted U-shape.

In a similar fashion, the extensive pursuitof strategic outsourcing through alliances canalso exhibit diminishing returns on firm-leveloutcomes (Rothaermel, 2001). Several theoreticalarguments support this relationship. First, firmsfrequently compete for the most promisingoutsourcing options, and thus enter them first.Based on the classical Ricardian rent model,this leaves only less productive alliance optionsas firms engage more intensively in strategicoutsourcing. Second, increasing reliance onstrategic outsourcing implies that firms engage inmultiple outsourcing agreements simultaneouslyat any given point in time, and thus managerialattention, frequently a constrained resource, maybecome overloaded and thus inadequate to overseea firm’s alliance activities. Increasing demandson managerial attention in turn accentuates thecognitive limitations of managers. Third, asfirms enter an increasing number of outsourcingalliances, their commensurate transaction andbureaucratic costs increase, beyond a point wheregains to additional alliances are outweighed bytheir marginal costs (Jones and Hill, 1988), thusproducing diminishing returns.

Hypothesis 3: The effects of a firm’s degree ofstrategic outsourcing on the size of its prod-uct portfolio (3a), new product success (3b),and firm performance (3c) are characterized bydiminishing returns such that the relationshipsresemble an inverted U-shape.

Product portfolio, product success, and firmperformance

Firms tend to add new products to their portfo-lios as they acquire new knowledge and integrateit with their existing knowledge base, in particularin highly dynamic industries. The new knowledgeoften builds upon the existing knowledge, allow-ing for improvements in existing products such ashigher quality and greater ability to satisfy con-sumer needs. As a result, this process of knowledgecreation and integration often improves the successof the related products in the portfolio. The mixof different knowledge stocks enriches the firm’scapability to expand its product portfolio and to

offer a greater variety of related products; in sodoing, the firm can better satisfy customer needsin a manner superior to competitors’ product offer-ings (Brown and Eisenhardt, 1997; Sirmon, Hitt,and Ireland, 2007).

In a highly dynamic industry, a larger productportfolio often is important to gain and maintain acompetitive advantage. Technological and marketuncertainties provide entrepreneurial opportunities(McGrath and MacMillan, 2000), which firms canexploit by developing new technologies and prod-ucts that satisfy market requirements. Firms withrich product portfolios are also in a position togain first mover advantages as opportunities arise(Lieberman and Montgomery, 1988). In addition toknowledge spillovers, a large portfolio of relatedproducts also helps to achieve economies of scaleand scope in production as well as marketing anddistribution, thus reducing unit production costswhile simultaneously increasing product variety(Kotha, 1995). Furthermore, firms that regularlydevelop and introduce new products to the markettend to improve their innovation processes overtime, and are thus in a better position to affectmarket change that favors their new product intro-ductions (Brown and Eisenhardt, 1997).

Product portfolios, however, must also bemanaged. While larger product portfolios providemultiple benefits, they also produce non-trivialmanagerial costs. Even related products engen-der coordination costs. As a result, managersmay experience information overload, as has beendemonstrated by firms with more diversified prod-uct portfolios (Hoskisson and Hitt, 1994). In fact,to create synergy and to gain the greatest valuefrom related products, substantial coordination isoften required, particularly with a large portfo-lio of related products. A large portfolio requiresmanagers to carefully coordinate a diverse set ofactivities along the value chain to meet marketdemand; the managerial attention that can be givento any one product is thus limited. In addition, mar-keting budgets can also be diluted in firms withoverly large product portfolios. As a result, over-all product success is likely to be lower if productportfolios are increased indiscriminately. Withoutefficient managerial coordination to compensatefor increasing production complexities, new prod-uct introductions can suffer from poor productquality, lack of necessary differentiation, and othershortcomings, directly reducing firm performance.



When striving to optimize new product successand firm performance, it is necessary for firms toachieve a close match between the size of theirproduct portfolios and external market demand,otherwise negative performance implications arelikely to exist. For example, the technology-basedfirm, Texas Instruments (TI), experienced a drasticerosion of its market share because of its single-minded focus on the experience curve concept,which required product portfolios with large scaleand low variety. Over time, TI had developed astrong core competence in high-volume, low-costmanufacturing (Prahalad and Hamel, 1990). Asmarket demand shifted, however, towards calcu-lators with more features, TI was surpassed bycompetitors, Casio and Hewlett-Packard, becauseits strong core competence became a core rigidity(Leonard-Barton, 1992). In contrast, Sony attemptsto maintain an innovative product portfolio thatprovides a variety of high-quality products whilesimultaneously maintaining its product portfolio ata reasonable size. Sony holds the actual numberof its Walkman models on the market more orless constant at any given time. It maintains thisbalance by continuously introducing new modelswhile retiring older ones (Sanderson and Uzumeri,1995). Integrating the set of arguments suggeststhat the effects of the size of the firm’s productportfolio on product success and firm performanceare curvilinear, such that they are positive for smalland medium size product portfolios, but turn neg-ative for large product portfolios.

Hypothesis 4: The effects of the size of a firm’sproduct portfolio on new product success (4a)and firm performance (4b) are characterized bydiminishing returns such that the relationshipsresemble an inverted U-shape.

METHODS

Research setting

The research setting for this study is the globalmicrocomputer industry. This industry is com-posed of firms manufacturing a variety of smallercomputers such as desktops, laptops/notebooks,servers/workstations, and handhelds. We chose themicrocomputer industry for several reasons.

First, this industry exhibited considerable growthin the past two decades, largely due to continuous

technological innovations that resulted in greatercomputing power and simultaneously lower prices.For example, Moore’s law predicting an expo-nential growth in the number of transistors perintegrated circuit every 12–18 months became aself-fulfilling prophecy. Second, a large number offirms compete in this industry, thereby enablingmeaningful firm-level statistical analysis over anumber of years. The microcomputer industry’sHerfindahl–Hirschman index, for example, was811 during the mid-1990s, indicating a fairly lowlevel of concentration in the industry. Third, firmsin this industry tend to introduce many new prod-ucts. As a result, the random factors that mightaffect the success of any single new product devel-opment effort can be balanced over time, resultingin more reliable measures of a firm’s product port-folio and product success. Moreover, the micro-computer industry is generally not characterizedby blockbuster successes based on a single productintroduction, unlike the pharmaceutical industry,which attenuates a potential skewness in proxiesfor product success. Finally, while many micro-computer firms tend to pursue some degree of taperintegration, they differ significantly in the extentto which they emphasize vertical integration andstrategic outsourcing, respectively. For example,IBM maintained a relatively high level of verti-cal integration over time, while Dell outsources alarge number of its components.

These characteristics explain why the microcom-puter industry is viewed as a hypercompetitive,high-velocity environment (D’Aveni, 1994; Bet-tis and Hitt, 1995; Brown and Eisenhardt, 1997).Overall, the microcomputer industry presents asuitable research setting to test our theoreticalmodel of the relationships among vertical integra-tion, strategic outsourcing, product portfolio, prod-uct success, and firm performance.

To test the hypotheses advanced, we createda longitudinal panel dataset covering the 4-yeartime period between 1994 and 1997. We focusedon this particular time period for several rea-sons. First, this period was characterized by incre-mental product innovations in the microcomputerindustry within a given dominant design (Teece,1986; Anderson and Tushman, 1990), thus allow-ing us to control for more radical innovations.Second, the microcomputer industry is a standards-driven industry to ensure that all componentswork together reasonably well. The microcomput-ing standard was established in the early 1980s



with the introduction of the IBM PC built on anopen architecture with a Microsoft operating sys-tem (MS-DOS) and Intel’s X86 microprocessorarchitecture (Hill, 1997). Finally, the time periodunder investigation witnessed the rise of multime-dia applications and the Internet, in addition to thecontinued convergence with the telecommunica-tions and consumer electronics industries. Thesefactors placed a premium on continuously intro-ducing successful new products. A similar timeframe was used by Brown and Eisenhardt (1997)in their study of continuous organizational changeand product innovations in the computer industry.

Sample and data

We gathered the data for this study from theComputer Select and Computer Company Profilesdatabases, numerous computer industry and com-pany publications, Security Data Company (SDCPlatinum), Lexis/Nexis, Compustat, Mergent FIS,SEC 10-K reports, and the U.S. Patent and Trade-mark Office. While several of the sources havebeen used in strategic management research, oth-ers are unique. The latter category includes theComputer Select and Computer Company Profilesdatabases published by Ziff Communications ofNew York. These databases are publicly availableand contain abstracts, full text, and graphics fromover 150 computer and general publications. Theyare published monthly, and each issue contains arolling year of data. The Computer Company Pro-files database is factual and objective in nature anddoes not include any comparative or evaluativeanalyses. This database lists qualitative informa-tion about each firm such as its participation inthe industry value chain, product portfolio, yearof founding, whether the firm is public or pri-vate, whether the firm is a U.S. or an internationalfirm, etc.

This sampling approach required us to addresscensoring issues. Some firms in the sample didnot offer applicable products at the beginning ofthe study, and other firms withdrew from the mar-ket within the time frame of the study. Ignoringdata from these firms creates a potential bias, sowe included all firms that received at least fourqualifying product reviews, which equates, on theaverage, to one product review per year for eachfirm during the study period. The purpose forchoosing this cut-off point was to ensure that an

adequate number of data points would be avail-able for each firm. Moreover, including multipleproduct reviews for a single firm enhanced thereliability and validity of the new product successmeasure.

The worldwide population of microcomputercompanies as listed in the Computer CompanyProfiles database consisted of 224 firms as ofDecember 1997. Applying the sampling framedescribed above, we were able to obtain com-plete data on 123 microcomputer manufactures(55% of the population), for a sample size of 492firm years over the 4-year study period. The sam-ple is quite diverse with small and large, publicand private, U.S. and non-U.S. firms. Comparingthe means of the sample with those of the pop-ulation reveals, however, that the sample is notsignificantly different statistically (annual revenuesp < 0.63; number of employees p < 0.40; publicvs. private ownership p < 0.56; U.S. vs. non-U.S.firms p < 0.53), which further enhances our con-fidence in the representativeness of the sampledrawn.

A second issue in the sample constructioninvolved methods of treating subsidiaries of alarger parent firm. We aggregated data fromsubsidiaries to obtain a composite score for asingle parent firm. For example, this was donein the case of AT&T and NCR. NCR wasacquired by AT&T in 1991 and subsequentlyspun-off in the fall 1996 trivestiture of AT&T,the new NCR, and Lucent Technologies. Allof the NCR products that qualified for productreviews were pre-trivestiture products and henceclassified as AT&T products. Similarly, all of thequalifying Packard-Bell product reviews occurredafter the July 1996 merger of Packard-Bell andNEC. Finally, all of the qualifying Zenith DataSystems reviews occurred prior to NEC’s February1996 acquisition of Zenith Data Systems. Adetailed review of the Lexis/Nexis Computing andTelecommunications database revealed no otherchanges in ownership status that affected themeasurement of the variables.

Variables and measures

Product portfolio

We proxied the size of a firm’s product portfolioby the number of microcomputer products thata firm offered in each year between 1994 and



1997. During this time period, all sample firmscombined had a total of 13,125 products in theirportfolios, which equates to an average of 27products per year for each firm. This measureincludes all microcomputer products that the firmssold during each year, rather than limiting it to theproducts that were newly introduced and reviewedin computer publications.

Product success

Measuring the performance of individual newproducts across a different range of productsand a diverse set of companies is challenging,because firms generally do not reveal revenuesfor individual products. Therefore, any attemptto measure the success of newly introducedproducts entails significant ambiguity. One of thegoals for this research, however, is to developand contribute a theoretically proximal and fine-grained measure of the performance of newlyintroduced products. Here, Godfrey and Hill(1995) suggest that researchers should track theobservable consequences of otherwise illusiveconstructs.

Thus, to assess the success of individual productintroductions in the microcomputer industry, werelied on experts’ ratings in industry publications.Expert evaluations have been used in priorresearch, for instance, in the assessment of theperformance of strategic alliances (Lane andLubatkin, 1998). To proxy a firm’s new productsuccess, we gathered individual new productevaluations from microcomputer product reviewscontained in the Computer Select publicationsdatabase. We limited the searches to the top-ten computer industry publications as measuredby computer industry advertising revenues.1 Weemployed two primary criteria in selecting newproduct reviews. First, we required that a minimumof four products be included in the review. Thisprocedure allowed for a more effective comparisonof products because product reviews with fewerthan four products tended to rate all productshighly, and thus were more akin to advertisements.

1 The top-ten industry publications by computer advertisingrevenues (in millions) in 1995 were: (1) PC Magazine ($213.9);(2) Wall Street Journal ($129.4); (3) PC Week ($118.0);(4) Computer Shopper ($103.3); (5) Computer World ($87.3);(6) PC World ($86.1); (7) Computer Reseller News ($80.0);(8) Info World ($79.6); (9) Business Week ($74.8); (10) PCComputing ($70.2) (Computer Industry Almanac, 1996).

Second, the new product reviews had to provide acompetitive evaluation indicating that one or moreof the products were favored over the others. Basedon these criteria, we calculated and recorded theindividual evaluations for 3,559 newly introducedmicrocomputer products during the study period.

We classified each of the 3,559 productsreviewed into one of four coding categories:clear winners = 4; runners-up = 3; neutral = 2;losers = 1.2 Next, we calculated an aggregatedmeasure of new product success at the firm levelof analysis. The development of this measure wascomplicated by a high variance in the selectivityof the different product reviews. For example, aproduct review article with 5 winners out of 100is much more selective than one with 5 winnersout of 10. We corrected for this selectivity inthe development of the product review scores. Wecalculated an adjusted review score to account forthe variance in difficulty inherent in the productreviews. Here, we set the adjusted review score(ARS ) equal to the assigned review score (RS )multiplied by one minus the ex ante probabilityof any single product in that review receiving awinning rating (p). We set the value of p equalto the number of winning products divided bythe total number of products in each review. Bydefinition, the adjusted review scores [ARS =RS × (1 − p)] varied between zero and four.Finally, we calculated the annual adjusted reviewscores for each firm. To accomplish this, we set thedenominator k equal to the number of productsthat were reviewed per firm in each year t . Thefollowing formula summarizes this procedure anddepicts the measure to proxy a firm’s new productsuccess in year t :

New product successt =

kt∑

1

[RS × (1 − p)]

kt

(1).

While the sample firms may compete in differ-ent segments of the microcomputer industry (desk-tops, laptops/notebooks, servers/workstations, andhandhelds), the vast majority of microcomputer

2 Assessment of inter-rater reliability was necessary for theComputer Shopper articles due to the lack of an objectivescheme underlying the review. A random sub-sample of 10Computer Shopper reviews was independently coded by asecond researcher. The inter-rater reliability for these ten articleswas equal to 1.0 (i.e., perfect inter-rater agreement).



products offered in this industry were desktopsand laptops/notebooks (hereafter referred to as lap-tops). The microcomputer industry largely con-sisted of two primary segments—desktops andlaptops—during the study period covering themid-1990s. This skewed distribution between dif-ferent types of microcomputer products is mir-rored in the frequency of product evaluations.Accordingly, the majority of products reviewedwere desktops (66.5% or 2,368 product evalua-tions) and laptops (29.3% or 1,043 product evalu-ations), together composing almost 96 percent (or3,411 product evaluations) of the 3,559 productevaluations. Servers/workstations composed only3.9 percent (138 product evaluations), and hand-helds made up merely 0.3 percent (10 productevaluations) of all products evaluated. To ensurea more homogeneous sample, we eliminated the148 product evaluations for servers/workstationsand handhelds, thereby focusing on the dominantsegments of desktops and laptops (this did not,however, affect the robustness of the results; seethe Appendix for more details). Thus, we used atotal 3,411 product evaluations in the final anal-yses. The average product reviewed received aneutral evaluation in terms of its adjusted reviewscore (2 out of 4), regardless of market segment.

To account, however, for the variation in thefrequency of product evaluations and to assesspotential idiosyncrasies of different market seg-ments, we developed three different measures toassess a firm’s new product success: (1) a com-posite new product performance score for thefull sample containing both desktop and laptopcomputers; (2) a new product performance scorefor desktop computers only; and (3) a new prod-uct performance score for laptop computers only.While the sample comprises 123 firms, 108 firmscompeted in desktops, and 69 firms competed inlaptops, with 54 firms competing in both seg-ments. All firms in the sample competed in desk-tops and/or laptops, and no single firm competedin either servers/workstations or handhelds exclu-sively. Thus, all firms were retained in the sample.

Firm performance

We proxied overall firm performance using totalannual revenues. If a firm’s products succeed in themarketplace, they capture higher revenues regard-less of whether a firm pursues a low-cost lead-ership or a differentiation strategy (Porter, 1985).

Further support for revenues as a performance met-ric is shown by the fact that the large majorityof the sample firms (92%) focus on computers astheir dominant business. In addition, revenue datawere more readily available than alternative per-formance measures such as net income or returnon assets, because the majority of the firms inthe sample were privately held (71%), and somewere international firms (10%). All revenue datawere converted into U.S. dollars corresponding tothe historic dollar value. Due to skewness in therevenue data, we applied a logarithmic transforma-tion. Further, we lagged firm revenues by one timeperiod (Lagged firm performance), and included itin the regression model to control for a potentialspecification bias arising from unobserved hetero-geneity (Jacobson, 1990).

Vertical integration

We based the measure of vertical integration oneach firm’s participation in different activities ofthe industry value chain. In particular, we assessedwhether a certain value chain activity was pursuedwithin the boundaries of the firm. To constructthe vertical integration measure, we combined adeductive approach to the value chain (Porter,1985) with detailed industry descriptions of themicrocomputer industry (e.g., Grove, 1996; Rivkinand Porter, 1999). The microcomputer industryvalue chain consists of five distinct activitiesdepicting a product’s progression from upstream todownstream stages in the value chain (Grove 1996:39-45): (1) chips; (2) computers (desktops and lap-tops); (3) software: operating system; (4) software:applications; (5) sales and service. Drawing ondata from the Computer Select database, we devel-oped a matrix of indicator variables to code eachfirm’s participation/non-participation in the differ-ent activities of the microcomputer industry valuechain (1 = activity pursued in-house). To obtaina vertical integration score at the firm level, wesummed the scores from the five different valuechain activities to obtain an aggregate measure toproxy each firm’s degree of vertical integration.

To assess the validity and reliability of theComputer Select data, a second researcher inde-pendently coded business press coverage duringthe study period drawn from Lexis/Nexis. Thesesources, such as the Wall Street Journal and Busi-ness Week, frequently provide information on the



value chain activities of microcomputer compa-nies. We found the intersource reliability to be r =0.68 (p < 0.001), and thus providing evidence ofacceptable reliability (Cohen et al., 2003). Giventhe variance in the sample with respect to firmsize, level of vertical integration, public and pri-vate ownership, U.S. and international firms, thesimilarity of these relatively coarse-grained datais heartening. Furthermore, the results reportedbelow are consistent regardless of the data sourceemployed, providing further evidence of accept-able reliability.

Based on the sample construction, all firms inthis study were involved in the second value chainstage, in-house manufacturing of microcomputers.Thus, the minimum score of vertical integrationis 1, while the maximum score is 5, implyingfull vertical integration. Indeed, the full range ofvertical integration is represented in the sample.3

Strategic outsourcing

Firms generally structure their outsourcing activ-ities through strategic alliances, defined as ‘vol-untary arrangements between firms involvingexchange, sharing, co-development of products,technologies, or services’ (Gulati, 1998: 293). Wefocused on contractual relationships between firmsbecause such inter-firm cooperation reflect formalcollaboration between independent firms, and thusare more likely to capture the theoretical constructof strategic outsourcing.

To develop fine-grained and theoretically proxi-mate strategic outsourcing measures required sev-eral steps. First, we proxied each firm’s degree ofstrategic outsourcing by the total number of strate-gic alliances in which a firm engaged for each yeart during the study period. To enhance the accuracyand reliability of the strategic outsourcing data, oneresearcher obtained alliance data from the strategic

3 It is important to note that a value of 1 for vertical integra-tion merely captures microcomputer manufacturing, and thus,one could argue, does not represent vertical integration. Apply-ing ranges of 1 to 5 for a firm’s value chain activities ratherthan ranges of 0 to 4, where one would only focus on verticalintegration beyond microcomputer manufacturing was neces-sary, however, to construct fine-grained measures of strategicoutsourcing discussed directly below. Moreover, given that allfirms participate at least in the same one value chain activity(in-house manufacturing), results are affected only when thereis one or more additional value chain activities in which the firmparticipates, producing the necessary variance.

alliance database published by Security Data Com-pany (SDC Platinum),4 while a second researcherindependently coded press coverage documentingthe sample firms’ alliance activity drawn from theLexis/Nexis database. The intersource reliabilitywas r = 0.86 (p < 0.001), suggesting high reli-ability (Cohen et al., 2003).

In a second step, to closely tie the strategic out-sourcing measure to the theory advanced above,we developed two fine-grained measures of strate-gic outsourcing based on a detailed content anal-ysis of each alliance. Overall, the firms in thesample entered a total of 1,205 alliances during thestudy period. Central to our theoretical argumentsis the concept of taper integration, which occurswhen a firm pursues the same value chain activityin-house as well as through strategic outsourcingwith external partners. Thus, to develop two dis-tinct strategic outsourcing measures, we comparedthe content description of each alliance to a firm’svalue chain activities pursued in-house, the proxyfor vertical integration. We coded alliances thatmapped to a value chain activity that a firm pur-sued in-house as strategic outsourcing taper , whilealliances that did not map onto a firm’s value-chainactivities pursued in-house were coded as strategicoutsourcing quasi . The latter construct captures thetheoretical notion of quasi integration (Harrigan,1984), because one part of the value chain is inter-nalized, while another part is conducted throughoutsourcing with external partners.

The 1,205 alliances represented 852 taperalliances (71%) and 353 quasi alliances (29%),and are count measures based on whether analliance mapped onto a value chain activity of thefirm (strategic outsourcing taper) or not (strategicoutsourcing quasi ). Moreover, the discriminantvalidity of the two strategic outsourcing proxiesis highlighted in their low bivariate correlationof r = 0.29 (8.4% common variance), wellbelow the common variance necessary to beconsidered a single construct (Cohen et al.,2003). About 41 percent of the sample firmspursued taper integration, and approximately 36percent of the firms engaged in quasi integration.Additionally, about 22 percent of the samplefirms simultaneously pursued both taper and quasiintegration.

4 For a recent application of the SDC alliance database instrategic management research see Anand and Khanna (2000).



Control variables

We included a diverse set of control variablesto account for other potential effects on afirm’s product portfolio, product success, and firmperformance, and thus to reduce the threat ofa potential specification bias due to unobservedheterogeneity.

Firm age

Research suggests that older firms tend tointroduce more innovations, albeit incrementalones (Sørensen and Stuart, 2000). Over time, firmsalso establish routines and overcome the liabilityof newness (Stinchcombe, 1965), which shouldenhance their performance and likelihood ofsurvival. We calculated firm age as the differencebetween the current year and the firm’s foundingdate for each year t during the study period. Theaverage microcomputer firm in the sample was14 years old.

Herfindahl–Hirschman Index

A firm’s market power, frequently reflective of itssize, changes over time as firms merge, pursuevertical integration, or exit certain industries, allof which affect firm performance. We controlledfor this potential effect when estimating the impactof vertical integration and strategic outsourcing onproduct portfolio and product success by includingin the analyses firm-level Herfindahl–Hirschmanindexes (HHI) for each year t , which wereproxied by each firm’s annual squared marketshare (Carlton and Perloff, 1994).

Employees

When estimating firm performance, a firm-levelHerfindahl–Hirschman index was not a suitablemeasure for firm size, because the constructionof the index directly depends on firm revenues,our proxy for performance. We thus used thenumber of employees as a control for firm sizewhen estimating firm performance, commonlyused to control for firm size in high-technologyindustries (Sørenson and Stuart, 2000; Rothaermeland Deeds, 2004). The average firm in thesample employed about 14,000 people. Becausethe employee data were only available for about60 percent of the sample, we needed to impute the

missing data, which did not affect the robustnessof the results (see Appendix for more detail). Itis noteworthy that other control variables suchas firm age, number of patents, and lagged firmperformance are highly correlated with firm size,and thus we are reasonably confident of ourability to isolate the effects of vertical integration,strategic outsourcing, and product portfolio on firmperformance beyond firm size effects.

Patents

Firm patents are a potentially important input intothe new product development process (Griliches,1990). We proxied each firm’s patenting activityusing a count of the total number of patentsreceived in each year t during the study period.Prior research has established the reliability ofpatent count data because it has shown that patentcount data are highly correlated with citation-weighted patent measures, thus proxying the sameunderlying theoretical construct (Hagedoorn andCloodt, 2003; Stuart, 2000). For example, thebivariate correlation between patent counts andcitation-weighted patents has been shown to beabove 0.77 (p < 0.001) in the pharmaceuticalindustry (Hagedoorn and Cloodt, 2003), and above0.80 (p < 0.001) in the semiconductor industry(Stuart, 2000), indicating some generalizability ofthis assertion. The average microcomputer firm inthe sample obtained about 75 patents per year.

U.S. firm

We controlled for institutional differences (Hen-nart, Roehl, and Zietlow, 1999) by including in theanalyses a dummy variable distinguishing U.S. andnon-U.S. based microcomputer companies (1 =U.S. firm) using the firm’s headquarters location;90 percent of the sample firms were headquarteredin the United States.

Public firm

We controlled for the ownership status of the firmby including a dummy variable with 1 = publicfirm and 0 = private firm. Only about 29 percentof the sample firms were public.

Dominant microcomputers

It is critical to control for the degree of overallfirm diversification when assessing the effect of



different organizing forms on product portfolio,product success, and firm performance in asingle industry. We assessed each firm’s level ofdiversification based on Rumelt’s (1974) seminalwork. He classified a firm’s business as dominant ifthe firm obtains 70 percent or more of its revenuesfrom a single business activity, which is indicativeof a low level of diversification. Building onthis definition, we created a dummy variableindicating that a firm’s dominant business is inmicrocomputers if the firm obtained 70 percentor more of its revenues from microcomputers(1 = dominant business is in microcomputers, 0 =all others). The vast majority of firms in thesample exhibited little unrelated diversification; as92 percent of the firms’ dominant business wasmicrocomputers.

Mergers and Acquisitions

A third alternative to structure innovation—inaddition to vertical integration and strategicoutsourcing—is to engage in mergers andacquisitions (Nicholls-Nixon and Woo, 2003).We controlled for this option by including acount variable indicating the number of mergersand acquisitions in which the focal firm wasengaged in year t during the study period(Mergers & Acquisitions). To assess the validityand reliability of the SDC data, a second researcherindependently coded business press coveragepertaining to M&A activity in the microcomputerindustry during the study period drawn fromLexis/Nexis. We found the intersource reliabilityto be r = 0.72 (p < 0.001), and thus above therecommended criterion (Cohen et al., 2003). Theaverage sample firm engaged in one merger oracquisition every 2 years.

Estimation procedure

The data used in this study are longitudinal, andthus represent a panel dataset. Panel data followa given set of companies over time, and thusprovide multiple observations on each firm. Inthis sample, we followed 123 firms over 4 years,which equals 492 firm years. It is important tonote that a majority of empirical work in strategicmanagement relies on cross-sectional data, anddoes therefore not allow for causal inferences (Hitt,Gimeno, and Hoskisson, 1998). Panel data areconsidered a superior alternative due to distinct

advantages over cross-sectional data (Hsiao, 2003).Panel data allow the researcher to control forthe initial values of the dependent variable andto recognize time lag effects. Panel data alsoenable the researcher to draw on a larger sample,and thereby increase statistical power and reducethe threat of multicollinearity among independentvariables, which in turn enhance the efficiency ofthe econometric estimates (Boyd, Gove, and Hitt,2005).

We used generalized least square (GLS) regres-sion analysis to test the hypotheses advanced(Greene, 2003).5 The GLS procedure producesmore efficient estimates than a general linearregression model, because it minimizes a weightedsum of squared residuals. GLS estimates are cor-rected for autocorrelation and cross-section het-eroscedasticity, while estimating weighted aver-ages of the within and between firm effects. Weapplied a more conservative approach by esti-mating the GLS regression models with Whiteheteroscedasticity-consistent standard errors andcovariances. This estimation procedure producescovariances that are robust to general heteroscedas-ticity, because variances within a cross-section areallowed to differ across time.

All hypotheses advanced in the theoretical modelabove indicate the inclusion of interaction terms,either as linear cross-products of two different vari-ables (Hypothesis 1) or as cross-products of thesame variable to create squared terms to assessthe potential for diminishing returns (Hypothe-ses 2–4). Testing moderated regression modelsrequires the inclusion of the direct effects as wellas the interaction effects. This approach is a rel-atively conservative method for examining inter-action effects, because the statistical significanceof the interaction term is evaluated after all lower-order effects have been controlled (Jaccard, Wan,and Turrisi, 1990).

To enhance the interpretability of the regressionresults and to reduce potential multicollinearity,we standardized all independent variables (Cohenet al., 2003). While neither degrading the qual-ity of the data nor affecting the statistical signif-icance levels, this procedure allows us to directlycompare beta coefficients across different variableswith different scales. The variables for the interac-tion terms were standardized prior to creating the

5 For a recent application of GLS regression analysis in strategicmanagement research, see Kotha and Nair (1995).



respective cross products. To assess the threat ofmulticollinearity, we calculated the variance infla-tion factors (VIFs) for each coefficient. The maxi-mum estimated VIF for all direct effects across thethree different dependent variables was 3.7, andthus well below the recommended ceiling of 10(Cohen et al., 2003).

As expected, the VIFs for the interaction terms,however, were somewhat elevated. This is becausethese interaction terms are either linear cross prod-ucts of two different variables or cross productsof the same variable to construct squared terms.One undesirable consequence of potential mul-ticollinearity is inflated standard errors that canresult in overall statistically significant regres-sion models, while failing to identify statisti-cally significant coefficients for individual vari-ables. Thus, multicollinearity can lead to TypeII errors. The results presented below, however,are not materially influenced by potential multi-collinearity, because most of the individual crossproducts and squared terms are statistically signif-icant. Moreover, the coefficients behave appropri-ately and consistently across different regressionmodels. It is also important to note that potentialmulticollinearity does not bias the coefficient esti-mates or influence the overall model fit (Kennedy,1996: 177).

RESULTS

Table 1 presents the descriptive statistics for andbivariate correlation among the variables, whileTables 2–4 contain the regression results. We firstestimated a respective baseline model for eachdependent variable, containing the control vari-ables only (Models 1, 5, 7, 9, and 11). All modelsused to assess the hypotheses represent a statisti-cally significant improvement over their respectivebaseline models (p < 0.001 in all cases).

In Hypothesis 1, we postulated that the inter-action between a firm’s degree of vertical inte-gration and level of strategic outsourcing has apositive effect on the number of related prod-ucts in the firm’s portfolio. Because the resultsof direct effects cannot be interpreted in a mean-ingful way when the regression contains bothdirect and interaction effects, based on the cross-products of the direct effects of interest (Cohenet al., 2003: 259–260), we first inserted the directeffects for strategic outsourcing quasi, strategic

outsourcing taper, and vertical integration to pre-dict a firm’s product portfolio (Model 2). Thisallows us to isolate the effect of each organiz-ing form individually, while controlling for theother organizing forms. The results in Model 2indicate that each organizing form has a positiveand statistically significant effect (at p < 0.01 orsmaller) on the size of a firm’s product portfo-lio. In Model 3, we inserted the interaction termsto assess the simultaneous effect of vertical inte-gration and strategic outsourcing on the size of afirm’s product portfolio. The statistically signifi-cant positive interaction terms between strategicoutsourcing quasi and vertical integration (p <

0.001) and between strategic outsourcing taper andvertical integration (p < 0.05) indicate that firmspursuing quasi integration or taper integration tendto have larger product portfolios. Thus, the individ-ual effects of strategic outsourcing are enhancedsynergistically in the presence of vertical integra-tion, above and beyond each organizing forms’direct effects. These results provide support forHypothesis 1.

In Hypothesis 2, we stated that the effectsof a firm’s degree of vertical integration on thesize of its product portfolio (Hypothesis 2a), newproduct success (Hypothesis 2b), and firm per-formance (Hypothesis 2c) are characterized bydiminishing returns such that the relationships areinverted U-shaped. Model 4 contains the linearterms for strategic outsourcing quasi, strategicoutsourcing taper, and vertical integration alongwith the squared terms for each organizing formto assess the diminishing returns hypothesis forthe product portfolio (Hypothesis 2a). Model 4does not contain the interaction effects that wereincluded in Model 3, because an inclusion of theinteraction terms would cause the direct effects,necessary to analyze Hypotheses 2 and 3, to beuninterpretable (Cohen et al., 2003: 259–260). Ifinteraction terms are included, the direct effectsof each organizational form can only be under-stood contingent upon the moderating variable, asin Model 3. Based on the results obtained in Model4, we find support for Hypothesis 2a, becausethe linear term of vertical integration is positiveand statistically significant (p < 0.001), while thesquared term is negative, and also statistically sig-nificant (p < 0.001).

We assessed Hypothesis 2b in Models 6, 8,and 10. When relating vertical integration to prod-uct success in the combined desktop and laptop



Tabl

e1.

Des

crip

tive

stat

istic

san

dbi

vari

ate

corr

elat

ion

mat

rix

Mea

nS.

D.

Min

.M

ax.

1.2.

3.4.

5.6.

7.8.

9.10

.11

.12

.13

.14

.15

.

1.Pr

oduc

tpo

rtfo

lio26

.70

31.2

91

189

2.Pr

oduc

tsu

cces

s(f

ull

sam

ple)

1.88

70.

421

0.75

3.77

0.20

3.Pr

oduc

tsu

cces

s(d

eskt

op)

1.90

50.

415

0.75

3.77

0.15

0.94

4.Pr

oduc

tsu

cces

s(l

apto

p)

1.88

00.

526

0.80

3.70

0.21

0.77

0.36

5.Fi

rmpe

rfor

man

ce(s

ales

mill

ion

$)

4,96

014

,715

0.73

87,0

000.

380.

060.

070.

08

6.Fi

rmag

e14

.27

18.8

90

109

0.34

0.02

0.06

0.08

0.64

7.H

erfin

dahl

–H

irsc

hman

inde

x

6.50

26.0

01.

41E

-08

173.

110.

31−0

.02

0.09

−0.0

10.

550.

62

8.E

mpl

oyee

s13

,954

42,7

212

250,

000

0.41

0.01

0.09

0.00

0.66

0.66

0.93

9.Pa

tent

s74

.66

262.

240

1,89

50.

400.

030.

010.

090.

530.

590.

470.

5710

.U

.S.

firm

0.90

0.30

01

0.08

0.08

0.05

0.08

−0.3

1−0

.31

−0.1

5−0

.17

−0.3

511

.Pu

blic

firm

0.29

0.45

01

0.24

−0.0

2−0

.09

0.10

0.50

0.37

0.22

0.29

0.39

−0.3

012

.D

omin

ant

mic

ro-

com

pute

rs

0.92

0.27

01

0.02

0.06

0.03

0.00

−0.4

8−0

.58

−0.3

2−0

.35

−0.5

00.

60−0

.43

13.

Mer

gers

&A

cqui

stio

ns0.

471.

780

150.

550.

080.

050.

090.

520.

490.

670.

770.

64−0

.07

0.29

−0.1

8

14.

Stra

tegi

cou

tsou

rcin

gqu

asi

0.72

3.16

037

0.42

0.03

0.06

0.03

0.42

0.34

0.30

0.38

0.42

−0.0

60.

29−0

.15

0.44

15.

Stra

tegi

cou

tsou

rcin

gta

per

1.73

8.65

099

0.63

0.10

0.07

0.10

0.39

0.42

0.47

0.57

0.57

0.02

0.24

−0.0

20.

720.

29

16.

Ver

tical

inte

grat

ion

1.68

0.90

15

0.50

0.09

0.01

0.24

0.22

0.17

0.15

0.26

0.19

0.25

0.14

0.09

0.37

0.13

0.53



Table 2. Regression results estimating the effects of vertical integration and strategic outsourcing on a firm’sproduct portfolio

Model 1Productportfolio




Constant 25.8400∗∗∗ 25.4988∗∗∗ 25.2716∗∗∗ 31.0313∗∗∗

(0.3038) (0.4465) (0.3803) (0.9175)Firm age 7.3412∗∗∗ 2.7037∗∗∗ 2.4779∗∗∗ 3.6585∗∗∗

(0.5645) (0.4216) (0.4634) (0.4443)Herfindahl–Hirschman index −0.8096 −1.5008∗ 0.4258 −2.8676∗∗∗

(1.2741) (0.7239) (0.4767) (0.8691)Patents 4.4980∗∗∗ 0.0537 4.1861∗∗∗ −0.9984

(0.3336) (0.9925) (0.3064) (1.3565)U.S. firm 1.6933∗∗∗ −0.7690† −0.0800 −1.5760∗∗∗

(0.5204) (0.4753) (0.5148) (0.4591)Public firm 3.6929∗∗∗ 0.9657∗∗∗ 0.8915∗∗∗ 1.9851∗∗∗

(0.4565) (0.2238) (0.1657) (0.3375)Dominant microcomputers 8.6366∗∗∗ 4.3545∗∗∗ 4.9466∗∗∗ 5.5537∗∗∗

(0.7605) (0.6847) (0.5606) (0.6175)Mergers & Acquisitions 10.7128∗∗∗ 4.6616∗∗ 0.3959 6.1638∗∗∗

(1.2346) (1.9680) (1.3676) (1.6369)Strategic outsourcing quasi 5.1958∗∗ 5.0253∗∗∗ 0.8534

(2.0644) (0.8762) (2.1289)(Strategic outsourcing quasi)2 0.1112

(0.2049)Strategic outsourcing taper 8.4980∗∗∗ 4.5643∗∗ 34.0091∗∗∗

(2.1589) (1.7964) (5.1525)(Strategic outsourcing taper)2 −2.1005∗∗∗

(0.3674)Vertical integration 8.0046∗∗∗ 7.3485∗∗∗ 8.7234∗∗∗

(0.2220) (0.3256) (0.4060)(Vertical integration)2 −4.1693∗∗∗

(0.9685)Strategic outsourcing quasi × Vertical integration 4.3557∗∗∗

(1.0310)Strategic outsourcing taper × Vertical integration 1.1831∗

(0.5808)

R2 0.39 0.50 0.54 0.52Improvement over base (�R2) 0.11∗∗∗ 0.15∗∗∗ 0.13∗∗∗

N 486 486 486 486

†p < 0.10; ∗ p < 0.05; ∗∗ p < 0.01; ∗∗∗ p < 0.001. Models are GLS, estimated with White heteroscedasticity-consistentstandard errors (in parentheses) and covariances (corrected for degrees of freedom).

sample (Model 6) and in the desktop market only(Model 8), we find that both the linear and squaredterms are each negative and statistically signifi-cant (p < 0.001). This implies that the relationshipbetween vertical integration and new product suc-cess is non-linearly negative rather than invertedU-shaped. We thus fail to find support for Hypoth-esis 2b in the combined desktop and laptop seg-ments, and in the desktop market. We do find sup-port, however, for a diminishing returns hypothesisbetween vertical integration and product successin the laptop market (Model 10). Here, the linear

term for vertical integration is, as expected, posi-tive and statistically significant (p < 0.001), whilethe squared term for vertical integration is negativeand statistically significant (p < 0.001).

We evaluated Hypothesis 2c, indicating dimin-ishing returns between vertical integration and firmperformance, in Model 12 presented in Table 4. Wefind that the squared term for vertical integrationis positive and statistically significant (p < 0.001).Thus, we fail to find support for Hypothesis 2c.Rather this result implies a non-linear positiveeffect of vertical integration on firm performance.



Table 3. Regression results estimating the effects of vertical integration, strategic outsourcing, and product portfolioon a firm’s product success across different microcomputer segments

Model 5Productsuccess

Full sample

Model 6Productsuccess

Full sample

Model 7ProductsuccessDesktop

Model 8ProductsuccessDesktop

Model 9ProductsuccessLaptop

Model 10ProductsuccessLaptop

Constant 1.8828∗∗∗ 1.9277∗∗∗ 1.9173∗∗∗ 1.9407∗∗∗ 1.8741∗∗∗ 1.9368∗∗∗

(0.0050) (0.0041) (0.0033) (0.0097) (0.0031) (0.0338)Firm age 0.0454∗∗∗ 0.0095 0.0190∗∗ 0.0125 0.0563∗∗∗ 0.0415∗∗

(0.0047) (0.0101) (0.0066) (0.0121) (0.0009) (0.0142)Herfindahl–Hirschman index −0.0805∗∗∗ −0.0493∗∗∗ 0.0332† 0.0566 −0.0778∗∗∗ −0.0729∗∗∗

(0.0058) (0.0119) (0.0211) (0.0456) (0.0084) (0.0133)Patents 0.0043 0.0150 −0.0108 −0.0212 −0.0219∗∗ −0.0445∗∗

(0.0104) (0.0180) (0.0146) (0.0256) (0.0096) (0.0164)U.S. firm 0.0302∗∗∗ 0.0365∗∗∗ 0.0167∗∗∗ 0.0261∗∗∗ 0.0641∗∗∗ 0.0023∗∗

(0.0052) (0.0042) (0.0022) (0.0035) (0.0067) (0.0101)Public firm −0.0123∗∗∗ −0.0193∗∗∗ −0.0295∗∗∗ −0.0269∗∗∗ 0.0552∗∗∗ 0.0611∗∗∗

(0.0009) (0.0034) (0.0014) (0.0039) (0.0043) (0.0066)Dominant microcomputers 0.0101 −0.0066 0.0104† −0.0080 0.0029 0.0014

(0.0149) (0.0127) (0.0066) (0.0090) (0.0118) (0.0154)Mergers & Acquisitions 0.0742∗∗∗ 0.0255∗ 0.0088 −0.0501∗∗∗ 0.0785∗∗∗ 0.0754∗∗∗

(0.0081) (0.0130) (0.0097) (0.0086) (0.0091) (0.0214)Strategic outsourcing quasi 0.0345∗∗∗ 0.0793∗∗∗ −0.1115∗∗

(0.0099) (0.0076) (0.0410)(Strategic outsourcing quasi)2 −0.0689∗∗∗ −0.0113∗∗∗ 0.0165†

(0.0008) (0.0024) (0.0124)Strategic outsourcing taper 0.0393∗∗∗ 0.0857∗∗∗ 0.2033∗

(0.0129) (0.0170) (0.1231)(Strategic outsourcing taper)2 −0.0017† −0.0044∗∗ −0.0163∗

(0.0013) (0.0018) (0.0087)Vertical integration −0.0221∗∗∗ −0.0367∗∗∗ 0.0683∗∗∗

(0.0046) (0.0047) (0.0119)(Vertical integration)2 −0.0128∗∗∗ −0.0145∗∗∗ −0.0507∗∗∗

(0.0023) (0.0039) (0.0110)Product portfolio 0.0982∗∗∗ 0.0569∗∗∗ 0.0830∗∗∗

(0.0073) (0.0088) (0.0078)(Product portfolio)2 −0.0071∗∗∗ 0.0002 −0.0161∗

(0.0012) (0.0025) (0.0095)

R2 0.03 0.06 0.02 0.05 0.05 0.09Improvement over base (�R2) 0.03∗∗∗ 0.03∗∗∗ 0.04∗∗∗

N 312 312 246 246 146 146

†p < 0.10; ∗ p < 0.05; ∗∗ p < 0.01; ∗∗∗ p < 0.001. Models are GLS, estimated with White heteroscedasticity-consistent standarderrors (in parentheses) and covariances (corrected for degrees of freedom).

Taken together, we find support for a diminishingreturns hypothesis between vertical integration andproduct portfolio (Hypothesis 2a) and between ver-tical integration and product success in the laptopmarket (Hypothesis 2b).

In parallel to the vertical integration hypothe-sis (Hypothesis 2), in Hypothesis 3 we suggestedthat the effects of a firm’s degree of strategicoutsourcing on the size of its product portfolio(Hypothesis 3a), new product success (Hypothe-sis 3b), and firm performance (Hypothesis 3c) arecharacterized by diminishing returns such that the

relationships are inverted U-shaped. Because thetheoretical focus of this study is on taper integra-tion, we consider quasi integration to be a controlvariable. Thus, we are able to assess the effectof strategic outsourcing taper above and beyonda firm’s level of strategic outsourcing quasi andvertical integration. We assessed Hypothesis 3a inModel 4, Hypothesis 3b in Models 6, 8, and 10,and Hypothesis 3c in Model 12.

The results in Model 4 provide support fora diminishing returns to the effects of strategicoutsourcing taper and the size of a firm’s product



portfolio, because the linear term for this variableis positive and statistically significant (p < 0.001),while the squared term is negative and also statisti-cally significant (p < 0.001). The results presentedin Models 6, 8, and 10 offer support for Hypoth-esis 3b, because in each model the linear term ofstrategic outsourcing taper is positive and statis-tically significant (at p < 0.05 or lower), whilethe squared term is negative and statistically sig-nificant (at p < 0.10 or lower). Thus, the resultssuggest that the relationship between strategic out-sourcing taper and new product success is invertedU-shaped, across the product categories of thecombined desktop and laptop sample, the desk-top segment alone and the laptop segment alone.The results presented in Model 12 lend support forHypothesis 3c, because the linear term of strategicoutsourcing taper is positive and statistically sig-nificant (p < 0.01), while the squared term is neg-ative and also statistically significant (p < 0.01).Taken together, the results provide support for adiminishing returns effect of strategic outsourcingtaper on the size of the firm’s product portfolio(Hypothesis 3a), new product success (Hypothesis3b), and firm performance (Hypothesis 3c). Theobserved curvilinear relationship between strategicoutsourcing taper and new product success appearsto be quite robust because it holds regardless of theunderlying microcomputer market segment.

In Hypothesis 4, we postulated that the effects ofthe size of a firm’s product portfolio on new prod-uct success (Hypothesis 4a) and firm performance(Hypothesis 4b) are characterized by diminishingreturns such that the relationships are inverted U-shaped. We tested Hypothesis 4a in Models 6,8, and 10. We find support for a diminishingreturns hypothesis of product portfolio size on newproduct success in the combined desktop and lap-top sample (Model 6) and in the laptop segment(Model 10), because in both cases the linear termof product portfolio is positive and statisticallysignificant (p < 0.001), while the squared termis negative and statistically significant (p < 0.05or lower). We fail to find support, however, forthe diminishing returns hypothesis of product port-folio size on new product success in the desk-top segment (Model 8). While the linear term is,as predicted, positive and statistically significant(p < 0.001), the squared term does not reach sta-tistical significance.

We assessed Hypothesis 4b in Model 12. Thediminishing returns hypothesis between product

Table 4. Regression results estimating the effects of ver-tical integration, strategic outsourcing, product portfolio,and product success on firm performance

Model 11Firm

performance

Model 12Firm

performance

Constant 18.6950∗∗∗ 18.4007∗∗∗

(0.2229) (0.2422)Firm age 0.6067∗∗∗ 0.5213∗∗∗

(0.0269) (0.0400)Employees 0.9088∗∗∗ 1.0076∗∗∗

(0.2086) (0.2501)Patents 0.1893∗∗∗ 0.0928∗

(0.0266) (0.0477)U.S. firm −0.1366∗∗∗ −0.1650∗∗∗

(0.0291) (0.0409)Public firm 0.8972∗∗∗ 0.7074∗∗∗

(0.0365) (0.0198)Dominant microcomputers −0.0899∗∗∗ −0.2035∗∗∗

(0.0216) (0.0245)Mergers & Acquisitions 0.0392 0.0070

(0.0536) (0.1269)Lagged firm performance 0.3804∗ 0.1127

(0.2131) (0.2218)Strategic outsourcing quasi 0.5163∗∗

(0.1670)(Strategic outsourcing −0.0363∗∗

quasi)2 (0.0152)Strategic outsourcing taper 0.5183∗∗

(0.2031)(Strategic outsourcing −0.0875∗∗

taper)2 (0.0248)Vertical integration 0.0176

(0.0209)(Vertical integration)2 0.1811∗∗∗

(0.0247)Product portfolio 0.2890∗∗∗

(0.0337)(Product portfolio)2 −0.1271∗∗∗

(0.0217)

R2 0.59 0.64Improvement over base

(�R2)0.05∗∗∗

N 486 486

†p < 0.10; ∗ p < 0.05; ∗∗ p < 0.01; ∗∗∗ p < 0.001. Models areGLS, estimated with White heteroscedasticity-consistent stan-dard errors (in parentheses) and covariances (corrected fordegrees of freedom).

portfolio size and firm performance is supported,because the linear term for product portfolio ispositive and statistically significant (p < 0.001),while the squared term is negative and statisticallysignificant (p < 0.001). To determine the theoreti-cally optimal size of a firm’s product portfolio thatmaximizes its revenues, we calculated the absolute



value of the partial derivative with respect to prod-uct portfolio size (0.2890/[2 × 0.1271] = 1.1369).We know that this is a maximum in the functionrelating product portfolio size to revenues becausethe second partial derivative with respect to prod-uct portfolio is negative. Because we standardizedall independent variables prior to entering themin the regression equations to enhance their inter-pretability and comparability, we had to transformthe standardized optimum into a non-standardizedsolution for accurate interpretation. Note that thestandardized optimum is a z-score; thus the non-standardized solution is obtained as follows:

X = mean + (standardized z-score

× standard deviation), or

X = 26.70 + (1.1369 × 31.29) = 62.27

Therefore, the average firm reaches its optimumproduct portfolio size when attempting to maxi-mize revenues at about 62 products in its portfolio.We need to be mindful, however, that this num-ber is more suggestive than normative because itis context dependent in terms of the underlyingsample and time period studied. In sum, we findsupport for Hypothesis 4a (except in the desktopsegment) and Hypothesis 4b, indicating that therelationship between the size of a firm’s productportfolio and new product success (Hypothesis 4a)and firm performance (Hypothesis 4b) is invertedU-shaped.6

DISCUSSION

Scholars in multiple disciplines have focused sus-tained attention on the boundaries of the firm.This emphasis has been particularly evident inthe prolific literature on vertical integration andstrategic alliances (cf. Gulati, 1998; Leiblein andMiller, 2003). Yet, these two significant streamsof research have not been integrated to under-stand the benefits and costs of internalizing somevalue chain activities and simultaneously outsourc-ing other value chain activities. This practice,however, has become increasingly common sinceHarrigan (1984) identified it over 20 years ago.Simultaneous internalization and outsourcing of

6 We conducted a number of robustness checks which we reportin the Appendix.

value chain activities are referred to as taper inte-gration and quasi integration; taper integration isof special importance for this study. We add valueto our knowledge of vertical integration and strate-gic outsourcing, because we relate the theoreticalarguments explaining the underlying benefits andcosts of taper integration to the development of thefirm’s product portfolio, the success of new prod-ucts introduced to the market, and ultimately tofirm financial performance.

The totality of the results provides strong evi-dence for the value of balancing vertical integra-tion and outsourcing of value chain activities, andespecially of engaging in taper integration. Taperintegration was found to have a positive effecton both development of related products for thefirm’s product portfolio and on the success of thosenew products upon introduction to the market-place. Taper integration provides several potentialbenefits.

In particular, taper integration, more than mostother forms of outsourcing, provides a significantpotential to access new external knowledge andto internalize it. By performing some activitiesin a particular stage of the value chain internallyand some externally, a firm keeps its pulse onexternal technology and new knowledge devel-oped outside the firm (Cohen and Levinthal, 1990).Importantly, the firm also has the absorptive capac-ity to learn the new knowledge and to quicklyapply it within the firm. In this way, the firmcan develop more effectively new products thatare technologically current. Because taper integra-tion allows a firm to be at the cutting edge oftechnology by internalizing it quickly, the firm ismore likely to enjoy success with its new prod-ucts introduced to the market. Additionally, taperintegration provides the benefits from the synergycreated by integrating complementary resourcesfrom the external partner with the firm’s inter-nal resources (Teece, 1986), and thus enables thefirm to develop new advantage-creating capabili-ties (Sirmon et al., 2007). Finally, taper integrationaffords more strategic flexibility to the firm. Oper-ating in dynamic environments common in high-technology industries, the use of taper integrationallows a firm to internalize a complete value chainstage or to outsource it. It can also more easilychange to use a new technology without signif-icant losses of current assets. For example, Delloften outsources design activities to firms suchas Flextronics, while maintaining in-house R&D



capabilities to develop many of the product com-ponents. Today, Flextronics can develop superiordesigns at a lower cost than many of its primarycustomers. For example, Flextronics reported thatit can design a high-end cell phone for about$10 million when the average cost to design sucha phone by the original equipment manufacturer(OEM) is $30–50 million (BusinessWeek Online,2005).

The benefits from taper integration are critical,but our research strongly suggests the importanceof balance. There are limits to the benefits oftaper integration, thus, increasing taper integrationbeyond some point begins to produce decreasingreturns. Overuse of external sources to completevalue chain activities could lead to opportunismand excessive transaction costs. It also reduces thefirm’s ability to absorb external knowledge therebydecreasing the learning opportunities. While exter-nal sourcing is likely to increase a firm’s flexibilityin the short term, it also increases its path depen-dence in the use of external sources. As the firmloses its internal capability to perform certain valuechain activities, it becomes increasingly dependenton its external partners to perform those activi-ties (Bettis, Bradley, and Hamel, 1992; Lei et al.,1996). Firms in which design is a core competenceshould not outsource the activity (Prahalad andHamel, 1990). For example, Steve Jobs claims thatApple is able to design ‘insanely great’ new prod-ucts that contribute to a competitive advantage.While Apple may design products with high mar-ket demand, it also engages in outsourcing (quasi)by having its products manufactured by outsourceddesign manufacturers (ODMs) (Burrows, 2005).

Other results in this study also support theimportance of balance. For example, while ver-tical integration was found to have an increasinglypositive effect on firm performance, it also hadan increasingly negative effect on new productsuccess (except in the laptop segment, where therelationship was inverted U-shaped). The positiveeffect on performance is likely due to the reduc-tion in transaction costs by internalizing the valuechain activities. Yet, vertical integration reducesaccess to new knowledge that can be used todevelop successful new products. Full vertical inte-gration, for example, could create an environmentin which the firm operates almost like a closedsystem because all value chain activities are inter-nalized with few external linkages. This in turnreduces a firm’s strategic flexibility to respond to

changing technologies and other contingencies inits environment, for example, and thus commen-surately enhances the probability of firm obso-lescence. Thus, while vertical integration reducestransaction costs, it also creates opportunity costswith potential negative performance implicationsfor the firm in the long term. The difference inthe effects of vertical integration on new productsuccess in the desktop market (non-linear negativeeffect) and the laptop market (diminishing returnseffect) could possibly be attributed to a differentiallevel of maturity in these two industry segments,with the desktop segment being more mature thanthe laptop segment.

The results also suggest that increases to afirm’s product portfolios generally improve newproduct success and lead to higher firm perfor-mance. Firms that increase their product portfo-lios too much, however, also experience diminish-ing returns. Larger product portfolios allow firmsto gain synergy from related products. This syn-ergy can materialize in the form of economiesof scale and scope, and thus result in lower unitproduct costs and higher sales through comple-mentary products that enhance customer demand.Alternatively, a large portfolio requires coordina-tion among the various products to achieve theseeconomies (Sanderson and Uzumeri, 1995). Anoverly large product portfolio can increase man-agerial costs to the point where they overcome thebenefits.

Limitations and future research

This study has limitations, which in turn pro-vide opportunities for future research. One limi-tation is the way we proxied some of our mea-sures. For example, while the results concerningnew product success are intuitive and fit wellwith the theoretical arguments advanced, the out-come could partly be due to the measure ofnew product success. The ratings of newly intro-duced products are conducted by individuals whooften have a technological lens. As a result, prod-ucts with cutting-edge technology might be ratedhighly but the advance in technology may begreater than desired by the customers to sat-isfy their needs or the costs are too high togenerate profits from the sales. This resonateswith Christensen and Bower’s (1996) observationthat, over time, the rate of technological progress



frequently exceeds the performance improvementsdemanded in a market.

While the possibility of a technology bias byindustry experts exists, careful steps were takenin this study to ensure the construct validity ofthe measures employed. In designing and exe-cuting this study, we followed many of the rec-ommendations made by Boyd et al. (2005). Insum, while this study is one of very few thathas measured the success of individual new prod-uct introductions, more research is needed toexamine the performance of newly introducedproducts.

Because we had to rely on firm revenues asa proxy for firm performance (due to the factthat the large majority of firms in this sam-ple were privately held), using more fine-grainedmeasures of firm performance in future researchmay help to validate the findings of this study.Additionally, more fine-grained measures of ver-tical integration might contribute to our knowl-edge of appropriate levels of internalization andespecially identifying which activities should bemaintained in-house and which ones should be out-sourced. This knowledge would provide the basefor future research to determine the appropriatebalance between vertical integration and strategicoutsourcing. This question is particularly interest-ing because firm boundaries are dynamic, and thuschange over time (Afuah, 2001). In particular, theeffects of taper integration on product portfolio,product success, and firm performance are likelyto vary depending on the stage of the industryevolution and the dynamism inherent in the indus-try (Leonard-Barton, 1992; Eisenhardt and Martin,2000).

While the sample is representative of the popu-lation of microcomputer manufacturers, there is aneed to test the theoretical model in other industrysettings to establish the generalizability of our find-ings. In addition, future research should study timeperiods characterized by both incremental and rad-ical innovations, because the emergence of a newdominant design is likely to demand a more contin-gent approach to organizing for innovation (Ander-son and Tushman, 1990; Afuah, 2001). Theseefforts would help in strengthening the externalvalidity of the theoretical model developed andtested herein. In sum, the measures for strategicoutsourcing (taper and quasi) provide a method-ological contribution. More importantly, we hope

that this contribution spurs future research on thisimportant phenomenon.

Conclusion

This study contributes to our understanding ofthe boundaries of the firm. Prior research focusedprimarily on vertical integration (internalization)or on external strategic alliances and networks.Thus, the results obtained in prior research mayhave been an artifact of unobserved heterogene-ity when specifying that outcomes were due toone specific organizing form, unless alternativeorganizing forms were controlled in the analyses,which was rare. By contrast, we simultaneouslyinvestigate the effect of vertical integration andstrategic outsourcing, while controlling for mergersand acquisitions. The new findings presented pro-vide broad support regarding the need for balancebetween internalization and outsourcing of valuechain activities. In particular, we show the valueof taper integration in developing new productsand their success in the marketplace.

This study also provides some relevant manage-rial implications. Our results sound a cautionarynote that pursuing either vertical integration orstrategic outsourcing in isolation appears to be sub-optimal. While this is clearly an important questionat the transaction level of analysis, a manager gen-erally faces questions pertaining to boundaries ofthe firm at a more aggregate level. In particu-lar, a manager frequently needs to determine thedegree of vertical integration and the extent ofstrategic outsourcing simultaneously. Here, balanc-ing vertical integration and strategic outsourcing toachieve a prudent level of taper integration seemsto be a valuable strategy in striving for superiorperformance.

When discussing options of dynamic firmboundaries within the context of the microcom-puter industry, we noted that integration into somevalue chain activities appears to be fairly effort-less, while integration into other parts of the valuechain seems to be rather difficult. If managersidentify value chain activities in which the firmshould participate, but are not able to verticallyintegrate into them, quasi integration provides apotentially valuable option. In this case, the neededvalue chain activities are accessed through strate-gic outsourcing.

Taken together, managers should strive to iden-tify the appropriate level of taper integration to



generate a product portfolio that contributes to the‘optimum’ levels of new product success and firmperformance. While finding the appropriate bal-ance between internalizing value chain activitiesand strategic external sourcing can be difficult,maintaining the right balance over time can proveeven more challenging because the competitivelandscape is often highly dynamic (Bettis and Hitt,1995; Brown and Eisenhardt, 1997). We thus sug-gest that matching the appropriate level of taperintegration with a firm’s resources and with theindustry environment is a firm dynamic capabil-ity, which has been described as a ‘firm’s abilityto integrate, build, and reconfigure internal andexternal competences to address rapidly changingenvironments’ (Teece, Pisano, and Shuen, 1997:516). In conclusion, taper integration seems tobe beneficial in organizing for innovation, but anintricate balance appears to exist that optimizesa firm’s product portfolio, product successes, andoverall performance. Discovering and maintainingthis balance between vertical integration and strate-gic outsourcing is a critical and challenging, butpotentially rewarding, task for managers.

ACKNOWLEDGEMENTS

An earlier version of this paper was presented atthe 2004 Strategic Management Society (SMS)Conference, and awarded the SMS ConferenceBest Paper Prize. We thank Associate EditorRichard Bettis, the anonymous reviewers, SharonAlvarez, Nathan Bennett, Terry Blum, WarrenBoeker, John Butler, John Hollenbeck, Tom Lee,Leonard Parsons, Susan Sanderson, Robert Wise-man, and the seminar participants at the RensselaerPolytechnic Institute, Texas Tech University, Uni-versity of California at Irvine, and the Universityof North Carolina for helpful comments and sug-gestions on earlier drafts of this paper. We alsothank Shanti Agung, Jay Cheng, Stephen Hin-man, Swastika Mukherjee, and Xiaoguang Yangfor research assistance, and Karyn Lu for edito-rial assistance. All errors and omissions remainentirely our own. Rothaermel is an Alfred P. SloanFoundation Industry Studies Fellow, and gratefullyacknowledges financial support from this fellow-ship. Rothaermel also acknowledges financial sup-port from the E. M. Kauffman Foundation.

REFERENCES

Afuah A. 2001. Dynamic boundaries of the firm: arefirms better off being vertically integrated in the faceof a technological change? Academy of ManagementJournal 44: 1211–1228.

Anand BN, Khanna T. 2000. Do firms learn to createvalue? The case of alliances. Strategic ManagementJournal , Special Issue 21(3): 295–315.

Anderson P, Tushman ML. 1990. Technological discon-tinuities and dominant designs: a cyclical model oftechnological change. Administrative Science Quar-terly 35: 604–633.

Argyres N. 1996. Evidence on the role of firm capabilitiesin vertical integration decision. Strategic ManagementJournal 17(2): 129–150.

Bayus BL, Erickson G, Jacobson R. 2003. The financialrewards of new product introductions in thepersonal computer industry. Management Science 49:197–210.

Bettis RA, Bradley SP, Hamel G. 1992. Outsourcing andindustrial decline. Academy of Management Executive6: 7–22.

Bettis RA, Hitt MA. 1995. The new competitivelandscape. Strategic Management Journal , SummerSpecial Issue 16: 7–20.

Boyd BK, Gove S, Hitt MA. 2005. Construct mea-surement in strategic management research: illu-sion or reality? Strategic Management Journal 26(3):239–257.

Brander J, Eaton J. 1985. Product line rivalry. AmericanEconomic Review 74: 323–334.

Brown SL, Eisenhardt KM. 1997. The art of continuouschange: linking complexity theory and time-pacedevolution in relentlessly shifting organizations.Administrative Science Quarterly 42: 1–34.

Burrows P. 2005. Commentary: Apple’s blueprint forgenius. BusinessWeek Online, www.businessweek.com, 21 March.

BusinessWeek Online. 2005. Design is a commodity.www.businessweek.com, 21 March.

Carlton DW, Perloff JM. 1994. Modern IndustrialOrganization (2nd edn). Harper Collins: NewYork.

Chandler AD. 1977. The Visible Hand. Harvard Univer-sity Press: Cambridge, MA.

Christensen CM, Bower JL. 1996. Customer power,strategic investment, and the failure of leading firms.Strategic Management Journal 17(3): 197–218.

Coase RH. 1937. The nature of the firm. Econometrica386–405.

Cohen P, Cohen J, West SG, Aiken LS. 2003. AppliedMultiple Regression/Correlation Analysis for theBehavioral Sciences (3rd edn). Erlbaum: Hillsdale,NJ.

Cohen W, Levinthal D. 1990. Absorptive capacity: a newperspective on learning and innovation. AdministrativeScience Quarterly 35: 128–152.

Collis DJ. 1991. A resource-based analysis of globalcompetition: the case of the bearings industry.Strategic Management Journal , Summer Special Issue12: 49–68.



Computer Industry Almanac. 1996. Brady: New York.D’Aveni RA. 1994. Hypercompetition: Managing the

Dynamics of Strategic Maneuvering. Free Press: NewYork.

Dyer JH. 1996. Specialized supplier networks as asource of competitive advantage: evidence from theauto industry. Strategic Management Journal 17(4):271–291.

Dyer JH, Singh H. 1998. The relational view: cooperativestrategy and sources of interorganizational competitiveadvantage. Academy of Management Review 23:660–679.

Eisenhardt KM, Martin JA. 2000. Dynamic capabilities:what are they? Strategic Management Journal ,Summer Special Issue 21(10–11): 1105–1121.

Engardio P, Einhorn B. 2005. Outsourcing innovation.Business Week 21 March: 84–94.

Folta TB. 1998. Governance and uncertainty: thetrade-off between administrative control and com-mitment. Strategic Management Journal 19(11):1007–1028.

Galbraith J. 1977. Organization Design . Addison Wesley:Reading, MA.

Godfrey PC, Hill CWL. 1995. The problem of unob-servables in strategic management research. StrategicManagement Journal 16(7): 519–533.

Greene WH. 2003. Econometric Analysis (5th edn).Prentice-Hall: Upper Saddle River, NJ.

Griliches Z. 1990. Patent statistics as economic indica-tors: a survey. Journal of Economic Literature 28:1661–1707.

Grove AS. 1996. Only the Paranoid Survive: Howto Exploit the Crisis Points that Challenge EveryCompany and Career. Currency Doubleday: NewYork.

Gulati R. 1998. Alliances and networks. StrategicManagement Journal , Special Issue 19: 293–317.

Hagedoorn J. 1993. Understanding the rationale ofstrategic technology partnering: interorganizationalmodes of cooperation and sectoral differences.Strategic Management Journal 14(5): 371–385.

Hagedoorn J, Cloodt M. 2003. Measuring innovativeperformance: is there an advantage in using multipleindicators? Research Policy 32: 1365–1379.

Harrigan KR. 1984. Formulating vertical integrationstrategies. Academy of Management Review 9:638–652.

Hennart J-F, Roehl T, Zietlow DS. 1999. Trojan horseor workhorse? The evolution of U.S.–Japanese jointventures in the United States. Strategic ManagementJournal 20(1): 15–29.

Hill CWL. 1997. Establishing a standard: competitivestrategy and technological standards in winner-take-all industries. Academy of Management Executive 11:7–25.

Hill CWL, Hoskisson RE. 1987. Strategy and structurein the multi-product firm. Academy of ManagementReview 12: 331–341.

Hill CWL, Rothaermel FT. 2003. The performance ofincumbent firms in the face of radical technologicalinnovation. Academy of Management Review 28:257–274.

Hitt MA, Gimeno J, Hoskisson RE. 1998. Current andfuture research methods in strategic management.Organizational Research Methods 1: 6–44.

Hoskisson RE, Hitt MA. 1994. Downscoping: How toTame the Diversified Firm . Oxford University Press:New York.

Hsiao C. 2003. Analysis of Panel Data (2nd edn).Cambridge University Press: Cambridge, U.K.

Ireland RD, Hitt MA, Vaidyanath D. 2002. Alliancemanagement as a source of competitive advantage.Journal of Management 28: 413–446.

Jaccard J, Wan CK, Turrisi R. 1990. The detection andinterpretation of interaction effects between contin-uous variables in multiple regression. MultivariateBehavioral Research 25: 467–478.

Jacobson R. 1990. Unobservable effects and businessperformance. Marketing Science 9: 74–85.

Jones GR, Hill CWL. 1988. Transaction cost analysisof strategy-structure choice. Strategic ManagementJournal 9(2): 159–172.

Kennedy P. 1996. A Guide to Econometrics (3rd edn).MIT Press: Cambridge, MA.

Kogut B. 1988. Joint ventures: theoretical and empiricalperspectives. Strategic Management Journal 9(4):319–332.

Kotabe M, Martin X, Domoto H. 2003. Gaining fromvertical partnerships: knowledge transfer, relationshipduration, and supplier performance improvement inthe U.S. and Japanese automotive industries. StrategicManagement Journal 24(4): 293–316.

Kotha S. 1995. Mass customization: implementingthe emerging paradigm for competitive advantage.Strategic Management Journal , Summer Special Issue16: 21–42.

Kotha S, Nair A. 1995. Strategy and environmentas determinants of performance: evidence fromthe Japanese machine tool industry. StrategicManagement Journal 16(7): 497–518.

Lane PJ, Lubatkin M. 1998. Relative absorptive capac-ity and interorganizational learning. Strategic Manage-ment Journal 19(5): 461–477.

Lei D, Hitt MA, Bettis RA. 1996. Dynamic corecompetences through meta-learning and strategiccontext. Journal of Management 22: 549–569.

Leiblein MJ, Miller DJ. 2003. An empirical examinationof transformation-and firm-level influences on thevertical boundaries of the firm. Strategic ManagementJournal 24(9): 839–859.

Leiblein MJ, Reuer JJ, Dalsace F. 2002. Do make orbuy decisions matter? The influence of organizationalgovernance on technological performance. StrategicManagement Journal 23(9): 817–833.

Leonard-Barton D. 1992. Core capabilities and corerigidities: a paradox in managing new product devel-opment. Strategic Management Journal , Summer Spe-cial Issue 13: 111–125.

Lieberman MB, Montgomery DB. 1988. First-moveradvantages. Strategic Management Journal , SummerSpecial Issue 9: 41–58.



McGrath RG, MacMillan IC. 2000. The EntrepreneurialMindset: Strategies for Continuously Creating Oppor-tunity in an Age of Uncertainty . Harvard BusinessSchool Press: Boston, MA.

Nicholls-Nixon C, Woo C. 2003. Technology sourcingand output of established firms in a regimeof encompassing technological change. StrategicManagement Journal 24(7): 651–666.

Pisano GP. 1990. The R&D boundaries of the firm: anempirical analysis. Administrative Science Quarterly35: 153–176.

Porter ME. 1985. Competitive Advantage: Creating andSustaining Superior Performance. Free Press: NewYork.

Powell WW, Koput KW, Smith-Doerr L. 1996. Interor-ganizational collaboration and the locus of innovation:networks of learning in biotechnology. AdministrativeScience Quarterly 41: 116–145.

Prahalad CK, Hamel G. 1990. The core competenceof the corporation. Harvard Business Review 68(3):79–91.

Rivkin JW, Porter ME. 1999. Matching Dell. HarvardBusiness School Case 9-799-158.

Rothaermel FT. 2001. Incumbent’s advantage throughexploiting complementary assets via interfirm coop-eration. Strategic Management Journal , Special Issue22(6–7): 687–699.

Rothaermel FT, Deeds DL. 2004. Exploration andexploitation alliances in biotechnology: a systemof new product development. Strategic ManagementJournal 25(3): 201–221.

Rumelt RP. 1974. Strategy, Structure and EconomicPerformance. Harvard Business School Press: Boston,MA.

Sanderson S, Uzumeri M. 1995. Managing productfamilies: the case of Sony Walkman. Research Policy24: 761–782.

Schumpeter JA. 1942. Capitalism, Socialism, and Democ-racy. Harper & Row: New York.

Sirmon DG, Hitt MA, Ireland RD. 2007. Managing firmresources in dynamic environments to create value:looking inside the black box. Academy of ManagementReview 32: in press.

Sørensen JB, Stuart TE. 2000. Aging, obsolescence,and organizational innovation. Administrative ScienceQuarterly 45: 81–112.

Spencer JW. 2003. Firms’ knowledge-sharing strategiesin the global innovation system: empirical evidencefrom the flat panel display industry. StrategicManagement Journal 24(3): 217–233.

Steensma HK, Corley KG. 2001. On the performanceof technology sourcing partnerships: the interactionbetween partner interdependence and technologyattributes. Academy of Management Journal 43:1045–1067.

Stinchcombe AL. 1965. Social structure in organizations.In Handbook of Organizations , March JG (ed). RandMcNally: Chicago, IL; 142–193.

Stuart TE. 2000. Interorganizational alliances and theperformance of firms: a study of growth andinnovation in a high-technology industry. StrategicManagement Journal 21(8): 791–811.

Tabachnick BG, Fidell LS. 2001. Using MultivariateStatistics (4th edn). Allyn & Bacon: Boston, MA.

Teece DJ. 1986. Profiting from technological innovation:implications for integration, collaboration, licensingand public policy. Research Policy 15: 285–305.

Teece DJ. 1992. Competition, cooperation, and innova-tion: organizational arrangements for regimes of rapidtechnological progress. Journal of Economic Behaviorand Organization 18: 1–25.

Teece DJ, Pisano G, Shuen A. 1997. Dynamic capabili-ties and strategic management. Strategic ManagementJournal 18(7): 509–533.

Thompson JD. 1967. Organizations in Action . McGrawHill: New York.

Walker G, Weber D. 1984. A transaction cost approachto the make–buy decisions. Administrative ScienceQuarterly 29: 373–391.

Williamson OE. 1975. Markets and Hierarchies . FreePress: New York.

Williamson OE. 1985. The Economic Institutions ofCapitalism . Free Press: New York.

APPENDIX: RELIABILITY, VALIDITY,AND ROBUSTNESS CHECKS

In designing this study, we took several actionsto ensure unbiased and objective coding of thedata proxying different organizational forms (ver-tical integration, strategic outsourcing, and M&Aactivity), because they pertain to a firm’s arrayof sourcing options. Moreover, vertical integrationand strategic outsourcing are the key theoreticalconstructs of this study. In total, we employeda team of five research assistants (all graduatestudents) to code the qualitative data. First, theresearch assistants were not aware of the theoryor hypotheses to be tested. Second, we used twoindependent data sources for each of the key inde-pendent variables. Third, the physical coding ofthe data from the two independent sources usedfor each measure was conducted at two differentinstitutions, and the research assistants were nei-ther in direct contact with one another nor did theyknow of each other.

In addition, we explored the robustness of theresults described above in several unreportedanalyses:

• We employed structural equation modeling(SEM). The results of SEM indicated statisticalsupport for most of the individual hypotheses(in the expected direction), but the overallfit of the structural equation model was notsatisfactory because of the small cross-sectionalsample size (N = 123 firms). SEM draws on



the bivariate correlation matrix; therefore, itdoes not leverage the econometric advantagesof longitudinal panel data (Hsaio, 2003).Tabachnick and Fidell (2001:659) describethe problem of estimating structural equationmodels based on a relatively small sampleas follows: ‘Covariances, like correlations, areless stable when estimated from small samples.Structural equation modeling (SEM) is based oncovariances. Parameter estimates and chi-squaretests of fit are also very sensitive to sample size.SEM, then, like factor analysis, is a large-sampletechnique . . ..’ Thus, applying GLS regressionanalysis is the preferred method here because itenables us to draw on a much larger sample(N = 492), and to access the advantages ofpanel data.

• To assess if the results were biased by M&Aactivity, we eliminated all cases where mergersor acquisitions occurred in which the researchand development for new microcomputer prod-ucts may have been conducted prior to themerger in the firm that did not survive, and thusmay bias the surviving firm’s post-merger newproduct portfolios and performance scores. Wereran the analysis on the revised sample.

• Because the equity acquired in a partner canspan the entire spectrum from minority to major-ity investments all the way to outright purchaseor merger, we also constructed the variableMergers & Acquisitions (majority), where theacquired firm purchased more than 50% of thetarget company. This variable captures the qual-itatively somewhat stronger M&A activity. Wereran the analysis using the revised variable.

• We eliminated all firms that either entered orexited the microcomputer industry during our

study period and reran the analysis on therevised sample.

• We added the 148 product reviews evaluatingservers/workstations and handhelds to the 3,441product evaluations for desktops and laptops,and reran the analysis including all 3,559 prod-uct evaluations.

• We varied the time window of the study periodby shortening it 1 year, and conducted newanalyses using the different time windows. Thus,we were able to assess the robustness of theresults on a moving 3-year window.

• We assessed the validity of revenues as a proxyof firm performance by applying return on assetson the sub-sample of firms for which this mea-sure was available.

• We reran the regression models on the sub-sample of firms for which employee data weredirectly available.

In all these additional analyses, the results wereconsistent with the ones presented above.

In a post hoc analysis, we assessed the inter-action effects of vertical integration and strategicoutsourcing quasi as well as those between verti-cal integration and strategic outsourcing taper onnew product success and firm performance. Wefound that the interactions were generally negativeand statistically significant. Recall that the inter-actions between vertical integration and strategicoutsourcing (quasi and taper) were positive. Takentogether, these results further highlight the impor-tance of balancing vertical integration and strategicoutsourcing in the pursuit of new product successand firm performance.


Documents

Balancing vertical integration and strategic outsourcing ... · BALANCING VERTICAL INTEGRATION AND ... forward integrated but rely on outsiders for a por- ... Balancing Vertical Integration